fix(ci): build gitnexus-shared before npm ci in publish workflow

The publish job failed because npm ci triggers the prepare script (tsc)
before gitnexus-shared types are available. Add the same build step
that the setup-gitnexus composite action uses in CI.

.claude-plugin/marketplace.json

@@ -0,0 +1,19 @@
+{
+  "name": "gitnexus-marketplace",
+  "owner": {
+    "name": "GitNexus",
+    "email": "nico@gitnexus.dev"
+  },
+  "metadata": {
+    "description": "Code intelligence powered by a knowledge graph — execution flows, blast radius, and semantic search",
+    "homepage": "https://github.com/nicosxt/gitnexus"
+  },
+  "plugins": [
+    {
+      "name": "gitnexus",
+      "version": "1.3.3",
+      "source": "./gitnexus-claude-plugin",
+      "description": "Code intelligence powered by a knowledge graph. Provides execution flow tracing, blast radius analysis, and augmented search across your codebase."
+    }
+  ]
+}

.claude/skills/gitnexus/gitnexus-cli/SKILL.md

@@ -0,0 +1,82 @@
+---
+name: gitnexus-cli
+description: "Use when the user needs to run GitNexus CLI commands like analyze/index a repo, check status, clean the index, generate a wiki, or list indexed repos. Examples: \"Index this repo\", \"Reanalyze the codebase\", \"Generate a wiki\""
+---
+
+# GitNexus CLI Commands
+
+All commands work via `npx` — no global install required.
+
+## Commands
+
+### analyze — Build or refresh the index
+
+```bash
+npx gitnexus analyze
+```
+
+Run from the project root. This parses all source files, builds the knowledge graph, writes it to `.gitnexus/`, and generates CLAUDE.md / AGENTS.md context files.
+
+| Flag           | Effect                                                           |
+| -------------- | ---------------------------------------------------------------- |
+| `--force`      | Force full re-index even if up to date                           |
+| `--embeddings` | Enable embedding generation for semantic search (off by default) |
+
+**When to run:** First time in a project, after major code changes, or when `gitnexus://repo/{name}/context` reports the index is stale. In Claude Code, a PostToolUse hook runs `analyze` automatically after `git commit` and `git merge`, preserving embeddings if previously generated.
+
+### status — Check index freshness
+
+```bash
+npx gitnexus status
+```
+
+Shows whether the current repo has a GitNexus index, when it was last updated, and symbol/relationship counts. Use this to check if re-indexing is needed.
+
+### clean — Delete the index
+
+```bash
+npx gitnexus clean
+```
+
+Deletes the `.gitnexus/` directory and unregisters the repo from the global registry. Use before re-indexing if the index is corrupt or after removing GitNexus from a project.
+
+| Flag      | Effect                                            |
+| --------- | ------------------------------------------------- |
+| `--force` | Skip confirmation prompt                          |
+| `--all`   | Clean all indexed repos, not just the current one |
+
+### wiki — Generate documentation from the graph
+
+```bash
+npx gitnexus wiki
+```
+
+Generates repository documentation from the knowledge graph using an LLM. Requires an API key (saved to `~/.gitnexus/config.json` on first use).
+
+| Flag                | Effect                                    |
+| ------------------- | ----------------------------------------- |
+| `--force`           | Force full regeneration                   |
+| `--model <model>`   | LLM model (default: minimax/minimax-m2.5) |
+| `--base-url <url>`  | LLM API base URL                          |
+| `--api-key <key>`   | LLM API key                               |
+| `--concurrency <n>` | Parallel LLM calls (default: 3)           |
+| `--gist`            | Publish wiki as a public GitHub Gist      |
+
+### list — Show all indexed repos
+
+```bash
+npx gitnexus list
+```
+
+Lists all repositories registered in `~/.gitnexus/registry.json`. The MCP `list_repos` tool provides the same information.
+
+## After Indexing
+
+1. **Read `gitnexus://repo/{name}/context`** to verify the index loaded
+2. Use the other GitNexus skills (`exploring`, `debugging`, `impact-analysis`, `refactoring`) for your task
+
+## Troubleshooting
+
+- **"Not inside a git repository"**: Run from a directory inside a git repo
+- **Index is stale after re-analyzing**: Restart Claude Code to reload the MCP server
+- **Embeddings slow**: Omit `--embeddings` (it's off by default) or set `OPENAI_API_KEY` for faster API-based embedding

.claude/skills/gitnexus/gitnexus-debugging/SKILL.md

@@ -0,0 +1,89 @@
+---
+name: gitnexus-debugging
+description: "Use when the user is debugging a bug, tracing an error, or asking why something fails. Examples: \"Why is X failing?\", \"Where does this error come from?\", \"Trace this bug\""
+---
+
+# Debugging with GitNexus
+
+## When to Use
+
+- "Why is this function failing?"
+- "Trace where this error comes from"
+- "Who calls this method?"
+- "This endpoint returns 500"
+- Investigating bugs, errors, or unexpected behavior
+
+## Workflow
+
+```
+1. gitnexus_query({query: "<error or symptom>"})            → Find related execution flows
+2. gitnexus_context({name: "<suspect>"})                    → See callers/callees/processes
+3. READ gitnexus://repo/{name}/process/{name}                → Trace execution flow
+4. gitnexus_cypher({query: "MATCH path..."})                 → Custom traces if needed
+```
+
+> If "Index is stale" → run `npx gitnexus analyze` in terminal.
+
+## Checklist
+
+```
+- [ ] Understand the symptom (error message, unexpected behavior)
+- [ ] gitnexus_query for error text or related code
+- [ ] Identify the suspect function from returned processes
+- [ ] gitnexus_context to see callers and callees
+- [ ] Trace execution flow via process resource if applicable
+- [ ] gitnexus_cypher for custom call chain traces if needed
+- [ ] Read source files to confirm root cause
+```
+
+## Debugging Patterns
+
+| Symptom              | GitNexus Approach                                          |
+| -------------------- | ---------------------------------------------------------- |
+| Error message        | `gitnexus_query` for error text → `context` on throw sites |
+| Wrong return value   | `context` on the function → trace callees for data flow    |
+| Intermittent failure | `context` → look for external calls, async deps            |
+| Performance issue    | `context` → find symbols with many callers (hot paths)     |
+| Recent regression    | `detect_changes` to see what your changes affect           |
+
+## Tools
+
+**gitnexus_query** — find code related to error:
+
+```
+gitnexus_query({query: "payment validation error"})
+→ Processes: CheckoutFlow, ErrorHandling
+→ Symbols: validatePayment, handlePaymentError, PaymentException
+```
+
+**gitnexus_context** — full context for a suspect:
+
+```
+gitnexus_context({name: "validatePayment"})
+→ Incoming calls: processCheckout, webhookHandler
+→ Outgoing calls: verifyCard, fetchRates (external API!)
+→ Processes: CheckoutFlow (step 3/7)
+```
+
+**gitnexus_cypher** — custom call chain traces:
+
+```cypher
+MATCH path = (a)-[:CodeRelation {type: 'CALLS'}*1..2]->(b:Function {name: "validatePayment"})
+RETURN [n IN nodes(path) | n.name] AS chain
+```
+
+## Example: "Payment endpoint returns 500 intermittently"
+
+```
+1. gitnexus_query({query: "payment error handling"})
+   → Processes: CheckoutFlow, ErrorHandling
+   → Symbols: validatePayment, handlePaymentError
+
+2. gitnexus_context({name: "validatePayment"})
+   → Outgoing calls: verifyCard, fetchRates (external API!)
+
+3. READ gitnexus://repo/my-app/process/CheckoutFlow
+   → Step 3: validatePayment → calls fetchRates (external)
+
+4. Root cause: fetchRates calls external API without proper timeout
+```

.claude/skills/gitnexus/gitnexus-exploring/SKILL.md

@@ -0,0 +1,78 @@
+---
+name: gitnexus-exploring
+description: "Use when the user asks how code works, wants to understand architecture, trace execution flows, or explore unfamiliar parts of the codebase. Examples: \"How does X work?\", \"What calls this function?\", \"Show me the auth flow\""
+---
+
+# Exploring Codebases with GitNexus
+
+## When to Use
+
+- "How does authentication work?"
+- "What's the project structure?"
+- "Show me the main components"
+- "Where is the database logic?"
+- Understanding code you haven't seen before
+
+## Workflow
+
+```
+1. READ gitnexus://repos                          → Discover indexed repos
+2. READ gitnexus://repo/{name}/context             → Codebase overview, check staleness
+3. gitnexus_query({query: "<what you want to understand>"})  → Find related execution flows
+4. gitnexus_context({name: "<symbol>"})            → Deep dive on specific symbol
+5. READ gitnexus://repo/{name}/process/{name}      → Trace full execution flow
+```
+
+> If step 2 says "Index is stale" → run `npx gitnexus analyze` in terminal.
+
+## Checklist
+
+```
+- [ ] READ gitnexus://repo/{name}/context
+- [ ] gitnexus_query for the concept you want to understand
+- [ ] Review returned processes (execution flows)
+- [ ] gitnexus_context on key symbols for callers/callees
+- [ ] READ process resource for full execution traces
+- [ ] Read source files for implementation details
+```
+
+## Resources
+
+| Resource                                | What you get                                            |
+| --------------------------------------- | ------------------------------------------------------- |
+| `gitnexus://repo/{name}/context`        | Stats, staleness warning (~150 tokens)                  |
+| `gitnexus://repo/{name}/clusters`       | All functional areas with cohesion scores (~300 tokens) |
+| `gitnexus://repo/{name}/cluster/{name}` | Area members with file paths (~500 tokens)              |
+| `gitnexus://repo/{name}/process/{name}` | Step-by-step execution trace (~200 tokens)              |
+
+## Tools
+
+**gitnexus_query** — find execution flows related to a concept:
+
+```
+gitnexus_query({query: "payment processing"})
+→ Processes: CheckoutFlow, RefundFlow, WebhookHandler
+→ Symbols grouped by flow with file locations
+```
+
+**gitnexus_context** — 360-degree view of a symbol:
+
+```
+gitnexus_context({name: "validateUser"})
+→ Incoming calls: loginHandler, apiMiddleware
+→ Outgoing calls: checkToken, getUserById
+→ Processes: LoginFlow (step 2/5), TokenRefresh (step 1/3)
+```
+
+## Example: "How does payment processing work?"
+
+```
+1. READ gitnexus://repo/my-app/context       → 918 symbols, 45 processes
+2. gitnexus_query({query: "payment processing"})
+   → CheckoutFlow: processPayment → validateCard → chargeStripe
+   → RefundFlow: initiateRefund → calculateRefund → processRefund
+3. gitnexus_context({name: "processPayment"})
+   → Incoming: checkoutHandler, webhookHandler
+   → Outgoing: validateCard, chargeStripe, saveTransaction
+4. Read src/payments/processor.ts for implementation details
+```

.claude/skills/gitnexus/gitnexus-guide/SKILL.md

@@ -0,0 +1,64 @@
+---
+name: gitnexus-guide
+description: "Use when the user asks about GitNexus itself — available tools, how to query the knowledge graph, MCP resources, graph schema, or workflow reference. Examples: \"What GitNexus tools are available?\", \"How do I use GitNexus?\""
+---
+
+# GitNexus Guide
+
+Quick reference for all GitNexus MCP tools, resources, and the knowledge graph schema.
+
+## Always Start Here
+
+For any task involving code understanding, debugging, impact analysis, or refactoring:
+
+1. **Read `gitnexus://repo/{name}/context`** — codebase overview + check index freshness
+2. **Match your task to a skill below** and **read that skill file**
+3. **Follow the skill's workflow and checklist**
+
+> If step 1 warns the index is stale, run `npx gitnexus analyze` in the terminal first.
+
+## Skills
+
+| Task                                         | Skill to read       |
+| -------------------------------------------- | ------------------- |
+| Understand architecture / "How does X work?" | `gitnexus-exploring`         |
+| Blast radius / "What breaks if I change X?"  | `gitnexus-impact-analysis`   |
+| Trace bugs / "Why is X failing?"             | `gitnexus-debugging`         |
+| Rename / extract / split / refactor          | `gitnexus-refactoring`       |
+| Tools, resources, schema reference           | `gitnexus-guide` (this file) |
+| Index, status, clean, wiki CLI commands      | `gitnexus-cli`               |
+
+## Tools Reference
+
+| Tool             | What it gives you                                                        |
+| ---------------- | ------------------------------------------------------------------------ |
+| `query`          | Process-grouped code intelligence — execution flows related to a concept |
+| `context`        | 360-degree symbol view — categorized refs, processes it participates in  |
+| `impact`         | Symbol blast radius — what breaks at depth 1/2/3 with confidence         |
+| `detect_changes` | Git-diff impact — what do your current changes affect                    |
+| `rename`         | Multi-file coordinated rename with confidence-tagged edits               |
+| `cypher`         | Raw graph queries (read `gitnexus://repo/{name}/schema` first)           |
+| `list_repos`     | Discover indexed repos                                                   |
+
+## Resources Reference
+
+Lightweight reads (~100-500 tokens) for navigation:
+
+| Resource                                       | Content                                   |
+| ---------------------------------------------- | ----------------------------------------- |
+| `gitnexus://repo/{name}/context`               | Stats, staleness check                    |
+| `gitnexus://repo/{name}/clusters`              | All functional areas with cohesion scores |
+| `gitnexus://repo/{name}/cluster/{clusterName}` | Area members                              |
+| `gitnexus://repo/{name}/processes`             | All execution flows                       |
+| `gitnexus://repo/{name}/process/{processName}` | Step-by-step trace                        |
+| `gitnexus://repo/{name}/schema`                | Graph schema for Cypher                   |
+
+## Graph Schema
+
+**Nodes:** File, Function, Class, Interface, Method, Community, Process
+**Edges (via CodeRelation.type):** CALLS, IMPORTS, EXTENDS, IMPLEMENTS, DEFINES, MEMBER_OF, STEP_IN_PROCESS
+
+```cypher
+MATCH (caller)-[:CodeRelation {type: 'CALLS'}]->(f:Function {name: "myFunc"})
+RETURN caller.name, caller.filePath
+```

.claude/skills/gitnexus/gitnexus-impact-analysis/SKILL.md

@@ -0,0 +1,97 @@
+---
+name: gitnexus-impact-analysis
+description: "Use when the user wants to know what will break if they change something, or needs safety analysis before editing code. Examples: \"Is it safe to change X?\", \"What depends on this?\", \"What will break?\""
+---
+
+# Impact Analysis with GitNexus
+
+## When to Use
+
+- "Is it safe to change this function?"
+- "What will break if I modify X?"
+- "Show me the blast radius"
+- "Who uses this code?"
+- Before making non-trivial code changes
+- Before committing — to understand what your changes affect
+
+## Workflow
+
+```
+1. gitnexus_impact({target: "X", direction: "upstream"})  → What depends on this
+2. READ gitnexus://repo/{name}/processes                   → Check affected execution flows
+3. gitnexus_detect_changes()                               → Map current git changes to affected flows
+4. Assess risk and report to user
+```
+
+> If "Index is stale" → run `npx gitnexus analyze` in terminal.
+
+## Checklist
+
+```
+- [ ] gitnexus_impact({target, direction: "upstream"}) to find dependents
+- [ ] Review d=1 items first (these WILL BREAK)
+- [ ] Check high-confidence (>0.8) dependencies
+- [ ] READ processes to check affected execution flows
+- [ ] gitnexus_detect_changes() for pre-commit check
+- [ ] Assess risk level and report to user
+```
+
+## Understanding Output
+
+| Depth | Risk Level       | Meaning                  |
+| ----- | ---------------- | ------------------------ |
+| d=1   | **WILL BREAK**   | Direct callers/importers |
+| d=2   | LIKELY AFFECTED  | Indirect dependencies    |
+| d=3   | MAY NEED TESTING | Transitive effects       |
+
+## Risk Assessment
+
+| Affected                       | Risk     |
+| ------------------------------ | -------- |
+| <5 symbols, few processes      | LOW      |
+| 5-15 symbols, 2-5 processes    | MEDIUM   |
+| >15 symbols or many processes  | HIGH     |
+| Critical path (auth, payments) | CRITICAL |
+
+## Tools
+
+**gitnexus_impact** — the primary tool for symbol blast radius:
+
+```
+gitnexus_impact({
+  target: "validateUser",
+  direction: "upstream",
+  minConfidence: 0.8,
+  maxDepth: 3
+})
+
+→ d=1 (WILL BREAK):
+  - loginHandler (src/auth/login.ts:42) [CALLS, 100%]
+  - apiMiddleware (src/api/middleware.ts:15) [CALLS, 100%]
+
+→ d=2 (LIKELY AFFECTED):
+  - authRouter (src/routes/auth.ts:22) [CALLS, 95%]
+```
+
+**gitnexus_detect_changes** — git-diff based impact analysis:
+
+```
+gitnexus_detect_changes({scope: "staged"})
+
+→ Changed: 5 symbols in 3 files
+→ Affected: LoginFlow, TokenRefresh, APIMiddlewarePipeline
+→ Risk: MEDIUM
+```
+
+## Example: "What breaks if I change validateUser?"
+
+```
+1. gitnexus_impact({target: "validateUser", direction: "upstream"})
+   → d=1: loginHandler, apiMiddleware (WILL BREAK)
+   → d=2: authRouter, sessionManager (LIKELY AFFECTED)
+
+2. READ gitnexus://repo/my-app/processes
+   → LoginFlow and TokenRefresh touch validateUser
+
+3. Risk: 2 direct callers, 2 processes = MEDIUM
+```

.claude/skills/gitnexus/gitnexus-pr-review/SKILL.md

@@ -0,0 +1,163 @@
+---
+name: gitnexus-pr-review
+description: "Use when the user wants to review a pull request, understand what a PR changes, assess risk of merging, or check for missing test coverage. Examples: \"Review this PR\", \"What does PR #42 change?\", \"Is this PR safe to merge?\""
+---
+
+# PR Review with GitNexus
+
+## When to Use
+
+- "Review this PR"
+- "What does PR #42 change?"
+- "Is this safe to merge?"
+- "What's the blast radius of this PR?"
+- "Are there missing tests for this PR?"
+- Reviewing someone else's code changes before merge
+
+## Workflow
+
+```
+1. gh pr diff <number>                                    → Get the raw diff
+2. gitnexus_detect_changes({scope: "compare", base_ref: "main"})  → Map diff to affected flows
+3. For each changed symbol:
+   gitnexus_impact({target: "<symbol>", direction: "upstream"})    → Blast radius per change
+4. gitnexus_context({name: "<key symbol>"})               → Understand callers/callees
+5. READ gitnexus://repo/{name}/processes                   → Check affected execution flows
+6. Summarize findings with risk assessment
+```
+
+> If "Index is stale" → run `npx gitnexus analyze` in terminal before reviewing.
+
+## Checklist
+
+```
+- [ ] Fetch PR diff (gh pr diff or git diff base...head)
+- [ ] gitnexus_detect_changes to map changes to affected execution flows
+- [ ] gitnexus_impact on each non-trivial changed symbol
+- [ ] Review d=1 items (WILL BREAK) — are callers updated?
+- [ ] gitnexus_context on key changed symbols to understand full picture
+- [ ] Check if affected processes have test coverage
+- [ ] Assess overall risk level
+- [ ] Write review summary with findings
+```
+
+## Review Dimensions
+
+| Dimension | How GitNexus Helps |
+| --- | --- |
+| **Correctness** | `context` shows callers — are they all compatible with the change? |
+| **Blast radius** | `impact` shows d=1/d=2/d=3 dependents — anything missed? |
+| **Completeness** | `detect_changes` shows all affected flows — are they all handled? |
+| **Test coverage** | `impact({includeTests: true})` shows which tests touch changed code |
+| **Breaking changes** | d=1 upstream items that aren't updated in the PR = potential breakage |
+
+## Risk Assessment
+
+| Signal | Risk |
+| --- | --- |
+| Changes touch <3 symbols, 0-1 processes | LOW |
+| Changes touch 3-10 symbols, 2-5 processes | MEDIUM |
+| Changes touch >10 symbols or many processes | HIGH |
+| Changes touch auth, payments, or data integrity code | CRITICAL |
+| d=1 callers exist outside the PR diff | Potential breakage — flag it |
+
+## Tools
+
+**gitnexus_detect_changes** — map PR diff to affected execution flows:
+
+```
+gitnexus_detect_changes({scope: "compare", base_ref: "main"})
+
+→ Changed: 8 symbols in 4 files
+→ Affected processes: CheckoutFlow, RefundFlow, WebhookHandler
+→ Risk: MEDIUM
+```
+
+**gitnexus_impact** — blast radius per changed symbol:
+
+```
+gitnexus_impact({target: "validatePayment", direction: "upstream"})
+
+→ d=1 (WILL BREAK):
+  - processCheckout (src/checkout.ts:42) [CALLS, 100%]
+  - webhookHandler (src/webhooks.ts:15) [CALLS, 100%]
+
+→ d=2 (LIKELY AFFECTED):
+  - checkoutRouter (src/routes/checkout.ts:22) [CALLS, 95%]
+```
+
+**gitnexus_impact with tests** — check test coverage:
+
+```
+gitnexus_impact({target: "validatePayment", direction: "upstream", includeTests: true})
+
+→ Tests that cover this symbol:
+  - validatePayment.test.ts [direct]
+  - checkout.integration.test.ts [via processCheckout]
+```
+
+**gitnexus_context** — understand a changed symbol's role:
+
+```
+gitnexus_context({name: "validatePayment"})
+
+→ Incoming calls: processCheckout, webhookHandler
+→ Outgoing calls: verifyCard, fetchRates
+→ Processes: CheckoutFlow (step 3/7), RefundFlow (step 1/5)
+```
+
+## Example: "Review PR #42"
+
+```
+1. gh pr diff 42 > /tmp/pr42.diff
+   → 4 files changed: payments.ts, checkout.ts, types.ts, utils.ts
+
+2. gitnexus_detect_changes({scope: "compare", base_ref: "main"})
+   → Changed symbols: validatePayment, PaymentInput, formatAmount
+   → Affected processes: CheckoutFlow, RefundFlow
+   → Risk: MEDIUM
+
+3. gitnexus_impact({target: "validatePayment", direction: "upstream"})
+   → d=1: processCheckout, webhookHandler (WILL BREAK)
+   → webhookHandler is NOT in the PR diff — potential breakage!
+
+4. gitnexus_impact({target: "PaymentInput", direction: "upstream"})
+   → d=1: validatePayment (in PR), createPayment (NOT in PR)
+   → createPayment uses the old PaymentInput shape — breaking change!
+
+5. gitnexus_context({name: "formatAmount"})
+   → Called by 12 functions — but change is backwards-compatible (added optional param)
+
+6. Review summary:
+   - MEDIUM risk — 3 changed symbols affect 2 execution flows
+   - BUG: webhookHandler calls validatePayment but isn't updated for new signature
+   - BUG: createPayment depends on PaymentInput type which changed
+   - OK: formatAmount change is backwards-compatible
+   - Tests: checkout.test.ts covers processCheckout path, but no webhook test
+```
+
+## Review Output Format
+
+Structure your review as:
+
+```markdown
+## PR Review: <title>
+
+**Risk: LOW / MEDIUM / HIGH / CRITICAL**
+
+### Changes Summary
+- <N> symbols changed across <M> files
+- <P> execution flows affected
+
+### Findings
+1. **[severity]** Description of finding
+   - Evidence from GitNexus tools
+   - Affected callers/flows
+
+### Missing Coverage
+- Callers not updated in PR: ...
+- Untested flows: ...
+
+### Recommendation
+APPROVE / REQUEST CHANGES / NEEDS DISCUSSION
+```

.claude/skills/gitnexus/gitnexus-refactoring/SKILL.md

@@ -0,0 +1,121 @@
+---
+name: gitnexus-refactoring
+description: "Use when the user wants to rename, extract, split, move, or restructure code safely. Examples: \"Rename this function\", \"Extract this into a module\", \"Refactor this class\", \"Move this to a separate file\""
+---
+
+# Refactoring with GitNexus
+
+## When to Use
+
+- "Rename this function safely"
+- "Extract this into a module"
+- "Split this service"
+- "Move this to a new file"
+- Any task involving renaming, extracting, splitting, or restructuring code
+
+## Workflow
+
+```
+1. gitnexus_impact({target: "X", direction: "upstream"})  → Map all dependents
+2. gitnexus_query({query: "X"})                            → Find execution flows involving X
+3. gitnexus_context({name: "X"})                           → See all incoming/outgoing refs
+4. Plan update order: interfaces → implementations → callers → tests
+```
+
+> If "Index is stale" → run `npx gitnexus analyze` in terminal.
+
+## Checklists
+
+### Rename Symbol
+
+```
+- [ ] gitnexus_rename({symbol_name: "oldName", new_name: "newName", dry_run: true}) — preview all edits
+- [ ] Review graph edits (high confidence) and ast_search edits (review carefully)
+- [ ] If satisfied: gitnexus_rename({..., dry_run: false}) — apply edits
+- [ ] gitnexus_detect_changes() — verify only expected files changed
+- [ ] Run tests for affected processes
+```
+
+### Extract Module
+
+```
+- [ ] gitnexus_context({name: target}) — see all incoming/outgoing refs
+- [ ] gitnexus_impact({target, direction: "upstream"}) — find all external callers
+- [ ] Define new module interface
+- [ ] Extract code, update imports
+- [ ] gitnexus_detect_changes() — verify affected scope
+- [ ] Run tests for affected processes
+```
+
+### Split Function/Service
+
+```
+- [ ] gitnexus_context({name: target}) — understand all callees
+- [ ] Group callees by responsibility
+- [ ] gitnexus_impact({target, direction: "upstream"}) — map callers to update
+- [ ] Create new functions/services
+- [ ] Update callers
+- [ ] gitnexus_detect_changes() — verify affected scope
+- [ ] Run tests for affected processes
+```
+
+## Tools
+
+**gitnexus_rename** — automated multi-file rename:
+
+```
+gitnexus_rename({symbol_name: "validateUser", new_name: "authenticateUser", dry_run: true})
+→ 12 edits across 8 files
+→ 10 graph edits (high confidence), 2 ast_search edits (review)
+→ Changes: [{file_path, edits: [{line, old_text, new_text, confidence}]}]
+```
+
+**gitnexus_impact** — map all dependents first:
+
+```
+gitnexus_impact({target: "validateUser", direction: "upstream"})
+→ d=1: loginHandler, apiMiddleware, testUtils
+→ Affected Processes: LoginFlow, TokenRefresh
+```
+
+**gitnexus_detect_changes** — verify your changes after refactoring:
+
+```
+gitnexus_detect_changes({scope: "all"})
+→ Changed: 8 files, 12 symbols
+→ Affected processes: LoginFlow, TokenRefresh
+→ Risk: MEDIUM
+```
+
+**gitnexus_cypher** — custom reference queries:
+
+```cypher
+MATCH (caller)-[:CodeRelation {type: 'CALLS'}]->(f:Function {name: "validateUser"})
+RETURN caller.name, caller.filePath ORDER BY caller.filePath
+```
+
+## Risk Rules
+
+| Risk Factor         | Mitigation                                |
+| ------------------- | ----------------------------------------- |
+| Many callers (>5)   | Use gitnexus_rename for automated updates |
+| Cross-area refs     | Use detect_changes after to verify scope  |
+| String/dynamic refs | gitnexus_query to find them               |
+| External/public API | Version and deprecate properly            |
+
+## Example: Rename `validateUser` to `authenticateUser`
+
+```
+1. gitnexus_rename({symbol_name: "validateUser", new_name: "authenticateUser", dry_run: true})
+   → 12 edits: 10 graph (safe), 2 ast_search (review)
+   → Files: validator.ts, login.ts, middleware.ts, config.json...
+
+2. Review ast_search edits (config.json: dynamic reference!)
+
+3. gitnexus_rename({symbol_name: "validateUser", new_name: "authenticateUser", dry_run: false})
+   → Applied 12 edits across 8 files
+
+4. gitnexus_detect_changes({scope: "all"})
+   → Affected: LoginFlow, TokenRefresh
+   → Risk: MEDIUM — run tests for these flows
+```

.cursor/index.mdc

@@ -0,0 +1,21 @@
+---
+alwaysApply: true
+---
+
+# GitNexus — Cursor project rules
+
+Last reviewed: 2026-03-24
+
+Canonical agent instructions: **[AGENTS.md](../AGENTS.md)** (GitNexus MCP rules, monorepo commands, Cursor Cloud notes). **[CLAUDE.md](../CLAUDE.md)** adds Claude Code-specific notes and points back to AGENTS.md for GitNexus.
+
+## Non-negotiables (always apply)
+
+- NEVER edit a function/class/method without running `gitnexus_impact` first.
+- NEVER rename symbols with find-and-replace — use `gitnexus_rename`.
+- NEVER commit without running `gitnexus_detect_changes()`.
+- NEVER ignore HIGH/CRITICAL risk warnings from impact analysis.
+- NEVER run `npx gitnexus analyze` without `--embeddings` if `.gitnexus/meta.json` shows stored embeddings.
+
+Full rules: **[AGENTS.md](../AGENTS.md)** (`gitnexus:start` block, Cursor Cloud section).
+
+**Rule architecture:** Prefer this file plus optional `.cursor/rules/*.mdc` globs (YAML `globs` in frontmatter). Legacy `.cursorrules` is deprecated; content lives here.

.cursor/rules/100-monorepo.mdc

@@ -0,0 +1,12 @@
+---
+globs:
+  - "gitnexus/**"
+  - "gitnexus-web/**"
+---
+
+# GitNexus build/test quick refs
+
+- CLI (`gitnexus/`): `npm test`; `npm run test:integration`; `npx tsc --noEmit`.
+- Web (`gitnexus-web/`): `npm test`; `npm run dev`; `npx tsc -b --noEmit`; `E2E=1 npx playwright test` (needs servers).
+- `npm install` in `gitnexus/` runs `prepare` (tsc build) and `postinstall` (tree-sitter patches); needs `python3`, `make`, `g++`.
+- LadybugDB locking tests may fail in containerized environments because of `/tmp` file locks (known issue, not a code bug).

.cursor/rules/200-eval.mdc

@@ -0,0 +1,14 @@
+---
+globs:
+  - "eval/**"
+---
+
+# GitNexus eval harness (Python)
+
+- **Run tests**: `cd eval && uv run pytest tests/`
+- **Run with coverage**: `cd eval && uv run coverage run -m pytest tests/ && uv run coverage report`
+- **Lint**: `cd eval && uv run ruff check .`
+- **Run eval**: `cd eval && uv run python run_eval.py --config configs/<config>.yaml`
+- Shared constants live in `eval/constants.py`; tool specs in `eval/tool_registry.py`.
+- Error logging uses `utils/errors.py` — set `GITNEXUS_EVAL_DEBUG=1` for full tracebacks.
+- Property-based tests use Hypothesis (`eval/tests/test_property_based.py`).

.cursorrules

@@ -0,0 +1,5 @@
+# Deprecated for Cursor Agent Mode
+
+Use **`.cursor/index.mdc`** (`alwaysApply: true`) for project rules. See [AGENTS.md](AGENTS.md).
+
+This file is kept only as a breadcrumb for older workflows.

.git-blame-ignore-revs

@@ -0,0 +1,5 @@
+# Prettier initial formatting (2026-03-28)
+afcc3d1523f99c77ff67c4fd1af12334660113f6
+
+# ESLint unused import removal (2026-03-28)
+1491826bc8da5436d3b1eb092d9274f2c9f028a7

.gitattributes

@@ -0,0 +1,2 @@
+* text=auto eol=lf
+.husky/* text eol=lf

.github/FUNDING.yml

@@ -0,0 +1,3 @@
+# These are supported funding model platforms
+
+github: abhigyanpatwari

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -0,0 +1,86 @@
+name: Bug report
+description: Report unexpected behavior or a regression
+labels: [bug]
+body:
+  - type: markdown
+    attributes:
+      value: |
+        **Goal:** capture enough context to reproduce and fix the issue quickly.
+        Use **one issue per bug**; split unrelated problems.
+
+  - type: dropdown
+    id: area
+    attributes:
+      label: Area
+      description: Where does the problem show up?
+      options:
+        - gitnexus (CLI / core / indexing / MCP server)
+        - gitnexus-web (browser UI / WASM / workers)
+        - CI / GitHub Actions
+        - Documentation / developer experience
+        - Other
+    validations:
+      required: true
+
+  - type: textarea
+    id: summary
+    attributes:
+      label: Summary
+      description: One sentence — what went wrong?
+    validations:
+      required: true
+
+  - type: textarea
+    id: context
+    attributes:
+      label: Context
+      description: What were you trying to do? Any relevant links, PRs, or commits?
+    validations:
+      required: false
+
+  - type: textarea
+    id: expected
+    attributes:
+      label: Expected behavior
+    validations:
+      required: true
+
+  - type: textarea
+    id: actual
+    attributes:
+      label: Actual behavior
+    validations:
+      required: true
+
+  - type: textarea
+    id: reproduce
+    attributes:
+      label: Steps to reproduce
+      description: Ordered steps, sample repo or minimal case, commands run.
+      placeholder: |
+        1. …
+        2. …
+        3. …
+    validations:
+      required: true
+
+  - type: textarea
+    id: environment
+    attributes:
+      label: Environment
+      description: OS, Node version, browser (if web), GitNexus version or commit SHA.
+      placeholder: |
+        - OS:
+        - Node:
+        - Browser (if applicable):
+        - Commit / version:
+    validations:
+      required: false
+
+  - type: textarea
+    id: logs
+    attributes:
+      label: Logs / screenshots
+      description: Paste errors, stack traces, or attach screenshots (redact secrets).
+    validations:
+      required: false

.github/ISSUE_TEMPLATE/config.yml

@@ -0,0 +1 @@
+blank_issues_enabled: true

.github/ISSUE_TEMPLATE/feature_request.yml

@@ -0,0 +1,74 @@
+name: Feature request
+description: Propose a new capability or improvement
+labels: [enhancement]
+body:
+  - type: markdown
+    attributes:
+      value: |
+        **Goal:** describe the problem and desired outcome so maintainers can size and prioritize.
+        Prefer **small, shippable** requests; split large ideas into phases.
+
+  - type: dropdown
+    id: area
+    attributes:
+      label: Area
+      description: Primary part of the monorepo this relates to.
+      options:
+        - gitnexus (CLI / core / indexing / MCP server)
+        - gitnexus-web (browser UI / WASM / workers)
+        - CI / release / packaging
+        - Documentation / developer experience
+        - Other
+    validations:
+      required: true
+
+  - type: textarea
+    id: problem
+    attributes:
+      label: Problem or opportunity
+      description: What pain point or gap exists today?
+    validations:
+      required: true
+
+  - type: textarea
+    id: proposal
+    attributes:
+      label: Proposed solution
+      description: What should happen instead? User-visible behavior, APIs, or UX.
+    validations:
+      required: true
+
+  - type: textarea
+    id: alternatives
+    attributes:
+      label: Alternatives considered
+      description: Other approaches you considered and why this one is preferred.
+    validations:
+      required: false
+
+  - type: textarea
+    id: acceptance
+    attributes:
+      label: Acceptance criteria
+      description: Testable conditions for “done” (bullets or checkboxes in prose).
+      placeholder: |
+        - When … then …
+        - Documentation / tests updated where appropriate
+    validations:
+      required: false
+
+  - type: textarea
+    id: constraints
+    attributes:
+      label: Constraints
+      description: Compatibility, performance, security, or “must not change” boundaries.
+    validations:
+      required: false
+
+  - type: checkboxes
+    id: willing
+    attributes:
+      label: Contribution
+      options:
+        - label: I am willing to open a PR for this (may need design discussion first).
+          required: false

.github/PULL_REQUEST_TEMPLATE.md

@@ -0,0 +1,52 @@
+## Summary
+
+<!-- One or two sentences: what does this PR change? -->
+
+## Motivation / context
+
+<!-- Why is this change needed? Link issues, ADRs, or prior discussion. -->
+
+## Areas touched
+
+<!-- Check all that apply -->
+
+- [ ] `gitnexus/` (CLI / core / MCP server)
+- [ ] `gitnexus-web/` (Vite / React UI)
+- [ ] `.github/` (workflows, actions)
+- [ ] `eval/` or other tooling
+- [ ] Docs / agent config only (`AGENTS.md`, `CLAUDE.md`, `.cursor/`, `llms.txt`, etc.)
+
+## Scope & constraints
+
+**In scope**
+
+- <!-- bullets -->
+
+**Explicitly out of scope / not done here**
+
+- <!-- bullets — prevents reviewers assuming missing work is an oversight -->
+
+## Implementation notes
+
+<!-- Optional: design choices, tradeoffs, follow-ups -->
+
+## Testing & verification
+
+<!-- What you ran; paste commands. Omit sections that do not apply. -->
+
+- [ ] `cd gitnexus && npm test`
+- [ ] `cd gitnexus && npm run test:integration` *(if core/indexing/MCP paths changed)*
+- [ ] `cd gitnexus && npx tsc --noEmit`
+- [ ] `cd gitnexus-web && npm test` *(if web changed)*
+- [ ] `cd gitnexus-web && npx tsc -b --noEmit` *(if web changed)*
+- [ ] Manual / Playwright E2E *(note environment — see `gitnexus-web/e2e/`)*
+
+## Risk & rollout
+
+<!-- Breaking changes, migrations, index refresh (`npx gitnexus analyze`), release notes -->
+
+## Checklist
+
+- [ ] PR body meets repo minimum length (workflow may label short descriptions)
+- [ ] If `AGENTS.md` / overlays changed: headers, scope block, and changelog updated per project conventions
+- [ ] No secrets, tokens, or machine-specific paths committed

.github/actions/setup-gitnexus-web/action.yml

@@ -0,0 +1,21 @@
+name: Setup GitNexus Web
+description: Setup Node.js 20, build gitnexus-shared, install web dependencies
+
+runs:
+  using: composite
+  steps:
+    - uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020 # v4
+      with:
+        node-version: 20
+        cache: npm
+        cache-dependency-path: gitnexus-web/package-lock.json
+
+    - name: Build gitnexus-shared
+      run: npm install && npm run build
+      shell: bash
+      working-directory: gitnexus-shared
+
+    - name: Install web dependencies
+      run: npm ci
+      shell: bash
+      working-directory: gitnexus-web

.github/actions/setup-gitnexus/action.yml

@@ -0,0 +1,33 @@
+name: Setup GitNexus
+description: Setup Node.js 20, install dependencies, and optionally build
+
+inputs:
+  build:
+    description: Whether to run npm run build after install
+    required: false
+    default: 'false'
+
+runs:
+  using: composite
+  steps:
+    - uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020 # v4
+      with:
+        node-version: 20
+        cache: npm
+        cache-dependency-path: gitnexus/package-lock.json
+
+    - name: Build gitnexus-shared
+      run: npm install && npm run build
+      shell: bash
+      working-directory: gitnexus-shared
+
+    - name: Install dependencies
+      run: npm ci
+      shell: bash
+      working-directory: gitnexus
+
+    - name: Build
+      if: ${{ inputs.build == 'true' }}
+      run: npm run build
+      shell: bash
+      working-directory: gitnexus

.github/release.yml

@@ -0,0 +1,45 @@
+changelog:
+  exclude:
+    labels:
+      - chore
+    authors:
+      - dependabot
+      - dependabot[bot]
+  categories:
+    - title: "\U0001F6A8 Security"
+      labels:
+        - security
+    - title: "\U0001F4A5 Breaking Changes"
+      labels:
+        - breaking
+    - title: "\U0001F680 Features"
+      labels:
+        - enhancement
+    - title: "\U0001F41B Bug Fixes"
+      labels:
+        - bug
+    - title: "\U0001F3CE\uFE0F Performance"
+      labels:
+        - performance
+    - title: "\U0001F9EA Tests"
+      labels:
+        - test
+    - title: "\U0001F504 Refactoring"
+      labels:
+        - refactor
+    - title: "\U0001F477 CI/CD"
+      labels:
+        - ci
+    - title: "\U0001F4E6 Dependencies"
+      labels:
+        - dependencies
+    - title: "\U0001F4DD Other Changes"
+      labels:
+        - '*'
+      exclude:
+        labels:
+          - dependencies
+          - ci
+          - test
+          - refactor
+          - chore

.github/scripts/triage/embedding_utils.py

@@ -0,0 +1,257 @@
+"""Pure math utilities for triage sweep embedding analysis.
+
+All functions are stateless and perform no I/O (except model loading by FastEmbed).
+Each function operates on numpy arrays and returns numpy arrays or plain Python types.
+"""
+from __future__ import annotations
+
+import numpy as np
+from numpy.typing import NDArray
+from fastembed import TextEmbedding
+from sklearn.decomposition import PCA
+from sklearn.covariance import EllipticEnvelope
+from sklearn.metrics.pairwise import cosine_similarity
+
+# FastEmbed model — BAAI/bge-small-en-v1.5 produces 384-dimensional embeddings.
+# ~46MB quantized ONNX, runs on CPU in ~0.5s per batch of 32.
+EMBEDDING_MODEL: str = "BAAI/bge-small-en-v1.5"
+
+# Embedding dimensionality (determined by model choice).
+EMBEDDING_DIM: int = 384
+
+# Batch size for FastEmbed. 32 balances memory and throughput on
+# a 2-vCPU GitHub Actions runner with ~7GB RAM.
+EMBEDDING_BATCH_SIZE: int = 32
+
+
+def embed_texts(texts: list[str]) -> NDArray[np.float32]:
+    """Embed a list of texts into dense vectors using FastEmbed.
+
+    Returns an array of shape (len(texts), 384) with dtype float32.
+    Empty input returns a (0, 384) array.
+    """
+    if not texts:
+        return np.empty((0, EMBEDDING_DIM), dtype=np.float32)
+
+    model = TextEmbedding(model_name=EMBEDDING_MODEL)
+    vectors = list(model.embed(texts, batch_size=EMBEDDING_BATCH_SIZE))
+    return np.vstack(vectors).astype(np.float32)
+
+
+def normalize_rows(matrix: NDArray[np.float32]) -> NDArray[np.float32]:
+    """L2-normalize each row to unit length.
+
+    Zero-norm rows (e.g. from empty text) remain zero vectors.
+    Uses eps=1e-10 in the denominator to avoid division by zero.
+    """
+    if matrix.shape[0] == 0:
+        return matrix
+
+    norms = np.linalg.norm(matrix, axis=1, keepdims=True)
+    return matrix / (norms + 1e-10)
+
+
+def reduce_dimensions(
+    matrix: NDArray[np.float32],
+    max_components: int,
+) -> NDArray[np.float32]:
+    """Reduce dimensionality via PCA.
+
+    Computes n_components = min(max_components, n-1, d). If n_components < 1,
+    returns the matrix unchanged. Logs explained variance for observability.
+    """
+    n, d = matrix.shape
+    if n <= 1:
+        return matrix
+
+    n_components = min(max_components, n - 1, d)
+    if n_components < 1:
+        return matrix
+
+    pca = PCA(n_components=n_components)
+    reduced = pca.fit_transform(matrix)
+    explained = pca.explained_variance_ratio_.sum()
+    print(f"PCA: {d}d -> {n_components}d, explained variance: {explained:.3f}")
+    return reduced.astype(np.float32)
+
+
+def detect_outliers(
+    matrix: NDArray[np.float32],
+    contamination: float = 0.1,
+    iqr_multiplier: float = 3.0,
+    max_outlier_pct: float = 0.05,
+) -> list[tuple[int, float]]:
+    """Flag items whose Mahalanobis distance exceeds an IQR-based cutoff.
+
+    Uses EllipticEnvelope (robust covariance via MCD) to estimate the
+    multivariate Gaussian, then computes sqrt(squared Mahalanobis distance)
+    for each sample. The cutoff is Q75 + iqr_multiplier * IQR, which
+    adapts to the actual distribution of distances.
+
+    A hard cap ensures no more than max_outlier_pct * n items are flagged;
+    when the cap is hit, only the most extreme items (sorted by distance
+    descending) are kept.
+
+    Returns (index, distance) tuples sorted by index ascending, along with
+    the cutoff value stored as an attribute on the returned list.
+    """
+    n = matrix.shape[0]
+    if n < 2:
+        return []
+
+    envelope = EllipticEnvelope(contamination=contamination, random_state=42)
+    envelope.fit(matrix)
+
+    # .mahalanobis() returns squared Mahalanobis distances
+    distances = np.sqrt(envelope.mahalanobis(matrix))
+
+    # IQR-based cutoff
+    q25, q75 = np.percentile(distances, [25, 75])
+    iqr = q75 - q25
+    cutoff = q75 + iqr_multiplier * iqr
+
+    outlier_mask = distances > cutoff
+    indices = np.where(outlier_mask)[0]
+
+    # Hard cap: keep at most max_outlier_pct * n items
+    max_count = max(1, int(max_outlier_pct * n))
+    if len(indices) > max_count:
+        # Sort by distance descending, take the most extreme
+        sorted_by_dist = sorted(indices, key=lambda i: distances[i], reverse=True)
+        indices = np.array(sorted_by_dist[:max_count])
+
+    # Sort by index ascending for stable output
+    indices = np.sort(indices)
+    result = [(int(idx), float(distances[idx])) for idx in indices]
+
+    # Attach cutoff as metadata so the report can use it
+    result = _OutlierResult(result)  # type: ignore[assignment]
+    result.cutoff = float(cutoff)  # type: ignore[attr-defined]
+    return result  # type: ignore[return-value]
+
+
+class _OutlierResult(list):
+    """A list subclass that carries metadata (cutoff) from outlier detection."""
+    cutoff: float = 0.0
+
+
+def find_duplicate_pairs(
+    matrix: NDArray[np.float32],
+    threshold: float,
+) -> list[tuple[int, int, float]]:
+    """Find pairs of items with cosine similarity above threshold.
+
+    Returns (i, j, similarity) tuples where i < j. The input should be
+    L2-normalized embeddings (full dimensionality, not PCA-reduced) so
+    cosine similarity equals the dot product.
+    """
+    n = matrix.shape[0]
+    if n <= 1:
+        return []
+
+    sim_matrix = cosine_similarity(matrix)
+    # Upper triangle indices (i < j), excluding diagonal
+    rows, cols = np.triu_indices(n, k=1)
+    similarities = sim_matrix[rows, cols]
+
+    mask = similarities > threshold
+    pairs: list[tuple[int, int, float]] = []
+    for idx in np.where(mask)[0]:
+        pairs.append((int(rows[idx]), int(cols[idx]), float(similarities[idx])))
+
+    return pairs
+
+
+# ── Label suggestion via z-score normalized embedding similarity ──────
+
+# Z-score threshold: a label must be this many standard deviations above
+# the column mean to be considered a match.
+LABEL_Z_THRESHOLD: float = 1.5
+
+# Margin gate: the top-1 label must beat the second-best by this many
+# z-score units to be accepted (subsequent labels don't need a margin).
+LABEL_Z_MARGIN: float = 0.5
+
+# Floor for per-column standard deviation to avoid division by near-zero.
+LABEL_Z_STD_FLOOR: float = 0.01
+
+# Minimum raw cosine similarity required even if z-score is high.
+# Prevents suggesting labels that are "relatively best" but still poor.
+MIN_RAW_SIMILARITY: float = 0.3
+
+# Maximum number of labels to suggest per item.
+MAX_LABELS_PER_ITEM: int = 3
+
+
+def suggest_labels(
+    item_embeddings: NDArray[np.float32],
+    label_embeddings: NDArray[np.float32],
+    label_names: list[str],
+    z_threshold: float = LABEL_Z_THRESHOLD,
+    z_margin: float = LABEL_Z_MARGIN,
+    std_floor: float = LABEL_Z_STD_FLOOR,
+    min_raw_sim: float = MIN_RAW_SIMILARITY,
+    max_per_item: int = MAX_LABELS_PER_ITEM,
+) -> list[list[tuple[str, float]]]:
+    """Suggest labels for each item using z-score normalized similarity.
+
+    1. Compute raw cosine similarity matrix (n items x m labels).
+    2. Column-wise z-score: for each label j, normalize across all items.
+    3. For each item, rank labels by z-score descending.
+    4. Accept a label only if z >= z_threshold AND raw_sim >= min_raw_sim.
+    5. Margin gate: the top-1 label must beat #2 by z_margin; subsequent
+       labels don't need a margin.
+    6. Cap at max_per_item.
+
+    Returns a list of length n, where each element is a list of
+    (label_name, raw_similarity) tuples. Empty list if nothing qualifies.
+    """
+    n = item_embeddings.shape[0]
+    m = label_embeddings.shape[0]
+    if n == 0 or m == 0:
+        return [[] for _ in range(n)]
+
+    # (n, m) raw similarity matrix
+    sim_matrix = cosine_similarity(item_embeddings, label_embeddings)
+
+    # Column-wise z-score normalization
+    col_means = sim_matrix.mean(axis=0)  # shape (m,)
+    col_stds = sim_matrix.std(axis=0)    # shape (m,)
+    col_stds = np.maximum(col_stds, std_floor)
+    z_matrix = (sim_matrix - col_means) / col_stds
+
+    suggestions: list[list[tuple[str, float]]] = []
+    for i in range(n):
+        z_row = z_matrix[i]
+        raw_row = sim_matrix[i]
+
+        # Rank labels by z-score descending
+        ranked = np.argsort(z_row)[::-1]
+
+        item_labels: list[tuple[str, float]] = []
+
+        # Margin gate: top-1 z-score must beat #2 by z_margin.
+        # If not, the assignment is ambiguous — skip this item entirely.
+        if len(ranked) > 1:
+            top1_z = float(z_row[ranked[0]])
+            top2_z = float(z_row[ranked[1]])
+            if top1_z - top2_z < z_margin:
+                suggestions.append(item_labels)
+                continue
+
+        for rank_pos, idx in enumerate(ranked):
+            if len(item_labels) >= max_per_item:
+                break
+
+            z_val = float(z_row[idx])
+            raw_val = float(raw_row[idx])
+
+            # Must pass both z-threshold and raw similarity floor
+            if z_val < z_threshold or raw_val < min_raw_sim:
+                continue
+
+            item_labels.append((label_names[idx], raw_val))
+
+        suggestions.append(item_labels)
+
+    return suggestions

.github/scripts/triage/requirements.txt

@@ -0,0 +1,4 @@
+fastembed>=0.5.0
+numpy>=1.26.0
+scikit-learn>=1.4.0
+scipy>=1.10.0

.github/scripts/triage/sweep.py

@@ -0,0 +1,600 @@
+"""Triage sweep: fetch open issues/PRs, detect outliers and duplicates, generate a report.
+
+Entrypoint script for the triage-sweep workflow. Fetches all open items via
+the GitHub REST API, delegates embedding and analysis to embedding_utils,
+generates a markdown report, and optionally creates a report issue.
+"""
+from __future__ import annotations
+
+import json
+import os
+import sys
+import urllib.request
+import urllib.parse
+from typing import TypedDict
+from datetime import datetime, timezone
+
+from embedding_utils import (
+    embed_texts,
+    normalize_rows,
+    reduce_dimensions,
+    detect_outliers,
+    find_duplicate_pairs,
+    suggest_labels,
+    LABEL_Z_THRESHOLD,
+    LABEL_Z_MARGIN,
+    LABEL_Z_STD_FLOOR,
+    MIN_RAW_SIMILARITY,
+    MAX_LABELS_PER_ITEM,
+)
+
+# ── Thresholds (overridable via workflow_dispatch inputs) ──────────────
+
+# IQR multiplier for outlier cutoff: cutoff = Q75 + IQR_MULTIPLIER * IQR.
+IQR_MULTIPLIER: float = float(os.environ.get("INPUT_IQR_MULTIPLIER", "3.0"))
+
+# Hard cap: at most this fraction of items can be flagged as outliers.
+MAX_OUTLIER_PCT: float = float(os.environ.get("INPUT_MAX_OUTLIER_PCT", "0.05"))
+
+# EllipticEnvelope contamination: expected fraction of outliers in the data.
+# Governs how aggressively the robust covariance downweights extreme points.
+CONTAMINATION: float = float(os.environ.get("INPUT_CONTAMINATION", "0.1"))
+
+# Cosine similarity above which two items are flagged as duplicates.
+# 0.92 catches near-identical issues while tolerating paraphrasing.
+COSINE_THRESHOLD: float = float(os.environ.get("INPUT_COSINE_THRESHOLD", "0.92"))
+
+# Hard cap on items to process. Prevents runaway costs on very large repos.
+MAX_ITEMS: int = int(os.environ.get("INPUT_MAX_ITEMS", "500"))
+
+# When true, print report to stdout/file but do not create a GitHub issue.
+DRY_RUN: bool = os.environ.get("INPUT_DRY_RUN", "false").lower() == "true"
+
+# ── Fixed constants (not user-configurable) ───────────────────────────
+
+# Minimum number of samples required for EllipticEnvelope to fit
+# a Gaussian reliably. Must be >= 3 * PCA_MAX_COMPONENTS so the
+# covariance matrix is estimated from enough data points.
+PCA_MAX_COMPONENTS: int = 20
+MIN_SAMPLES_FOR_OUTLIER_DETECTION: int = 100
+
+# Max character length for embedding input text. bge-small-en-v1.5 has a
+# 512-token context window (~4 chars/token). We keep title + body under
+# this limit so the model sees the full text instead of silently truncating.
+MAX_EMBED_CHARS: int = 2000
+
+# GitHub REST API page size (max allowed is 100).
+API_PAGE_SIZE: int = 100
+
+# Report issue label.
+REPORT_LABEL: str = "triage-report"
+
+# Report file path (written for the summary step to pick up).
+REPORT_FILE: str = "/tmp/triage-report.md"
+
+
+class TriageItem(TypedDict):
+    """One open issue or PR, with only the fields we need."""
+    number: int
+    title: str
+    html_url: str
+    is_pr: bool
+    labels: list[str]
+    created_at: str
+    # title + body concatenated, used as embedding input
+    text: str
+
+
+def github_api_get(path: str) -> list[dict]:
+    """Make a single authenticated GET request to the GitHub REST API.
+
+    Reads GITHUB_TOKEN and GITHUB_REPOSITORY from env. Raises SystemExit
+    with the HTTP status and response body on any non-2xx response.
+    """
+    token = os.environ["GITHUB_TOKEN"]
+    repo = os.environ["GITHUB_REPOSITORY"]
+    url = f"https://api.github.com/repos/{repo}{path}"
+
+    req = urllib.request.Request(url)
+    req.add_header("Accept", "application/vnd.github+json")
+    req.add_header("Authorization", f"Bearer {token}")
+    req.add_header("X-GitHub-Api-Version", "2022-11-28")
+
+    try:
+        with urllib.request.urlopen(req, timeout=30) as resp:
+            return json.loads(resp.read().decode("utf-8"))
+    except urllib.error.HTTPError as e:
+        body = e.read().decode("utf-8", errors="replace")
+        print(f"::error::GitHub API {e.code}: {body}")
+        sys.exit(1)
+
+
+def fetch_all_open_items() -> list[TriageItem]:
+    """Paginate through all open issues and PRs.
+
+    Returns up to MAX_ITEMS TriageItem dicts. Items with a pull_request
+    key are marked is_pr=True. The text field is title + body concatenated.
+    """
+    items: list[TriageItem] = []
+    page = 1
+
+    while len(items) < MAX_ITEMS:
+        path = (
+            f"/issues?state=open&per_page={API_PAGE_SIZE}"
+            f"&sort=created&direction=desc&page={page}"
+        )
+        data = github_api_get(path)
+
+        if not data:
+            break
+
+        for raw in data:
+            if len(items) >= MAX_ITEMS:
+                break
+
+            body = raw.get("body", "") or ""
+            full_text = f"{raw['title']}\n\n{body}"
+            # Truncate to fit the embedding model's token window.
+            # Title is always preserved; body gets clipped if needed.
+            if len(full_text) > MAX_EMBED_CHARS:
+                full_text = full_text[:MAX_EMBED_CHARS]
+            items.append(TriageItem(
+                number=raw["number"],
+                title=raw["title"],
+                html_url=raw["html_url"],
+                is_pr="pull_request" in raw,
+                labels=[lbl["name"] for lbl in raw.get("labels", [])],
+                created_at=raw["created_at"],
+                text=full_text,
+            ))
+
+        if len(data) < API_PAGE_SIZE:
+            break
+
+        page += 1
+
+    return items
+
+
+class RepoLabel(TypedDict):
+    """A label from the repo with its embedding text."""
+    name: str
+    description: str
+    # "name: description" concatenated for embedding
+    text: str
+
+
+def fetch_repo_labels() -> list[RepoLabel]:
+    """Fetch all labels from the repository, paginating if needed.
+
+    Returns labels with name, description, and a text field suitable
+    for embedding ("name: description"). Labels with no description
+    use just the name.
+    """
+    labels: list[RepoLabel] = []
+    page = 1
+
+    while True:
+        data = github_api_get(f"/labels?per_page={API_PAGE_SIZE}&page={page}")
+        for raw in data:
+            name = raw["name"]
+            desc = raw.get("description", "") or ""
+            text = f"{name}: {desc}" if desc else name
+            labels.append(RepoLabel(name=name, description=desc, text=text))
+
+        if len(data) < API_PAGE_SIZE:
+            break
+        page += 1
+
+    return labels
+
+
+def apply_labels_to_item(item_number: int, labels: list[str]) -> None:
+    """Add labels to a single issue/PR via the GitHub API.
+
+    Skips silently if labels list is empty. Uses POST which adds labels
+    without removing existing ones.
+    """
+    if not labels:
+        return
+
+    token = os.environ["GITHUB_TOKEN"]
+    repo = os.environ["GITHUB_REPOSITORY"]
+    url = f"https://api.github.com/repos/{repo}/issues/{item_number}/labels"
+
+    payload = json.dumps({"labels": labels}).encode("utf-8")
+    req = urllib.request.Request(url, data=payload, method="POST")
+    req.add_header("Accept", "application/vnd.github+json")
+    req.add_header("Authorization", f"Bearer {token}")
+    req.add_header("X-GitHub-Api-Version", "2022-11-28")
+    req.add_header("Content-Type", "application/json")
+
+    try:
+        with urllib.request.urlopen(req, timeout=30) as resp:
+            resp.read()
+    except urllib.error.HTTPError as e:
+        body = e.read().decode("utf-8", errors="replace")
+        # Non-fatal: log warning but don't abort the sweep
+        print(f"::warning::Failed to label #{item_number}: {e.code} {body}")
+
+
+def _item_age(created_at: str) -> str:
+    """Compute a human-readable age string from an ISO 8601 created_at timestamp."""
+    try:
+        created = datetime.fromisoformat(created_at.replace("Z", "+00:00"))
+        delta = datetime.now(timezone.utc) - created
+        days = delta.days
+        if days < 1:
+            return "<1d"
+        if days < 30:
+            return f"{days}d"
+        if days < 365:
+            return f"{days // 30}mo"
+        return f"{days // 365}y"
+    except (ValueError, TypeError):
+        return "?"
+
+
+def _suggested_action(a: TriageItem, b: TriageItem) -> str:
+    """Determine a suggested action for a duplicate pair based on types and age."""
+    if a["is_pr"] and b["is_pr"]:
+        return "Review for overlap"
+    if not a["is_pr"] and not b["is_pr"]:
+        # Both issues — close the newer one
+        try:
+            a_dt = datetime.fromisoformat(a["created_at"].replace("Z", "+00:00"))
+            b_dt = datetime.fromisoformat(b["created_at"].replace("Z", "+00:00"))
+            newer = b if b_dt > a_dt else a
+        except (ValueError, TypeError):
+            newer = b
+        return f"Close #{newer['number']} as duplicate"
+    # One issue, one PR
+    return "Link PR to issue"
+
+
+def generate_report(
+    items: list[TriageItem],
+    outlier_results: list[tuple[int, float]],
+    duplicate_pairs: list[tuple[int, int, float]],
+    label_suggestions: list[list[tuple[str, float]]] | None = None,
+) -> str:
+    """Generate a structured markdown triage report."""
+    now = datetime.now(timezone.utc).strftime("%Y-%m-%d %H:%M:%S")
+    repo = os.environ.get("GITHUB_REPOSITORY", "unknown/repo")
+
+    # Compute label suggestion counts early for the health table
+    outlier_set = {idx for idx, _ in outlier_results}
+    suggested_count = 0
+    if label_suggestions is not None:
+        suggested_count = sum(
+            1 for i, s in enumerate(label_suggestions)
+            if s and not items[i]["labels"] and i not in outlier_set
+        )
+
+    # ── Health summary table at the top ──────────────────────────────
+    lines: list[str] = [
+        "## Triage Sweep Report",
+        "",
+        f"**Run:** {now} UTC",
+        f"**Items analyzed:** {len(items)}",
+        f"**Thresholds:** IQR multiplier {IQR_MULTIPLIER}, Cosine > {COSINE_THRESHOLD}",
+        "",
+        "### Health Summary",
+        "",
+        "| Metric | Value |",
+        "|--------|-------|",
+        f"| Items analyzed | {len(items)} |",
+        f"| Outliers flagged | {len(outlier_results)} |",
+        f"| Duplicate pairs | {len(duplicate_pairs)} |",
+        f"| Label suggestions | {suggested_count} |",
+        "",
+    ]
+
+    # ── Outlier section ──────────────────────────────────────────────
+    # Determine cutoff for high-confidence split
+    cutoff = getattr(outlier_results, "cutoff", 0.0)
+    high_conf_cutoff = 2 * cutoff if cutoff > 0 else float("inf")
+
+    high_conf = [(idx, d) for idx, d in outlier_results if d > high_conf_cutoff]
+    borderline = [(idx, d) for idx, d in outlier_results if d <= high_conf_cutoff]
+
+    lines.extend([
+        f"### Potential Outliers / Spam ({len(outlier_results)})",
+        "",
+        "Items with unusually high Mahalanobis distance from the distribution center.",
+        "These may be spam, off-topic, or poorly described.",
+        "",
+    ])
+
+    if high_conf:
+        lines.append(f"**High Confidence** ({len(high_conf)} items, distance > 2x cutoff)")
+        lines.append("")
+        lines.append("| # | Type | Title | Distance | Age |")
+        lines.append("|---|------|-------|----------|-----|")
+        for idx, distance in high_conf:
+            item = items[idx]
+            kind = "PR" if item["is_pr"] else "Issue"
+            age = _item_age(item["created_at"])
+            title = item["title"][:80] + ("..." if len(item["title"]) > 80 else "")
+            lines.append(
+                f"| [#{item['number']}]({item['html_url']}) "
+                f"| {kind} | {title} | {distance:.2f} | {age} |"
+            )
+        lines.append("")
+
+    if borderline:
+        lines.append("<details>")
+        lines.append(f"<summary>Borderline ({len(borderline)} items)</summary>")
+        lines.append("")
+        lines.append("| # | Type | Title | Distance | Age |")
+        lines.append("|---|------|-------|----------|-----|")
+        for idx, distance in borderline:
+            item = items[idx]
+            kind = "PR" if item["is_pr"] else "Issue"
+            age = _item_age(item["created_at"])
+            title = item["title"][:80] + ("..." if len(item["title"]) > 80 else "")
+            lines.append(
+                f"| [#{item['number']}]({item['html_url']}) "
+                f"| {kind} | {title} | {distance:.2f} | {age} |"
+            )
+        lines.append("")
+        lines.append("</details>")
+        lines.append("")
+
+    if not outlier_results:
+        lines.append("None found.")
+
+    # ── Duplicate pairs section ──────────────────────────────────────
+    lines.extend([
+        "",
+        f"### Potential Duplicates ({len(duplicate_pairs)} pairs)",
+        "",
+        "Pairs of items with cosine similarity above the threshold.",
+        "",
+    ])
+
+    if duplicate_pairs:
+        lines.append("| Item A | Item B | Similarity | Suggested Action |")
+        lines.append("|--------|--------|------------|------------------|")
+        for i, j, sim in duplicate_pairs:
+            a = items[i]
+            b = items[j]
+            kind_a = "PR" if a["is_pr"] else "Issue"
+            kind_b = "PR" if b["is_pr"] else "Issue"
+            action = _suggested_action(a, b)
+            lines.append(
+                f"| [#{a['number']}]({a['html_url']}) {kind_a}: {a['title']} "
+                f"| [#{b['number']}]({b['html_url']}) {kind_b}: {b['title']} "
+                f"| {sim:.3f} | {action} |"
+            )
+    else:
+        lines.append("None found.")
+
+    # ── Label suggestions section ────────────────────────────────────
+    if label_suggestions is not None:
+        # High confidence: top-1 label with raw_sim >= 0.5
+        # Low confidence: top-1 label with raw_sim < 0.5
+        high_conf_labels: list[tuple[int, list[tuple[str, float]]]] = []
+        low_conf_labels: list[tuple[int, list[tuple[str, float]]]] = []
+        for i, sugs in enumerate(label_suggestions):
+            if sugs and not items[i]["labels"] and i not in outlier_set:
+                top1 = sugs[:1]
+                if top1[0][1] >= 0.5:
+                    high_conf_labels.append((i, top1))
+                else:
+                    low_conf_labels.append((i, top1))
+
+        total_suggestions = len(high_conf_labels) + len(low_conf_labels)
+        lines.extend([
+            "",
+            f"### Suggested Labels ({total_suggestions} unlabeled items)",
+            "",
+            "Labels suggested by z-score normalized embedding similarity against repo label descriptions.",
+            "Only shown for unlabeled items that were not flagged as outliers.",
+            "",
+        ])
+
+        # Label concentration warning
+        if total_suggestions > 0:
+            label_counts: dict[str, int] = {}
+            for _, sugs in high_conf_labels + low_conf_labels:
+                for name, _ in sugs:
+                    label_counts[name] = label_counts.get(name, 0) + 1
+            for name, count in label_counts.items():
+                if count > total_suggestions * 0.5:
+                    lines.append(
+                        f"> **Warning:** Label `{name}` accounts for "
+                        f"{count}/{total_suggestions} suggestions "
+                        f"({count * 100 // total_suggestions}%). "
+                        f"Consider reviewing label descriptions for specificity."
+                    )
+                    lines.append("")
+
+        if high_conf_labels:
+            lines.append("| # | Type | Title | Suggested Label |")
+            lines.append("|---|------|-------|--------------------|")
+            for idx, sugs in high_conf_labels:
+                item = items[idx]
+                kind = "PR" if item["is_pr"] else "Issue"
+                label_strs = [f"`{name}` ({score:.2f})" for name, score in sugs]
+                lines.append(
+                    f"| [#{item['number']}]({item['html_url']}) "
+                    f"| {kind} | {item['title']} | {', '.join(label_strs)} |"
+                )
+
+        if low_conf_labels:
+            lines.append("")
+            lines.append("<details>")
+            lines.append(f"<summary>Low-confidence suggestions ({len(low_conf_labels)} items)</summary>")
+            lines.append("")
+            lines.append("| # | Type | Title | Suggested Label |")
+            lines.append("|---|------|-------|--------------------|")
+            for idx, sugs in low_conf_labels:
+                item = items[idx]
+                kind = "PR" if item["is_pr"] else "Issue"
+                label_strs = [f"`{name}` ({score:.2f})" for name, score in sugs]
+                lines.append(
+                    f"| [#{item['number']}]({item['html_url']}) "
+                    f"| {kind} | {item['title']} | {', '.join(label_strs)} |"
+                )
+            lines.append("")
+            lines.append("</details>")
+
+        if not high_conf_labels and not low_conf_labels:
+            lines.append("No unlabeled items need suggestions.")
+
+    lines.extend([
+        "",
+        "### Summary",
+        "",
+        f"- {len(outlier_results)} outliers flagged for review",
+        f"- {len(duplicate_pairs)} duplicate pairs found",
+        f"- {len(items)} items analyzed in total",
+    ])
+
+    if label_suggestions is not None:
+        lines.append(f"- {suggested_count} items suggested for labeling")
+
+    lines.extend([
+        "",
+        "---",
+        f"*Generated by [triage-sweep](https://github.com/{repo}/actions) — no LLM was used.*",
+    ])
+
+    return "\n".join(lines)
+
+
+def create_report_issue(report_body: str) -> None:
+    """Create a GitHub issue with the triage report.
+
+    Posts to the issues API with the triage-report label.
+    Raises SystemExit on non-201 response.
+    """
+    token = os.environ["GITHUB_TOKEN"]
+    repo = os.environ["GITHUB_REPOSITORY"]
+    url = f"https://api.github.com/repos/{repo}/issues"
+
+    today = datetime.now(timezone.utc).strftime("%Y-%m-%d")
+    payload = json.dumps({
+        "title": f"Triage Sweep Report — {today}",
+        "body": report_body,
+        "labels": [REPORT_LABEL],
+    }).encode("utf-8")
+
+    req = urllib.request.Request(url, data=payload, method="POST")
+    req.add_header("Accept", "application/vnd.github+json")
+    req.add_header("Authorization", f"Bearer {token}")
+    req.add_header("X-GitHub-Api-Version", "2022-11-28")
+    req.add_header("Content-Type", "application/json")
+
+    try:
+        with urllib.request.urlopen(req, timeout=30) as resp:
+            resp_body = resp.read().decode("utf-8")
+            if resp.status != 201:
+                print(f"::error::Failed to create issue: {resp.status} {resp_body}")
+                sys.exit(1)
+            result = json.loads(resp_body)
+            print(f"Created issue: {result.get('html_url', 'unknown')}")
+    except urllib.error.HTTPError as e:
+        body = e.read().decode("utf-8", errors="replace")
+        print(f"::error::Failed to create issue: {e.code} {body}")
+        sys.exit(1)
+
+
+def write_report(report: str) -> None:
+    """Write the report to the file system for the summary step."""
+    with open(REPORT_FILE, "w", encoding="utf-8") as f:
+        f.write(report)
+
+
+def main() -> None:
+    """Orchestrate the full triage sweep."""
+    # 1. Validate environment
+    for var in ("GITHUB_TOKEN", "GITHUB_REPOSITORY"):
+        if not os.environ.get(var):
+            print(f"::error::Missing required environment variable: {var}")
+            sys.exit(1)
+
+    # 2. Fetch all open issues + PRs
+    items = fetch_all_open_items()
+    print(f"Fetched {len(items)} open items")
+
+    if len(items) == 0:
+        report = "## Triage Sweep Report\n\nNo open issues or PRs found."
+        write_report(report)
+        print("No items to analyze.")
+        return
+
+    # 3. Extract texts for embedding
+    texts: list[str] = [item["text"] for item in items]
+
+    # 4. Embed all texts (returns numpy float32 array of shape [n, 384])
+    embeddings = embed_texts(texts)
+
+    # 5. L2-normalize
+    embeddings = normalize_rows(embeddings)
+
+    # 6. Outlier detection (Mahalanobis via EllipticEnvelope)
+    outlier_results: list[tuple[int, float]] = []
+    if len(items) >= MIN_SAMPLES_FOR_OUTLIER_DETECTION:
+        reduced = reduce_dimensions(embeddings, PCA_MAX_COMPONENTS)
+        outlier_results = detect_outliers(
+            reduced,
+            contamination=CONTAMINATION,
+            iqr_multiplier=IQR_MULTIPLIER,
+            max_outlier_pct=MAX_OUTLIER_PCT,
+        )
+    else:
+        print(
+            f"Skipping outlier detection: {len(items)} items < "
+            f"{MIN_SAMPLES_FOR_OUTLIER_DETECTION} minimum"
+        )
+
+    # 7. Duplicate detection (pairwise cosine similarity)
+    duplicate_pairs = find_duplicate_pairs(embeddings, COSINE_THRESHOLD)
+
+    # 8. Label suggestion via embedding similarity
+    label_suggestions: list[list[tuple[str, float]]] | None = None
+    repo_labels = fetch_repo_labels()
+    if repo_labels:
+        label_texts = [lbl["text"] for lbl in repo_labels]
+        label_names = [lbl["name"] for lbl in repo_labels]
+        label_embeddings = embed_texts(label_texts)
+        label_embeddings = normalize_rows(label_embeddings)
+        label_suggestions = suggest_labels(embeddings, label_embeddings, label_names)
+        print(f"Computed label suggestions against {len(repo_labels)} repo labels")
+
+        # NOTE: Auto-labeling is disabled. The report shows suggestions for
+        # human review. To re-enable, uncomment the block below.
+        #
+        # # Apply top label to unlabeled items (unless dry run)
+        # # Skip outliers — flagged items shouldn't get categorized
+        # outlier_set = {idx for idx, _ in outlier_results}
+        # if not DRY_RUN:
+        #     applied_count = 0
+        #     for i, sugs in enumerate(label_suggestions):
+        #         if sugs and not items[i]["labels"] and i not in outlier_set:
+        #             # Apply only the top-1 label (highest confidence)
+        #             apply_labels_to_item(items[i]["number"], [sugs[0][0]])
+        #             applied_count += 1
+        #     print(f"Applied labels to {applied_count} unlabeled items")
+    else:
+        print("No repo labels found — skipping label suggestions")
+
+    # 9. Generate report
+    report = generate_report(items, outlier_results, duplicate_pairs, label_suggestions)
+
+    # 10. Write report to file (for summary step)
+    write_report(report)
+
+    # 11. Create report issue (unless dry run)
+    if DRY_RUN:
+        print("Dry run — skipping issue creation and label application.")
+        print(report)
+    else:
+        create_report_issue(report)
+        print("Report issue created.")
+
+
+if __name__ == "__main__":
+    main()

.github/scripts/triage/test_embedding_utils.py

@@ -0,0 +1,468 @@
+"""Tests for embedding_utils.py — all embedding model calls are mocked."""
+from __future__ import annotations
+
+import sys
+from unittest.mock import patch, MagicMock
+import numpy as np
+import pytest
+
+# Mock fastembed before importing the module under test (persistent)
+if "fastembed" not in sys.modules:
+    sys.modules["fastembed"] = MagicMock()
+
+from embedding_utils import (
+    embed_texts,
+    normalize_rows,
+    reduce_dimensions,
+    detect_outliers,
+    find_duplicate_pairs,
+    suggest_labels,
+    EMBEDDING_DIM,
+    EMBEDDING_MODEL,
+    EMBEDDING_BATCH_SIZE,
+    LABEL_Z_THRESHOLD,
+    LABEL_Z_MARGIN,
+    LABEL_Z_STD_FLOOR,
+    MIN_RAW_SIMILARITY,
+    MAX_LABELS_PER_ITEM,
+)
+
+
+class TestEmbedTexts:
+    """Tests for the embed_texts function."""
+
+    def test_empty_list_returns_empty_array(self):
+        result = embed_texts([])
+        assert result.shape == (0, EMBEDDING_DIM)
+        assert result.dtype == np.float32
+
+    @patch("embedding_utils.TextEmbedding")
+    def test_single_text(self, mock_cls):
+        mock_model = MagicMock()
+        mock_cls.return_value = mock_model
+        vec = np.random.randn(EMBEDDING_DIM).astype(np.float32)
+        mock_model.embed.return_value = iter([vec])
+
+        result = embed_texts(["hello world"])
+
+        mock_cls.assert_called_once_with(model_name=EMBEDDING_MODEL)
+        mock_model.embed.assert_called_once_with(
+            ["hello world"], batch_size=EMBEDDING_BATCH_SIZE
+        )
+        assert result.shape == (1, EMBEDDING_DIM)
+        assert result.dtype == np.float32
+        np.testing.assert_array_almost_equal(result[0], vec)
+
+    @patch("embedding_utils.TextEmbedding")
+    def test_multiple_texts(self, mock_cls):
+        mock_model = MagicMock()
+        mock_cls.return_value = mock_model
+        vecs = [
+            np.random.randn(EMBEDDING_DIM).astype(np.float32)
+            for _ in range(5)
+        ]
+        mock_model.embed.return_value = iter(vecs)
+
+        result = embed_texts(["a", "b", "c", "d", "e"])
+        assert result.shape == (5, EMBEDDING_DIM)
+        assert result.dtype == np.float32
+
+
+class TestNormalizeRows:
+    """Tests for L2 row normalization."""
+
+    def test_empty_matrix(self):
+        m = np.empty((0, 10), dtype=np.float32)
+        result = normalize_rows(m)
+        assert result.shape == (0, 10)
+
+    def test_single_row(self):
+        m = np.array([[3.0, 4.0]], dtype=np.float32)
+        result = normalize_rows(m)
+        # Norm should be ~1.0
+        norm = np.linalg.norm(result[0])
+        assert abs(norm - 1.0) < 1e-5
+
+    def test_multiple_rows(self):
+        rng = np.random.default_rng(42)
+        m = rng.standard_normal((10, 50)).astype(np.float32)
+        result = normalize_rows(m)
+        norms = np.linalg.norm(result, axis=1)
+        np.testing.assert_allclose(norms, 1.0, atol=1e-5)
+
+    def test_zero_row_stays_near_zero(self):
+        m = np.array([[0.0, 0.0, 0.0], [1.0, 0.0, 0.0]], dtype=np.float32)
+        result = normalize_rows(m)
+        # Zero row divided by eps -> very small values
+        assert np.linalg.norm(result[0]) < 1e-3
+        # Non-zero row should be unit norm
+        assert abs(np.linalg.norm(result[1]) - 1.0) < 1e-5
+
+    def test_preserves_direction(self):
+        m = np.array([[2.0, 0.0], [0.0, 3.0]], dtype=np.float32)
+        result = normalize_rows(m)
+        np.testing.assert_allclose(result[0], [1.0, 0.0], atol=1e-5)
+        np.testing.assert_allclose(result[1], [0.0, 1.0], atol=1e-5)
+
+
+class TestReduceDimensions:
+    """Tests for PCA dimensionality reduction."""
+
+    def test_single_sample_returns_unchanged(self):
+        m = np.random.randn(1, 50).astype(np.float32)
+        result = reduce_dimensions(m, 10)
+        np.testing.assert_array_equal(result, m)
+
+    def test_reduces_dimensions(self):
+        rng = np.random.default_rng(42)
+        m = rng.standard_normal((100, 50)).astype(np.float32)
+        result = reduce_dimensions(m, 10)
+        assert result.shape == (100, 10)
+        assert result.dtype == np.float32
+
+    def test_caps_at_n_minus_1(self):
+        rng = np.random.default_rng(42)
+        # 5 samples, 20 features -> max components = 4 (n-1)
+        m = rng.standard_normal((5, 20)).astype(np.float32)
+        result = reduce_dimensions(m, 50)
+        assert result.shape == (5, 4)
+
+    def test_caps_at_d(self):
+        rng = np.random.default_rng(42)
+        # 100 samples, 3 features -> max components = 3
+        m = rng.standard_normal((100, 3)).astype(np.float32)
+        result = reduce_dimensions(m, 50)
+        assert result.shape == (100, 3)
+
+    def test_max_components_respected(self):
+        rng = np.random.default_rng(42)
+        m = rng.standard_normal((50, 30)).astype(np.float32)
+        result = reduce_dimensions(m, 5)
+        assert result.shape[1] == 5
+
+
+class TestDetectOutliers:
+    """Tests for IQR-based outlier detection."""
+
+    def test_single_sample_returns_empty(self):
+        m = np.random.randn(1, 5).astype(np.float32)
+        result = detect_outliers(m)
+        assert result == []
+
+    def test_empty_returns_empty(self):
+        # n < 2 case
+        m = np.empty((0, 5), dtype=np.float32)
+        result = detect_outliers(m)
+        assert result == []
+
+    def test_finds_outliers_in_synthetic_data(self):
+        rng = np.random.default_rng(42)
+        # Create a tight cluster with one obvious outlier
+        cluster = rng.standard_normal((50, 3)).astype(np.float32) * 0.1
+        outlier = np.array([[100.0, 100.0, 100.0]], dtype=np.float32)
+        m = np.vstack([cluster, outlier])
+        result = detect_outliers(m)
+        # The outlier (index 50) should be detected
+        outlier_indices = [idx for idx, _ in result]
+        assert 50 in outlier_indices
+
+    def test_returns_list_of_index_distance_tuples(self):
+        rng = np.random.default_rng(42)
+        # Tight cluster + outlier to guarantee at least one result
+        cluster = rng.standard_normal((20, 3)).astype(np.float32) * 0.1
+        far_point = np.array([[50.0, 50.0, 50.0]], dtype=np.float32)
+        m = np.vstack([cluster, far_point])
+        result = detect_outliers(m)
+        assert isinstance(result, list)
+        for item in result:
+            assert isinstance(item, tuple)
+            assert len(item) == 2
+            idx, dist = item
+            assert isinstance(idx, int)
+            assert isinstance(dist, float)
+            assert dist > 0
+
+    def test_iqr_cutoff_behavior(self):
+        """Lower IQR multiplier should flag more items than higher multiplier."""
+        rng = np.random.default_rng(42)
+        m = rng.standard_normal((100, 3)).astype(np.float32)
+        low = detect_outliers(m, iqr_multiplier=1.0, max_outlier_pct=0.5)
+        high = detect_outliers(m, iqr_multiplier=5.0, max_outlier_pct=0.5)
+        assert len(low) >= len(high)
+
+    def test_dimension_aware_no_mass_flagging(self):
+        """High-dimensional clean Gaussian data should not flag everything."""
+        rng = np.random.default_rng(42)
+        # 500 samples, 10 dims — well-conditioned for robust covariance
+        m = rng.standard_normal((500, 10)).astype(np.float32)
+        result = detect_outliers(m)
+        # With IQR-based cutoff on clean Gaussian data,
+        # only a small fraction should be flagged (well under 50%)
+        assert len(result) < 250
+
+    def test_contamination_parameter(self):
+        rng = np.random.default_rng(42)
+        m = rng.standard_normal((50, 3)).astype(np.float32)
+        # Should not raise with different contamination values
+        result = detect_outliers(m, contamination=0.05)
+        assert isinstance(result, list)
+
+    def test_max_outlier_pct_hard_cap(self):
+        """The hard cap should limit outlier count to max_outlier_pct * n."""
+        rng = np.random.default_rng(42)
+        # Create data with many potential outliers (bimodal)
+        cluster = rng.standard_normal((80, 3)).astype(np.float32) * 0.1
+        outliers = rng.standard_normal((20, 3)).astype(np.float32) * 50.0
+        m = np.vstack([cluster, outliers])
+        # Very low IQR multiplier to flag a lot, but cap at 5%
+        result = detect_outliers(m, iqr_multiplier=0.5, max_outlier_pct=0.05)
+        max_allowed = max(1, int(0.05 * 100))  # 5
+        assert len(result) <= max_allowed
+
+    def test_hard_cap_keeps_most_extreme(self):
+        """When capped, the most extreme items (highest distance) should be kept."""
+        rng = np.random.default_rng(42)
+        cluster = rng.standard_normal((90, 3)).astype(np.float32) * 0.1
+        # Create outliers with increasing extremity
+        outliers = np.array([
+            [10.0, 10.0, 10.0],
+            [20.0, 20.0, 20.0],
+            [50.0, 50.0, 50.0],
+        ], dtype=np.float32)
+        m = np.vstack([cluster, outliers])
+        # Cap at ~1 item (0.01 * 93 = 0, but min is 1)
+        result = detect_outliers(m, iqr_multiplier=0.5, max_outlier_pct=0.02)
+        if len(result) > 0:
+            # The most extreme (index 92, distance for [50,50,50]) should be kept
+            indices = [idx for idx, _ in result]
+            assert 92 in indices
+
+    def test_cutoff_attribute(self):
+        """Returned result should carry a cutoff attribute."""
+        rng = np.random.default_rng(42)
+        m = rng.standard_normal((50, 3)).astype(np.float32)
+        result = detect_outliers(m)
+        assert hasattr(result, "cutoff")
+        assert isinstance(result.cutoff, float)
+        assert result.cutoff > 0
+
+
+class TestFindDuplicatePairs:
+    """Tests for cosine similarity duplicate detection."""
+
+    def test_single_item_returns_empty(self):
+        m = np.random.randn(1, 10).astype(np.float32)
+        result = find_duplicate_pairs(m, 0.9)
+        assert result == []
+
+    def test_empty_returns_empty(self):
+        m = np.empty((0, 10), dtype=np.float32)
+        result = find_duplicate_pairs(m, 0.9)
+        assert result == []
+
+    def test_identical_vectors_detected(self):
+        vec = np.random.randn(10).astype(np.float32)
+        vec = vec / np.linalg.norm(vec)
+        m = np.vstack([vec, vec, np.random.randn(10).astype(np.float32)])
+        result = find_duplicate_pairs(m, 0.99)
+        # Items 0 and 1 are identical, should be found
+        assert any(i == 0 and j == 1 for i, j, _ in result)
+
+    def test_orthogonal_vectors_not_detected(self):
+        m = np.eye(5, dtype=np.float32)
+        result = find_duplicate_pairs(m, 0.5)
+        assert result == []
+
+    def test_returns_correct_format(self):
+        vec = np.random.randn(10).astype(np.float32)
+        vec = vec / np.linalg.norm(vec)
+        m = np.vstack([vec, vec])
+        result = find_duplicate_pairs(m, 0.5)
+        assert len(result) >= 1
+        for item in result:
+            assert len(item) == 3
+            i, j, sim = item
+            assert isinstance(i, int)
+            assert isinstance(j, int)
+            assert isinstance(sim, float)
+            assert i < j
+
+    def test_i_less_than_j(self):
+        rng = np.random.default_rng(42)
+        # Create some similar vectors
+        base = rng.standard_normal(10).astype(np.float32)
+        m = np.vstack([base + rng.standard_normal(10) * 0.01 for _ in range(5)])
+        result = find_duplicate_pairs(m, 0.5)
+        for i, j, _ in result:
+            assert i < j
+
+    def test_high_threshold_fewer_pairs(self):
+        rng = np.random.default_rng(42)
+        m = rng.standard_normal((10, 20)).astype(np.float32)
+        # Normalize for meaningful cosine similarities
+        norms = np.linalg.norm(m, axis=1, keepdims=True)
+        m = m / norms
+        low = find_duplicate_pairs(m, 0.3)
+        high = find_duplicate_pairs(m, 0.9)
+        assert len(low) >= len(high)
+
+
+class TestSuggestLabels:
+    """Tests for z-score normalized label suggestion."""
+
+    def test_empty_items_returns_empty_lists(self):
+        items = np.empty((0, 10), dtype=np.float32)
+        labels = np.random.randn(3, 10).astype(np.float32)
+        result = suggest_labels(items, labels, ["a", "b", "c"])
+        assert result == []
+
+    def test_empty_labels_returns_empty_per_item(self):
+        items = np.random.randn(5, 10).astype(np.float32)
+        labels = np.empty((0, 10), dtype=np.float32)
+        result = suggest_labels(items, labels, [])
+        assert len(result) == 5
+        assert all(s == [] for s in result)
+
+    def test_identical_embedding_gets_that_label(self):
+        """If an item embedding strongly matches one label, z-score should highlight it."""
+        # Create multiple items so z-score normalization is meaningful
+        rng = np.random.default_rng(42)
+        # 10 random items + 1 item that matches label "bug" exactly
+        random_items = rng.standard_normal((10, 3)).astype(np.float32)
+        bug_vec = np.array([[1.0, 0.0, 0.0]], dtype=np.float32)
+        items = np.vstack([random_items, bug_vec])
+        labels = np.array([[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], dtype=np.float32)
+        result = suggest_labels(
+            items, labels, ["bug", "feature", "docs"],
+            z_threshold=0.5, z_margin=0.0, min_raw_sim=0.1,
+        )
+        # The last item (matching bug_vec) should get "bug" as top suggestion
+        last_item_sugs = result[-1]
+        if last_item_sugs:
+            assert last_item_sugs[0][0] == "bug"
+
+    def test_z_score_suppresses_dominant_label(self):
+        """When all items are similar to one label, z-scores should be low
+        (none stands out) and that label should not be blindly suggested."""
+        # All items identical — z-score for every item on every label is 0
+        items = np.ones((10, 3), dtype=np.float32)
+        labels = np.array([[1.0, 1.0, 1.0], [0.0, 1.0, 0.0]], dtype=np.float32)
+        result = suggest_labels(
+            items, labels, ["catch-all", "specific"],
+            z_threshold=1.5, min_raw_sim=0.3,
+        )
+        # With identical items, std=0 -> z-scores are all 0 -> nothing passes z_threshold
+        for sugs in result:
+            assert sugs == []
+
+    def test_margin_gate_blocks_top1(self):
+        """Top-1 label must beat #2 by z_margin to be accepted as position 0."""
+        rng = np.random.default_rng(99)
+        # 20 items, each slightly different, 2 labels
+        items = rng.standard_normal((20, 5)).astype(np.float32)
+        # Two labels that are nearly identical -> margin gate should block top-1
+        labels = np.array([[1.0, 0.5, 0.0, 0.0, 0.0],
+                           [1.0, 0.5, 0.01, 0.0, 0.0]], dtype=np.float32)
+        result = suggest_labels(
+            items, labels, ["label-a", "label-b"],
+            z_threshold=0.0, z_margin=10.0, min_raw_sim=0.0, max_per_item=1,
+        )
+        # With a huge margin requirement and max_per_item=1, nothing should pass
+        # because the only candidate (top-1) is blocked by margin gate,
+        # and max_per_item=1 prevents falling through to position 2
+        for sugs in result:
+            assert sugs == []
+
+    def test_margin_gate_passes_when_clear_winner(self):
+        """When top-1 clearly beats #2, it should pass the margin gate."""
+        # Create items where one strongly matches label 0 vs label 1
+        items = np.array([
+            [1.0, 0.0, 0.0, 0.0, 0.0],   # strongly matches label-a
+            [0.0, 0.0, 0.0, 0.0, 1.0],   # matches neither well
+        ] * 5, dtype=np.float32)  # 10 items for stable z-scores
+        labels = np.array([
+            [1.0, 0.0, 0.0, 0.0, 0.0],   # label-a
+            [0.0, 1.0, 0.0, 0.0, 0.0],   # label-b (orthogonal)
+        ], dtype=np.float32)
+        result = suggest_labels(
+            items, labels, ["label-a", "label-b"],
+            z_threshold=0.5, z_margin=0.3, min_raw_sim=0.1,
+        )
+        # Items matching label-a should get it suggested (clear z-score advantage)
+        got_label_a = sum(1 for sugs in result if sugs and sugs[0][0] == "label-a")
+        assert got_label_a > 0
+
+    def test_min_raw_similarity_filter(self):
+        """Even with high z-score, low raw similarity should be filtered out."""
+        # Items are orthogonal to all labels -> raw similarity near 0
+        items = np.array([[1.0, 0.0, 0.0]], dtype=np.float32)
+        labels = np.array([[0.0, 0.0, 1.0]], dtype=np.float32)
+        result = suggest_labels(
+            items, labels, ["irrelevant"],
+            z_threshold=0.0, z_margin=0.0, min_raw_sim=0.9,
+        )
+        # Raw similarity is ~0, which is below min_raw_sim=0.9
+        assert result[0] == []
+
+    def test_max_per_item_respected(self):
+        """Even if many labels qualify, max_per_item caps the results."""
+        rng = np.random.default_rng(42)
+        # Create items with some variance so z-scores differentiate
+        items = rng.standard_normal((20, 10)).astype(np.float32)
+        base = items[0]
+        # All labels very similar to item 0
+        labels = np.array([base + rng.standard_normal(10) * 0.01 for _ in range(10)])
+        names = [f"label-{i}" for i in range(10)]
+        result = suggest_labels(
+            items, labels, names,
+            z_threshold=0.0, z_margin=0.0, min_raw_sim=0.0, max_per_item=2,
+        )
+        for sugs in result:
+            assert len(sugs) <= 2
+
+    def test_returns_raw_similarity_not_z_score(self):
+        """Returned scores should be raw cosine similarity, not z-scores."""
+        rng = np.random.default_rng(42)
+        items = rng.standard_normal((15, 5)).astype(np.float32)
+        labels = rng.standard_normal((3, 5)).astype(np.float32)
+        names = ["bug", "feature", "docs"]
+        result = suggest_labels(
+            items, labels, names,
+            z_threshold=0.0, z_margin=0.0, min_raw_sim=-1.0,
+        )
+        # Raw cosine similarity should be in [-1, 1] range
+        for sugs in result:
+            for name, score in sugs:
+                assert -1.0 <= score <= 1.0 + 1e-5
+                assert isinstance(name, str)
+                assert isinstance(score, float)
+
+    def test_returns_correct_format(self):
+        rng = np.random.default_rng(42)
+        items = rng.standard_normal((3, 10)).astype(np.float32)
+        labels = rng.standard_normal((5, 10)).astype(np.float32)
+        names = ["bug", "feature", "docs", "ci", "test"]
+        result = suggest_labels(
+            items, labels, names,
+            z_threshold=0.0, z_margin=0.0, min_raw_sim=-1.0,
+        )
+        assert len(result) == 3
+        for sugs in result:
+            for name, score in sugs:
+                assert isinstance(name, str)
+                assert isinstance(score, float)
+                assert name in names
+
+    def test_text_truncation_in_labels(self):
+        """Label names should be returned as-is even when very long."""
+        rng = np.random.default_rng(42)
+        items = rng.standard_normal((10, 5)).astype(np.float32)
+        long_name = "a" * 200
+        labels = rng.standard_normal((1, 5)).astype(np.float32)
+        result = suggest_labels(
+            items, labels, [long_name],
+            z_threshold=0.0, z_margin=0.0, min_raw_sim=-1.0,
+        )
+        for sugs in result:
+            if sugs:
+                assert sugs[0][0] == long_name

.github/scripts/triage/test_sweep.py

@@ -0,0 +1,873 @@
+"""Tests for sweep.py — all external calls (API, embedding) are mocked."""
+from __future__ import annotations
+
+import json
+import os
+import sys
+from io import BytesIO
+from unittest.mock import patch, MagicMock, mock_open
+from urllib.error import HTTPError
+
+import numpy as np
+import pytest
+
+# Mock fastembed before importing sweep (which imports embedding_utils)
+sys.modules["fastembed"] = MagicMock()
+
+# Set required env vars before importing sweep (module-level constants read env)
+os.environ.setdefault("GITHUB_TOKEN", "test-token")
+os.environ.setdefault("GITHUB_REPOSITORY", "owner/repo")
+
+from sweep import (
+    github_api_get,
+    fetch_all_open_items,
+    fetch_repo_labels,
+    apply_labels_to_item,
+    generate_report,
+    create_report_issue,
+    write_report,
+    main,
+    TriageItem,
+    RepoLabel,
+    REPORT_FILE,
+    REPORT_LABEL,
+    API_PAGE_SIZE,
+    MIN_SAMPLES_FOR_OUTLIER_DETECTION,
+    PCA_MAX_COMPONENTS,
+    MAX_EMBED_CHARS,
+    IQR_MULTIPLIER,
+    MAX_OUTLIER_PCT,
+    _item_age,
+    _suggested_action,
+)
+
+
+def _make_api_issue(number: int, title: str = "Test issue", is_pr: bool = False,
+                    body: str = "Issue body", labels: list[str] | None = None,
+                    created_at: str = "2026-03-21T00:00:00Z") -> dict:
+    """Helper to build a mock GitHub API issue response object."""
+    result: dict = {
+        "number": number,
+        "title": title,
+        "html_url": f"https://github.com/owner/repo/issues/{number}",
+        "body": body,
+        "created_at": created_at,
+        "labels": [{"name": lbl} for lbl in (labels or [])],
+    }
+    if is_pr:
+        result["pull_request"] = {"url": "..."}
+    return result
+
+
+class TestGithubApiGet:
+    """Tests for the github_api_get function."""
+
+    @patch("sweep.urllib.request.urlopen")
+    def test_successful_request(self, mock_urlopen):
+        mock_resp = MagicMock()
+        mock_resp.read.return_value = json.dumps([{"id": 1}]).encode()
+        mock_resp.__enter__ = lambda s: s
+        mock_resp.__exit__ = MagicMock(return_value=False)
+        mock_urlopen.return_value = mock_resp
+
+        result = github_api_get("/issues?state=open")
+        assert result == [{"id": 1}]
+
+    @patch("sweep.urllib.request.urlopen")
+    def test_http_error_exits(self, mock_urlopen):
+        error = HTTPError(
+            url="https://api.github.com/repos/owner/repo/issues",
+            code=403,
+            msg="Forbidden",
+            hdrs=None,  # type: ignore[arg-type]
+            fp=BytesIO(b'{"message": "rate limited"}'),
+        )
+        mock_urlopen.side_effect = error
+
+        with pytest.raises(SystemExit) as exc_info:
+            github_api_get("/issues")
+        assert exc_info.value.code == 1
+
+
+class TestConstants:
+    """Tests for module-level constants."""
+
+    def test_min_samples_is_at_least_3x_pca_max(self):
+        """MIN_SAMPLES must be >= 3 * PCA_MAX_COMPONENTS for reliable covariance."""
+        assert MIN_SAMPLES_FOR_OUTLIER_DETECTION >= 3 * PCA_MAX_COMPONENTS
+
+    def test_min_samples_is_100(self):
+        assert MIN_SAMPLES_FOR_OUTLIER_DETECTION == 100
+
+    def test_pca_max_components_is_20(self):
+        assert PCA_MAX_COMPONENTS == 20
+
+    def test_iqr_multiplier_default(self):
+        assert IQR_MULTIPLIER == 3.0
+
+    def test_max_outlier_pct_default(self):
+        assert MAX_OUTLIER_PCT == 0.05
+
+
+class TestFetchAllOpenItems:
+    """Tests for fetch_all_open_items."""
+
+    @patch("sweep.github_api_get")
+    def test_empty_repo(self, mock_get):
+        mock_get.return_value = []
+        items = fetch_all_open_items()
+        assert items == []
+
+    @patch("sweep.github_api_get")
+    def test_single_page(self, mock_get):
+        mock_get.return_value = [
+            _make_api_issue(1, "Bug report"),
+            _make_api_issue(2, "Feature request", is_pr=True),
+        ]
+        items = fetch_all_open_items()
+        assert len(items) == 2
+        assert items[0]["number"] == 1
+        assert items[0]["is_pr"] is False
+        assert items[1]["is_pr"] is True
+
+    @patch("sweep.github_api_get")
+    def test_text_field_constructed(self, mock_get):
+        mock_get.return_value = [
+            _make_api_issue(1, "My Title", body="My Body"),
+        ]
+        items = fetch_all_open_items()
+        assert items[0]["text"] == "My Title\n\nMy Body"
+
+    @patch("sweep.github_api_get")
+    def test_long_body_truncated(self, mock_get):
+        """Bodies exceeding MAX_EMBED_CHARS are truncated to fit the token window."""
+        long_body = "x" * (MAX_EMBED_CHARS + 500)
+        mock_get.return_value = [
+            _make_api_issue(1, "Title", body=long_body),
+        ]
+        items = fetch_all_open_items()
+        assert len(items[0]["text"]) == MAX_EMBED_CHARS
+
+    @patch("sweep.github_api_get")
+    def test_short_body_not_truncated(self, mock_get):
+        """Bodies under the limit are left intact."""
+        mock_get.return_value = [
+            _make_api_issue(1, "Title", body="Short body"),
+        ]
+        items = fetch_all_open_items()
+        assert items[0]["text"] == "Title\n\nShort body"
+
+    @patch("sweep.github_api_get")
+    def test_null_body_handled(self, mock_get):
+        issue = _make_api_issue(1, "No body")
+        issue["body"] = None
+        mock_get.return_value = [issue]
+        items = fetch_all_open_items()
+        assert items[0]["text"] == "No body\n\n"
+
+    @patch("sweep.github_api_get")
+    def test_labels_extracted(self, mock_get):
+        mock_get.return_value = [
+            _make_api_issue(1, "Labeled", labels=["bug", "high-priority"]),
+        ]
+        items = fetch_all_open_items()
+        assert items[0]["labels"] == ["bug", "high-priority"]
+
+    @patch("sweep.MAX_ITEMS", 3)
+    @patch("sweep.github_api_get")
+    def test_max_items_cap(self, mock_get):
+        mock_get.return_value = [_make_api_issue(i) for i in range(100)]
+        items = fetch_all_open_items()
+        assert len(items) == 3
+
+    @patch("sweep.API_PAGE_SIZE", 2)
+    @patch("sweep.github_api_get")
+    def test_pagination(self, mock_get):
+        # First page: 2 items (full page), second page: 1 item (partial -> stop)
+        mock_get.side_effect = [
+            [_make_api_issue(1), _make_api_issue(2)],
+            [_make_api_issue(3)],
+        ]
+        items = fetch_all_open_items()
+        assert len(items) == 3
+        assert mock_get.call_count == 2
+
+
+class TestItemAge:
+    """Tests for _item_age helper."""
+
+    def test_recent_item(self):
+        from datetime import datetime, timezone, timedelta
+        recent = (datetime.now(timezone.utc) - timedelta(hours=12)).isoformat()
+        assert _item_age(recent) == "<1d"
+
+    def test_days_old(self):
+        from datetime import datetime, timezone, timedelta
+        old = (datetime.now(timezone.utc) - timedelta(days=15)).isoformat()
+        assert _item_age(old) == "15d"
+
+    def test_months_old(self):
+        from datetime import datetime, timezone, timedelta
+        old = (datetime.now(timezone.utc) - timedelta(days=90)).isoformat()
+        assert _item_age(old) == "3mo"
+
+    def test_years_old(self):
+        from datetime import datetime, timezone, timedelta
+        old = (datetime.now(timezone.utc) - timedelta(days=400)).isoformat()
+        assert _item_age(old) == "1y"
+
+    def test_invalid_date(self):
+        assert _item_age("not-a-date") == "?"
+
+
+class TestSuggestedAction:
+    """Tests for _suggested_action helper."""
+
+    def test_both_issues_close_newer(self):
+        a = TriageItem(
+            number=1, title="A", html_url="u", is_pr=False, labels=[],
+            created_at="2026-01-01T00:00:00Z", text="t",
+        )
+        b = TriageItem(
+            number=2, title="B", html_url="u", is_pr=False, labels=[],
+            created_at="2026-02-01T00:00:00Z", text="t",
+        )
+        result = _suggested_action(a, b)
+        assert "Close #2 as duplicate" in result
+
+    def test_both_prs_review(self):
+        a = TriageItem(
+            number=1, title="A", html_url="u", is_pr=True, labels=[],
+            created_at="2026-01-01T00:00:00Z", text="t",
+        )
+        b = TriageItem(
+            number=2, title="B", html_url="u", is_pr=True, labels=[],
+            created_at="2026-01-01T00:00:00Z", text="t",
+        )
+        assert _suggested_action(a, b) == "Review for overlap"
+
+    def test_issue_pr_link(self):
+        a = TriageItem(
+            number=1, title="A", html_url="u", is_pr=False, labels=[],
+            created_at="2026-01-01T00:00:00Z", text="t",
+        )
+        b = TriageItem(
+            number=2, title="B", html_url="u", is_pr=True, labels=[],
+            created_at="2026-01-01T00:00:00Z", text="t",
+        )
+        assert _suggested_action(a, b) == "Link PR to issue"
+
+
+class TestGenerateReport:
+    """Tests for the markdown report generator."""
+
+    def test_no_findings(self):
+        items = [
+            TriageItem(
+                number=1, title="Test", html_url="https://example.com/1",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text="Test",
+            ),
+        ]
+        report = generate_report(items, [], [])
+        assert "## Triage Sweep Report" in report
+        assert "Items analyzed:** 1" in report
+        assert "None found." in report
+        assert "0 outliers flagged" in report
+        assert "0 duplicate pairs found" in report
+
+    def test_health_summary_table(self):
+        items = [
+            TriageItem(
+                number=1, title="Test", html_url="https://example.com/1",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text="Test",
+            ),
+        ]
+        report = generate_report(items, [], [])
+        assert "### Health Summary" in report
+        assert "| Metric | Value |" in report
+        assert "| Items analyzed | 1 |" in report
+
+    def test_iqr_multiplier_in_thresholds(self):
+        """Report should show IQR multiplier, not percentile."""
+        items = [
+            TriageItem(
+                number=1, title="Test", html_url="https://example.com/1",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text="Test",
+            ),
+        ]
+        report = generate_report(items, [], [])
+        assert "IQR multiplier" in report
+        assert "percentile" not in report.lower().split("thresholds")[0]  # not in thresholds line
+
+    def test_with_outliers_shows_distance_and_age(self):
+        items = [
+            TriageItem(
+                number=10, title="Spam Issue", html_url="https://example.com/10",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text="spam",
+            ),
+            TriageItem(
+                number=20, title="Good Issue", html_url="https://example.com/20",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text="good",
+            ),
+        ]
+        report = generate_report(items, [(0, 12.34)], [])
+        assert "#10" in report
+        assert "Spam Issue" in report
+        assert "12.34" in report
+        assert "1 outliers flagged" in report
+        # Age column should be present
+        assert "| Age |" in report
+
+    def test_outlier_borderline_in_details(self):
+        """Borderline outliers should be in a <details> section."""
+        from embedding_utils import _OutlierResult
+        items = [
+            TriageItem(
+                number=10, title="Borderline", html_url="https://example.com/10",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text="spam",
+            ),
+        ]
+        # Create outlier results with cutoff=10.0, distance=12.0 (< 2*cutoff=20)
+        outlier_results = _OutlierResult([(0, 12.0)])
+        outlier_results.cutoff = 10.0
+        report = generate_report(items, outlier_results, [])
+        assert "<details>" in report
+        assert "Borderline" in report
+
+    def test_outlier_high_confidence(self):
+        """Items with distance > 2x cutoff should be in high confidence section."""
+        from embedding_utils import _OutlierResult
+        items = [
+            TriageItem(
+                number=10, title="Definite Spam", html_url="https://example.com/10",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text="spam",
+            ),
+        ]
+        outlier_results = _OutlierResult([(0, 25.0)])
+        outlier_results.cutoff = 10.0
+        report = generate_report(items, outlier_results, [])
+        assert "High Confidence" in report
+
+    def test_with_duplicates_suggested_action(self):
+        items = [
+            TriageItem(
+                number=1, title="First", html_url="https://example.com/1",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text="a",
+            ),
+            TriageItem(
+                number=2, title="Second", html_url="https://example.com/2",
+                is_pr=True, labels=[], created_at="2026-02-01T00:00:00Z", text="b",
+            ),
+        ]
+        report = generate_report(items, [], [(0, 1, 0.954)])
+        assert "#1" in report
+        assert "#2" in report
+        assert "0.954" in report
+        assert "1 duplicate pairs found" in report
+        assert "Suggested Action" in report
+        assert "Link PR to issue" in report
+
+    def test_duplicate_both_issues_close_newer(self):
+        items = [
+            TriageItem(
+                number=1, title="First", html_url="https://example.com/1",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text="a",
+            ),
+            TriageItem(
+                number=2, title="Second", html_url="https://example.com/2",
+                is_pr=False, labels=[], created_at="2026-02-01T00:00:00Z", text="b",
+            ),
+        ]
+        report = generate_report(items, [], [(0, 1, 0.95)])
+        assert "Close #2 as duplicate" in report
+
+    def test_duplicate_both_prs_review(self):
+        items = [
+            TriageItem(
+                number=1, title="PR A", html_url="https://example.com/1",
+                is_pr=True, labels=[], created_at="2026-01-01T00:00:00Z", text="a",
+            ),
+            TriageItem(
+                number=2, title="PR B", html_url="https://example.com/2",
+                is_pr=True, labels=[], created_at="2026-01-01T00:00:00Z", text="b",
+            ),
+        ]
+        report = generate_report(items, [], [(0, 1, 0.95)])
+        assert "Review for overlap" in report
+
+    def test_pr_type_label(self):
+        items = [
+            TriageItem(
+                number=5, title="PR Title", html_url="https://example.com/5",
+                is_pr=True, labels=[], created_at="2026-01-01T00:00:00Z", text="pr",
+            ),
+        ]
+        report = generate_report(items, [(0, 8.5)], [])
+        assert "| PR |" in report
+
+    def test_footer_present(self):
+        items = [
+            TriageItem(
+                number=1, title="T", html_url="u",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text="t",
+            ),
+        ]
+        report = generate_report(items, [], [])
+        assert "no LLM was used" in report
+
+
+class TestCreateReportIssue:
+    """Tests for creating the report GitHub issue."""
+
+    @patch("sweep.urllib.request.urlopen")
+    def test_successful_creation(self, mock_urlopen):
+        mock_resp = MagicMock()
+        mock_resp.status = 201
+        mock_resp.read.return_value = json.dumps({
+            "html_url": "https://github.com/owner/repo/issues/99",
+        }).encode()
+        mock_resp.__enter__ = lambda s: s
+        mock_resp.__exit__ = MagicMock(return_value=False)
+        mock_urlopen.return_value = mock_resp
+
+        # Should not raise
+        create_report_issue("# Test Report")
+
+    @patch("sweep.urllib.request.urlopen")
+    def test_http_error_exits(self, mock_urlopen):
+        error = HTTPError(
+            url="https://api.github.com/repos/owner/repo/issues",
+            code=422,
+            msg="Unprocessable",
+            hdrs=None,  # type: ignore[arg-type]
+            fp=BytesIO(b'{"message": "validation failed"}'),
+        )
+        mock_urlopen.side_effect = error
+
+        with pytest.raises(SystemExit) as exc_info:
+            create_report_issue("# Test Report")
+        assert exc_info.value.code == 1
+
+
+class TestWriteReport:
+    """Tests for the write_report helper."""
+
+    @patch("builtins.open", mock_open())
+    def test_writes_to_file(self):
+        write_report("# Report Content")
+        from builtins import open as builtin_open  # noqa
+        # Verify open was called with the right path
+        from unittest.mock import call
+        open_mock = open  # The patched version
+        open_mock.assert_called_once_with(REPORT_FILE, "w", encoding="utf-8")  # type: ignore[attr-defined]
+        open_mock().write.assert_called_once_with("# Report Content")  # type: ignore[attr-defined]
+
+
+class TestFetchRepoLabels:
+    """Tests for fetch_repo_labels."""
+
+    @patch("sweep.github_api_get")
+    def test_fetches_and_constructs_labels(self, mock_get):
+        mock_get.return_value = [
+            {"name": "bug", "description": "Something isn't working"},
+            {"name": "enhancement", "description": "New feature or request"},
+            {"name": "docs", "description": ""},
+        ]
+        labels = fetch_repo_labels()
+        assert len(labels) == 3
+        assert labels[0]["name"] == "bug"
+        assert labels[0]["text"] == "bug: Something isn't working"
+        assert labels[2]["text"] == "docs"  # no description, just name
+
+    @patch("sweep.github_api_get")
+    def test_empty_repo_labels(self, mock_get):
+        mock_get.return_value = []
+        labels = fetch_repo_labels()
+        assert labels == []
+
+    @patch("sweep.github_api_get")
+    def test_null_description_handled(self, mock_get):
+        mock_get.return_value = [
+            {"name": "wontfix", "description": None},
+        ]
+        labels = fetch_repo_labels()
+        assert labels[0]["text"] == "wontfix"
+
+    @patch("sweep.API_PAGE_SIZE", 2)
+    @patch("sweep.github_api_get")
+    def test_label_pagination(self, mock_get):
+        """Repos with more labels than one page should fetch all pages."""
+        mock_get.side_effect = [
+            # First page: full (2 items = API_PAGE_SIZE)
+            [
+                {"name": "bug", "description": "Broken"},
+                {"name": "feature", "description": "New"},
+            ],
+            # Second page: partial (1 item < API_PAGE_SIZE) -> stop
+            [
+                {"name": "docs", "description": "Documentation"},
+            ],
+        ]
+        labels = fetch_repo_labels()
+        assert len(labels) == 3
+        assert mock_get.call_count == 2
+        assert labels[0]["name"] == "bug"
+        assert labels[2]["name"] == "docs"
+
+
+class TestApplyLabelsToItem:
+    """Tests for apply_labels_to_item."""
+
+    def test_empty_labels_skips(self):
+        # Should not make any API call
+        apply_labels_to_item(1, [])
+
+    @patch("sweep.urllib.request.urlopen")
+    def test_successful_label_application(self, mock_urlopen):
+        mock_resp = MagicMock()
+        mock_resp.read.return_value = b'[{"name": "bug"}]'
+        mock_resp.__enter__ = lambda s: s
+        mock_resp.__exit__ = MagicMock(return_value=False)
+        mock_urlopen.return_value = mock_resp
+
+        # Should not raise
+        apply_labels_to_item(42, ["bug", "enhancement"])
+
+    @patch("sweep.urllib.request.urlopen")
+    def test_http_error_is_non_fatal(self, mock_urlopen):
+        error = HTTPError(
+            url="https://api.github.com/repos/owner/repo/issues/1/labels",
+            code=404,
+            msg="Not Found",
+            hdrs=None,  # type: ignore[arg-type]
+            fp=BytesIO(b'{"message": "not found"}'),
+        )
+        mock_urlopen.side_effect = error
+
+        # Should NOT raise — labeling failures are warnings, not fatal
+        apply_labels_to_item(1, ["bug"])
+
+
+class TestGenerateReportWithLabels:
+    """Tests for label suggestions in the report."""
+
+    def test_report_includes_label_section_high_confidence(self):
+        """High-confidence label (raw_sim >= 0.5) should appear in main table."""
+        items = [
+            TriageItem(
+                number=1, title="Fix crash", html_url="https://example.com/1",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text="crash",
+            ),
+        ]
+        suggestions = [[("bug", 0.85)]]
+        report = generate_report(items, [], [], label_suggestions=suggestions)
+        assert "Suggested Labels" in report
+        assert "`bug` (0.85)" in report
+        assert "1 items suggested for labeling" in report
+
+    def test_report_low_confidence_in_details(self):
+        """Low-confidence label (raw_sim < 0.5) should be in <details> section."""
+        items = [
+            TriageItem(
+                number=1, title="Something", html_url="https://example.com/1",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text="something",
+            ),
+        ]
+        suggestions = [[("maybe-bug", 0.35)]]
+        report = generate_report(items, [], [], label_suggestions=suggestions)
+        assert "Low-confidence suggestions" in report
+        assert "<details>" in report
+        assert "`maybe-bug` (0.35)" in report
+
+    def test_report_skips_already_labeled_items(self):
+        items = [
+            TriageItem(
+                number=1, title="Already labeled", html_url="https://example.com/1",
+                is_pr=False, labels=["bug"], created_at="2026-01-01T00:00:00Z", text="bug",
+            ),
+        ]
+        suggestions = [[("bug", 0.95)]]
+        report = generate_report(items, [], [], label_suggestions=suggestions)
+        assert "0 items suggested for labeling" in report
+        assert "No unlabeled items" in report
+
+    def test_report_excludes_outliers_from_suggestions(self):
+        items = [
+            TriageItem(
+                number=1, title="Spam garbage", html_url="https://example.com/1",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text="spam",
+            ),
+            TriageItem(
+                number=2, title="Real bug", html_url="https://example.com/2",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text="bug",
+            ),
+        ]
+        suggestions = [[("bug", 0.85)], [("bug", 0.90)]]
+        # Item 0 is an outlier (with distance) — should be excluded from label suggestions
+        report = generate_report(items, [(0, 15.2)], [], label_suggestions=suggestions)
+        assert "1 unlabeled items" in report  # only item 2
+        assert "#2" in report
+        # Item 0 (outlier) should NOT be in the suggestions table
+        assert "Spam garbage" not in report.split("Suggested Labels")[1]
+
+    def test_report_without_label_suggestions(self):
+        items = [
+            TriageItem(
+                number=1, title="T", html_url="u",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text="t",
+            ),
+        ]
+        report = generate_report(items, [], [], label_suggestions=None)
+        assert "Suggested Labels" not in report
+
+    def test_label_concentration_warning(self):
+        """When >50% of suggestions point to the same label, a warning should appear."""
+        items = [
+            TriageItem(
+                number=i, title=f"Item {i}", html_url=f"https://example.com/{i}",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text=f"text {i}",
+            )
+            for i in range(4)
+        ]
+        # 3 out of 4 items get "bug" label -> 75% concentration
+        suggestions = [
+            [("bug", 0.85)],
+            [("bug", 0.80)],
+            [("bug", 0.75)],
+            [("enhancement", 0.90)],
+        ]
+        report = generate_report(items, [], [], label_suggestions=suggestions)
+        assert "Warning" in report
+        assert "`bug`" in report
+        assert "3/4" in report
+
+
+class TestMain:
+    """Tests for the main orchestration function."""
+
+    @patch.dict(os.environ, {"GITHUB_TOKEN": "", "GITHUB_REPOSITORY": "owner/repo"})
+    def test_missing_token_exits(self):
+        with pytest.raises(SystemExit) as exc_info:
+            main()
+        assert exc_info.value.code == 1
+
+    @patch.dict(os.environ, {"GITHUB_TOKEN": "tok", "GITHUB_REPOSITORY": ""})
+    def test_missing_repo_exits(self):
+        with pytest.raises(SystemExit) as exc_info:
+            main()
+        assert exc_info.value.code == 1
+
+    @patch("sweep.write_report")
+    @patch("sweep.fetch_all_open_items", return_value=[])
+    def test_no_items(self, mock_fetch, mock_write):
+        main()
+        mock_write.assert_called_once()
+        report = mock_write.call_args[0][0]
+        assert "No open issues or PRs found" in report
+
+    @patch("sweep.create_report_issue")
+    @patch("sweep.write_report")
+    @patch("sweep.suggest_labels", return_value=[])
+    @patch("sweep.find_duplicate_pairs", return_value=[])
+    @patch("sweep.detect_outliers", return_value=[])
+    @patch("sweep.reduce_dimensions")
+    @patch("sweep.normalize_rows")
+    @patch("sweep.embed_texts")
+    @patch("sweep.fetch_repo_labels")
+    @patch("sweep.fetch_all_open_items")
+    def test_full_flow_with_enough_items(
+        self, mock_fetch, mock_labels, mock_embed, mock_norm, mock_reduce,
+        mock_outliers, mock_dupes, mock_suggest, mock_write, mock_create,
+    ):
+        """Test the full flow with >= MIN_SAMPLES items (outlier detection runs)."""
+        n = MIN_SAMPLES_FOR_OUTLIER_DETECTION
+        items = [
+            TriageItem(
+                number=i, title=f"Item {i}", html_url=f"https://example.com/{i}",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text=f"text {i}",
+            )
+            for i in range(n)
+        ]
+        mock_fetch.return_value = items
+        mock_labels.return_value = [
+            RepoLabel(name="bug", description="Something broken", text="bug: Something broken"),
+        ]
+
+        embeddings = np.random.randn(n, 384).astype(np.float32)
+        mock_embed.return_value = embeddings
+        mock_norm.return_value = embeddings
+        mock_reduce.return_value = np.random.randn(n, 10).astype(np.float32)
+
+        main()
+
+        mock_fetch.assert_called_once()
+        mock_labels.assert_called_once()
+        # embed_texts called twice: once for items, once for labels
+        assert mock_embed.call_count == 2
+        mock_norm.assert_called()
+        mock_reduce.assert_called_once()
+        mock_outliers.assert_called_once()
+        mock_dupes.assert_called_once()
+        mock_suggest.assert_called_once()
+        mock_write.assert_called_once()
+        mock_create.assert_called_once()
+
+    @patch("sweep.create_report_issue")
+    @patch("sweep.write_report")
+    @patch("sweep.suggest_labels", return_value=[])
+    @patch("sweep.find_duplicate_pairs", return_value=[])
+    @patch("sweep.detect_outliers")
+    @patch("sweep.reduce_dimensions")
+    @patch("sweep.normalize_rows")
+    @patch("sweep.embed_texts")
+    @patch("sweep.fetch_repo_labels", return_value=[])
+    @patch("sweep.fetch_all_open_items")
+    def test_skips_outlier_detection_for_few_items(
+        self, mock_fetch, mock_labels, mock_embed, mock_norm, mock_reduce,
+        mock_outliers, mock_dupes, mock_suggest, mock_write, mock_create,
+    ):
+        """With < MIN_SAMPLES items, outlier detection should be skipped."""
+        n = MIN_SAMPLES_FOR_OUTLIER_DETECTION - 1
+        items = [
+            TriageItem(
+                number=i, title=f"Item {i}", html_url=f"https://example.com/{i}",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text=f"text {i}",
+            )
+            for i in range(n)
+        ]
+        mock_fetch.return_value = items
+
+        embeddings = np.random.randn(n, 384).astype(np.float32)
+        mock_embed.return_value = embeddings
+        mock_norm.return_value = embeddings
+
+        main()
+
+        # Outlier detection should not have been called
+        mock_reduce.assert_not_called()
+        mock_outliers.assert_not_called()
+        # But duplicates should still be checked
+        mock_dupes.assert_called_once()
+
+    @patch.dict(os.environ, {"INPUT_DRY_RUN": "true"})
+    @patch("sweep.DRY_RUN", True)
+    @patch("sweep.write_report")
+    @patch("sweep.create_report_issue")
+    @patch("sweep.apply_labels_to_item")
+    @patch("sweep.suggest_labels", return_value=[[("bug", 0.85)]])
+    @patch("sweep.find_duplicate_pairs", return_value=[])
+    @patch("sweep.normalize_rows")
+    @patch("sweep.embed_texts")
+    @patch("sweep.fetch_repo_labels")
+    @patch("sweep.fetch_all_open_items")
+    def test_dry_run_skips_issue_creation_and_labeling(
+        self, mock_fetch, mock_labels, mock_embed, mock_norm,
+        mock_dupes, mock_suggest, mock_apply, mock_create, mock_write,
+    ):
+        items = [
+            TriageItem(
+                number=1, title="Item", html_url="https://example.com/1",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text="text",
+            )
+        ]
+        mock_fetch.return_value = items
+        mock_labels.return_value = [
+            RepoLabel(name="bug", description="Broken", text="bug: Broken"),
+        ]
+        embeddings = np.random.randn(1, 384).astype(np.float32)
+        mock_embed.return_value = embeddings
+        mock_norm.return_value = embeddings
+
+        main()
+
+        mock_create.assert_not_called()
+        mock_apply.assert_not_called()
+        mock_write.assert_called_once()
+
+    @patch("sweep.create_report_issue")
+    @patch("sweep.write_report")
+    @patch("sweep.apply_labels_to_item")
+    @patch("sweep.suggest_labels")
+    @patch("sweep.find_duplicate_pairs", return_value=[])
+    @patch("sweep.normalize_rows")
+    @patch("sweep.embed_texts")
+    @patch("sweep.fetch_repo_labels")
+    @patch("sweep.fetch_all_open_items")
+    def test_labels_not_auto_applied(
+        self, mock_fetch, mock_labels, mock_embed, mock_norm,
+        mock_dupes, mock_suggest, mock_apply, mock_write, mock_create,
+    ):
+        """Auto-labeling is disabled; labels should appear in report only."""
+        items = [
+            TriageItem(
+                number=1, title="Crash bug", html_url="https://example.com/1",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text="crash",
+            ),
+            TriageItem(
+                number=2, title="Already labeled", html_url="https://example.com/2",
+                is_pr=False, labels=["enhancement"], created_at="2026-01-01T00:00:00Z", text="feat",
+            ),
+        ]
+        mock_fetch.return_value = items
+        mock_labels.return_value = [
+            RepoLabel(name="bug", description="Broken", text="bug: Broken"),
+        ]
+        mock_suggest.return_value = [
+            [("bug", 0.90)],
+            [("bug", 0.45)],
+        ]
+
+        embeddings = np.random.randn(2, 384).astype(np.float32)
+        mock_embed.return_value = embeddings
+        mock_norm.return_value = embeddings
+
+        main()
+
+        # Auto-labeling is disabled — apply_labels_to_item should never be called
+        mock_apply.assert_not_called()
+
+    @patch("sweep.create_report_issue")
+    @patch("sweep.write_report")
+    @patch("sweep.apply_labels_to_item")
+    @patch("sweep.suggest_labels")
+    @patch("sweep.find_duplicate_pairs", return_value=[])
+    @patch("sweep.detect_outliers")
+    @patch("sweep.reduce_dimensions")
+    @patch("sweep.normalize_rows")
+    @patch("sweep.embed_texts")
+    @patch("sweep.fetch_repo_labels")
+    @patch("sweep.fetch_all_open_items")
+    def test_outliers_excluded_from_report_suggestions(
+        self, mock_fetch, mock_labels, mock_embed, mock_norm, mock_reduce,
+        mock_outliers, mock_dupes, mock_suggest, mock_apply, mock_write, mock_create,
+    ):
+        """Items flagged as outliers should not appear in report label suggestions."""
+        n = MIN_SAMPLES_FOR_OUTLIER_DETECTION
+        items = [
+            TriageItem(
+                number=i, title=f"Item {i}", html_url=f"https://example.com/{i}",
+                is_pr=False, labels=[], created_at="2026-01-01T00:00:00Z", text=f"text {i}",
+            )
+            for i in range(n)
+        ]
+        mock_fetch.return_value = items
+        mock_labels.return_value = [
+            RepoLabel(name="bug", description="Broken", text="bug: Broken"),
+        ]
+        mock_outliers.return_value = [(0, 12.5), (5, 15.3)]
+        mock_suggest.return_value = [[("bug", 0.85)] for _ in range(n)]
+
+        embeddings = np.random.randn(n, 384).astype(np.float32)
+        mock_embed.return_value = embeddings
+        mock_norm.return_value = embeddings
+        mock_reduce.return_value = np.random.randn(n, 10).astype(np.float32)
+
+        main()
+
+        # Auto-labeling is disabled
+        mock_apply.assert_not_called()
+        # Report should still be generated (outliers excluded from suggestions in report)
+        mock_write.assert_called_once()
+        report = mock_write.call_args[0][0]
+        # Outlier items 0 and 5 should not appear in the label suggestions section
+        assert "Item 0" not in report.split("Suggested Labels")[1] if "Suggested Labels" in report else True

.github/workflows/ci-e2e.yml

@@ -0,0 +1,83 @@
+name: E2E Tests
+
+on:
+  workflow_call:
+
+jobs:
+  check-changes:
+    name: Check web module changes
+    runs-on: ubuntu-latest
+    timeout-minutes: 5
+    outputs:
+      web_changed: ${{ steps.filter.outputs.web }}
+    steps:
+      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+      - uses: dorny/paths-filter@de90cc6fb38fc0963ad72b210f1f284cd68cea36 # v3
+        id: filter
+        with:
+          filters: |
+            web:
+              - 'gitnexus-web/**'
+
+  e2e:
+    name: e2e (chromium)
+    needs: check-changes
+    if: needs.check-changes.result == 'success' && needs.check-changes.outputs.web_changed == 'true'
+    runs-on: ubuntu-latest
+    timeout-minutes: 20
+    steps:
+      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+
+      - uses: ./.github/actions/setup-gitnexus-web
+
+      - name: Install Playwright browsers
+        run: npx playwright install --with-deps chromium
+        working-directory: gitnexus-web
+
+      - name: Install backend dependencies
+        run: npm ci
+        working-directory: gitnexus
+
+      - name: Build backend
+        run: npm run build
+        working-directory: gitnexus
+
+      - name: Analyze repository (index for backend)
+        run: |
+          node gitnexus/dist/cli/index.js analyze || true
+          if [ ! -d ".gitnexus" ]; then
+            echo "::error::No .gitnexus index created"
+            exit 1
+          fi
+
+      - name: Start backend server
+        run: node dist/cli/index.js serve &
+        working-directory: gitnexus
+
+      - name: Wait for backend readiness
+        run: npx wait-on http://localhost:4747/api/repos --timeout 30000
+        working-directory: gitnexus-web
+
+      - name: Start Vite dev server
+        run: npm run dev &
+        working-directory: gitnexus-web
+
+      - name: Wait for Vite dev server
+        run: npx wait-on http://localhost:5173 --timeout 30000
+        working-directory: gitnexus-web
+
+      - name: Run E2E tests
+        run: npx playwright test
+        working-directory: gitnexus-web
+        env:
+          E2E: '1'
+
+      - name: Upload test results
+        if: always()
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4
+        with:
+          name: e2e-results
+          path: |
+            gitnexus-web/test-results/
+            gitnexus-web/playwright-report/
+          retention-days: 5

.github/workflows/ci-quality.yml

@@ -0,0 +1,49 @@
+name: Quality Checks
+
+on:
+  workflow_call:
+
+jobs:
+  format:
+    runs-on: ubuntu-latest
+    timeout-minutes: 5
+    steps:
+      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+      - uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020 # v4
+        with:
+          node-version: 20
+          cache: npm
+          cache-dependency-path: package-lock.json
+      - run: npm ci
+      - run: npx prettier --check .
+
+  lint:
+    runs-on: ubuntu-latest
+    timeout-minutes: 10
+    steps:
+      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+      - uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020 # v4
+        with:
+          node-version: 20
+          cache: npm
+          cache-dependency-path: package-lock.json
+      - run: npm ci
+      - run: npx eslint .
+
+  typecheck:
+    runs-on: ubuntu-latest
+    timeout-minutes: 10
+    steps:
+      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+      - uses: ./.github/actions/setup-gitnexus
+      - run: npx tsc --noEmit
+        working-directory: gitnexus
+
+  typecheck-web:
+    runs-on: ubuntu-latest
+    timeout-minutes: 10
+    steps:
+      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+      - uses: ./.github/actions/setup-gitnexus-web
+      - run: npx tsc -b --noEmit
+        working-directory: gitnexus-web

.github/workflows/ci-report.yml

@@ -0,0 +1,413 @@
+name: CI Report
+
+# Triggered after the CI workflow completes.  Because workflow_run
+# always runs code from the *default branch*, it receives a read/write
+# GITHUB_TOKEN — even when the triggering PR comes from a fork.
+
+on:
+  workflow_run:
+    workflows: ['CI']
+    types: [completed]
+
+permissions:
+  actions: read # needed to list/download workflow run artifacts
+  contents: read # needed for sparse checkout of vitest.config.ts
+  pull-requests: write # needed to post sticky PR comment
+
+jobs:
+  pr-report:
+    name: PR Report
+    # Only run for pull-request CI runs
+    if: >-
+      github.event.workflow_run.event == 'pull_request' &&
+      github.event.workflow_run.conclusion != 'cancelled'
+    runs-on: ubuntu-latest
+    timeout-minutes: 5
+    steps:
+      # ── Download artifacts from the CI run ────────────────────────
+      - name: Download artifacts
+        uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7
+        with:
+          script: |
+            const fs = require('fs');
+            const path = require('path');
+            const runId = context.payload.workflow_run.id;
+
+            const allArtifacts = await github.rest.actions.listWorkflowRunArtifacts({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              run_id: runId,
+            });
+
+            async function downloadArtifact(name, dest) {
+              const match = allArtifacts.data.artifacts.find(a => a.name === name);
+              if (!match) {
+                core.warning(`Artifact "${name}" not found`);
+                return false;
+              }
+              const zip = await github.rest.actions.downloadArtifact({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                artifact_id: match.id,
+                archive_format: 'zip',
+              });
+              fs.mkdirSync(dest, { recursive: true });
+              fs.writeFileSync(path.join(dest, `${name}.zip`), Buffer.from(zip.data));
+              return true;
+            }
+
+            const temp = process.env.RUNNER_TEMP;
+            await downloadArtifact('pr-meta', path.join(temp, 'dl'));
+            await downloadArtifact('test-reports', path.join(temp, 'dl'));
+
+      - name: Extract artifacts
+        shell: bash
+        run: |
+          cd "$RUNNER_TEMP/dl"
+          # Extract each artifact into its own directory to avoid filename collisions
+          for z in *.zip; do
+            [ -f "$z" ] || continue
+            name="${z%.zip}"
+            mkdir -p "$RUNNER_TEMP/artifacts/$name"
+            unzip -o "$z" -d "$RUNNER_TEMP/artifacts/$name"
+          done
+
+      - name: Read PR metadata
+        id: meta
+        shell: bash
+        run: |
+          DIR="$RUNNER_TEMP/artifacts/pr-meta"
+          if [ ! -f "$DIR/pr_number" ]; then
+            echo "skip=true" >> "$GITHUB_OUTPUT"
+            echo "::warning::pr_number artifact missing — skipping report"
+            exit 0
+          fi
+
+          # Validate PR number is a positive integer (artifact comes from
+          # untrusted fork code, so treat contents defensively).
+          PR_NUM=$(cat "$DIR/pr_number" | tr -d '[:space:]')
+          if ! [[ "$PR_NUM" =~ ^[0-9]+$ ]]; then
+            echo "skip=true" >> "$GITHUB_OUTPUT"
+            echo "::error::Invalid PR number in artifact: '$PR_NUM'"
+            exit 0
+          fi
+
+          echo "skip=false"              >> "$GITHUB_OUTPUT"
+          echo "pr_number=$PR_NUM"       >> "$GITHUB_OUTPUT"
+          # Validate job-result strings against known GitHub Actions values.
+          # Artifact contents come from the PR workflow (potentially untrusted
+          # fork code), so we whitelist to prevent newline injection into
+          # GITHUB_OUTPUT.
+          validate_result() {
+            local val
+            val=$(cat "$1" | tr -d '[:space:]')
+            case "$val" in
+              success|failure|cancelled|skipped) echo "$val" ;;
+              *) echo "unknown" ;;
+            esac
+          }
+
+          echo "quality=$(validate_result "$DIR/quality_result")"        >> "$GITHUB_OUTPUT"
+          echo "tests=$(validate_result "$DIR/tests_result")"            >> "$GITHUB_OUTPUT"
+          echo "e2e=$(validate_result "$DIR/e2e_result")"                >> "$GITHUB_OUTPUT"
+
+      - name: Checkout (for vitest config)
+        if: steps.meta.outputs.skip != 'true'
+        uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+        with:
+          sparse-checkout: gitnexus/vitest.config.ts
+          sparse-checkout-cone-mode: false
+
+      # ── Fetch base branch coverage for delta reporting ───────────
+      - name: Fetch base branch coverage
+        if: steps.meta.outputs.skip != 'true'
+        id: base-coverage
+        uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7
+        with:
+          script: |
+            const fs = require('fs');
+            const path = require('path');
+
+            // Find the latest successful CI run on main
+            const runs = await github.rest.actions.listWorkflowRuns({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              workflow_id: 'ci.yml',
+              branch: 'main',
+              status: 'success',
+              per_page: 1,
+            });
+
+            if (runs.data.workflow_runs.length === 0) {
+              core.setOutput('found', 'false');
+              core.info('No successful main branch CI runs found');
+              return;
+            }
+
+            const mainRunId = runs.data.workflow_runs[0].id;
+            const artifacts = await github.rest.actions.listWorkflowRunArtifacts({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              run_id: mainRunId,
+            });
+
+            const testReports = artifacts.data.artifacts.find(a => a.name === 'test-reports');
+            if (!testReports) {
+              core.setOutput('found', 'false');
+              core.info('No test-reports artifact on main branch');
+              return;
+            }
+
+            const zip = await github.rest.actions.downloadArtifact({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              artifact_id: testReports.id,
+              archive_format: 'zip',
+            });
+
+            const dest = path.join(process.env.RUNNER_TEMP, 'base-coverage');
+            fs.mkdirSync(dest, { recursive: true });
+            fs.writeFileSync(path.join(dest, 'base.zip'), Buffer.from(zip.data));
+            core.setOutput('found', 'true');
+            core.setOutput('dir', dest);
+
+      - name: Extract base coverage
+        if: steps.meta.outputs.skip != 'true' && steps.base-coverage.outputs.found == 'true'
+        shell: bash
+        run: |
+          cd "${{ steps.base-coverage.outputs.dir }}"
+          mkdir -p base
+          unzip -o base.zip -d base
+
+      - name: Build report
+        if: steps.meta.outputs.skip != 'true'
+        id: report
+        shell: bash
+        env:
+          QUALITY: ${{ steps.meta.outputs.quality }}
+          TESTS: ${{ steps.meta.outputs.tests }}
+          E2E: ${{ steps.meta.outputs.e2e }}
+          BASE_FOUND: ${{ steps.base-coverage.outputs.found }}
+          BASE_DIR: ${{ steps.base-coverage.outputs.dir }}
+          RUN_URL: ${{ github.event.workflow_run.html_url }}
+        run: |
+          DIR="$RUNNER_TEMP/artifacts"
+
+          # ── Helper: read coverage summary into prefixed vars ──
+          read_cov() {
+            local prefix=$1 file=$2
+            if [ -n "$file" ] && [ -f "$file" ]; then
+              local val
+              val=$(jq -r '.total.statements.pct // "N/A"' "$file" 2>/dev/null) || val="N/A"
+              printf -v "${prefix}_STMTS" '%s' "$val"
+              val=$(jq -r '.total.branches.pct // "N/A"' "$file" 2>/dev/null) || val="N/A"
+              printf -v "${prefix}_BRANCH" '%s' "$val"
+              val=$(jq -r '.total.functions.pct // "N/A"' "$file" 2>/dev/null) || val="N/A"
+              printf -v "${prefix}_FUNCS" '%s' "$val"
+              val=$(jq -r '.total.lines.pct // "N/A"' "$file" 2>/dev/null) || val="N/A"
+              printf -v "${prefix}_LINES" '%s' "$val"
+              val=$(jq -r '"\(.total.statements.covered)/\(.total.statements.total)"' "$file" 2>/dev/null) || val=""
+              printf -v "${prefix}_STMTS_COV" '%s' "$val"
+              val=$(jq -r '"\(.total.branches.covered)/\(.total.branches.total)"' "$file" 2>/dev/null) || val=""
+              printf -v "${prefix}_BRANCH_COV" '%s' "$val"
+              val=$(jq -r '"\(.total.functions.covered)/\(.total.functions.total)"' "$file" 2>/dev/null) || val=""
+              printf -v "${prefix}_FUNCS_COV" '%s' "$val"
+              val=$(jq -r '"\(.total.lines.covered)/\(.total.lines.total)"' "$file" 2>/dev/null) || val=""
+              printf -v "${prefix}_LINES_COV" '%s' "$val"
+              return 0
+            else
+              printf -v "${prefix}_STMTS" '%s' "N/A"
+              printf -v "${prefix}_BRANCH" '%s' "N/A"
+              printf -v "${prefix}_FUNCS" '%s' "N/A"
+              printf -v "${prefix}_LINES" '%s' "N/A"
+              printf -v "${prefix}_STMTS_COV" '%s' ""
+              printf -v "${prefix}_BRANCH_COV" '%s' ""
+              printf -v "${prefix}_FUNCS_COV" '%s' ""
+              printf -v "${prefix}_LINES_COV" '%s' ""
+              return 1
+            fi
+          }
+
+          # ── Read coverage reports ──
+          UNIT_SUMMARY=$(find "$DIR/test-reports" -name "coverage-summary.json" -type f 2>/dev/null | head -1)
+
+          read_cov "U" "$UNIT_SUMMARY"
+
+          # ── Read base branch coverage (main) ──
+          BASE_SUMMARY=""
+          if [ "$BASE_FOUND" = "true" ] && [ -n "$BASE_DIR" ]; then
+            BASE_SUMMARY=$(find "$BASE_DIR/base" -name "coverage-summary.json" -type f 2>/dev/null | head -1)
+          fi
+          read_cov "B" "$BASE_SUMMARY"
+
+          # ── Locate test results ──
+          RESULTS_FILE=$(find "$DIR/test-reports" -name "test-results.json" -type f 2>/dev/null | head -1)
+          WEB_RESULTS_FILE=$(find "$DIR/test-reports" -name "web-test-results.json" -type f 2>/dev/null | head -1)
+
+          sum_results() {
+            local file=$1
+            if [ -n "$file" ] && [ -f "$file" ]; then
+              jq -r '"\(.numTotalTests) \(.numPassedTests) \(.numFailedTests) \(.numPendingTests) \(.numTotalTestSuites) \(((.testResults | map(.endTime) | max) - (.startTime)) / 1000 | floor)"' "$file" 2>/dev/null || echo "0 0 0 0 0 0"
+            else
+              echo "0 0 0 0 0 0"
+            fi
+          }
+
+          read CLI_T CLI_P CLI_F CLI_S CLI_SU CLI_D <<< "$(sum_results "$RESULTS_FILE")"
+          read WEB_T WEB_P WEB_F WEB_S WEB_SU WEB_D <<< "$(sum_results "$WEB_RESULTS_FILE")"
+
+          TOTAL=$((CLI_T + WEB_T))
+          PASSED=$((CLI_P + WEB_P))
+          FAILED=$((CLI_F + WEB_F))
+          SKIPPED=$((CLI_S + WEB_S))
+          SUITES=$((CLI_SU + WEB_SU))
+          DURATION=$((CLI_D > WEB_D ? CLI_D : WEB_D))
+
+          # ── Status helpers ──
+          status_icon() {
+            case "$1" in
+              success) echo "✅" ;;
+              failure) echo "❌" ;;
+              cancelled) echo "⏭️" ;;
+              *) echo "❓" ;;
+            esac
+          }
+
+          # Validate a value looks like a number (integer or decimal, optional
+          # leading minus).  Returns 1 for anything else — guards against awk
+          # injection when artifact values come from untrusted fork code.
+          is_numeric() { [[ "$1" =~ ^-?[0-9]+(\.[0-9]+)?$ ]]; }
+
+          cov_delta() {
+            local pct=$1 base=$2
+            if [ "$pct" = "N/A" ] || [ "$base" = "N/A" ]; then echo "—"; return; fi
+            if ! is_numeric "$pct" || ! is_numeric "$base"; then echo "—"; return; fi
+            local diff
+            diff=$(awk -v p="$pct" -v b="$base" 'BEGIN { printf "%.1f", p - b }')
+            if [ "$(awk -v p="$pct" -v b="$base" 'BEGIN { print (p > b) ? 1 : 0 }')" = "1" ]; then
+              echo "📈 +${diff}"
+            elif [ "$(awk -v p="$pct" -v b="$base" 'BEGIN { print (p < b) ? 1 : 0 }')" = "1" ]; then
+              echo "📉 ${diff}"
+            else
+              echo "＝ ${diff}"
+            fi
+          }
+
+          cov_bar() {
+            local pct=$1 base=$2
+            if [ "$pct" = "N/A" ] || ! is_numeric "$pct"; then echo "—"; return; fi
+            local filled
+            filled=$(awk -v p="$pct" 'BEGIN { printf "%d", p / 5 }')
+            (( filled < 0 )) && filled=0
+            (( filled > 20 )) && filled=20
+            local empty=$((20 - filled))
+            local bar=""
+            for ((i=0; i<filled; i++)); do bar+="█"; done
+            for ((i=0; i<empty; i++)); do bar+="░"; done
+            # Green if >= base (or base unavailable), red if dropped
+            if [ "$base" = "N/A" ] || ! is_numeric "$base" || [ "$(awk -v p="$pct" -v b="$base" 'BEGIN { print (p >= b) ? 1 : 0 }')" = "1" ]; then
+              echo "🟢 ${bar}"
+            else
+              echo "🔴 ${bar}"
+            fi
+          }
+
+          # ── Overall status ──
+          if [[ "$QUALITY" == "success" && "$TESTS" == "success" && ("$E2E" == "success" || "$E2E" == "skipped") ]]; then
+            OVERALL="✅ **All checks passed**"
+          else
+            OVERALL="❌ **Some checks failed**"
+          fi
+
+          # ── Build markdown ──
+          {
+            echo "body<<GITNEXUS_CI_REPORT_EOF_7f3a"
+            echo "## CI Report"
+            echo ""
+            echo "${OVERALL}"
+            echo ""
+            echo "### Pipeline Status"
+            echo ""
+            echo "| Stage | Status | Details |"
+            echo "|-------|--------|---------|"
+            echo "| $(status_icon "$QUALITY") Typecheck | \`${QUALITY}\` | tsc --noEmit |"
+            echo "| $(status_icon "$TESTS") Tests | \`${TESTS}\` | unit tests, 3 platforms |"
+            echo "| $(status_icon "$E2E") E2E | \`${E2E}\` | gitnexus-web changes only |"
+            echo ""
+
+            if [ "$TOTAL" -gt 0 ] 2>/dev/null; then
+              echo "### Test Results"
+              echo ""
+              echo "| Tests | Passed | Failed | Skipped | Duration |"
+              echo "|-------|--------|--------|---------|----------|"
+              echo "| ${TOTAL} | ${PASSED} | ${FAILED} | ${SKIPPED} | ${DURATION}s |"
+              echo ""
+
+              if [ "$FAILED" = "0" ]; then
+                echo "✅ All **${PASSED}** tests passed"
+              else
+                echo "❌ **${FAILED}** failed / **${PASSED}** passed"
+              fi
+              if [ "$SKIPPED" != "0" ]; then
+                echo ""
+                echo "<details>"
+                echo "<summary>${SKIPPED} test(s) skipped — expand for details</summary>"
+                echo ""
+                for rf in "$RESULTS_FILE" "$WEB_RESULTS_FILE"; do
+                  if [ -n "$rf" ] && [ -f "$rf" ]; then
+                    jq -r '
+                      .testResults[]
+                      | .assertionResults[]?
+                      | select(.status == "pending" or .status == "skipped")
+                      | "- \(.ancestorTitles | join(" > ")) > \(.title)"
+                    ' "$rf" 2>/dev/null || true
+                  fi
+                done
+                echo ""
+                echo "</details>"
+              fi
+              echo ""
+            fi
+
+            # ── Coverage table helper ──
+            cov_table() {
+              local label=$1 s=$2 b=$3 f=$4 l=$5 sc=$6 bc=$7 fc=$8 lc=$9
+              shift 9
+              local bs=$1 bb=$2 bf=$3 bl=$4
+              echo "#### ${label}"
+              echo ""
+              echo "| Metric | Coverage | Covered | Base | Delta | Status |"
+              echo "|--------|----------|---------|------|-------|--------|"
+              echo "| Statements | **${s}%** | ${sc} | ${bs}% | $(cov_delta "$s" "$bs") | $(cov_bar "$s" "$bs") |"
+              echo "| Branches | **${b}%** | ${bc} | ${bb}% | $(cov_delta "$b" "$bb") | $(cov_bar "$b" "$bb") |"
+              echo "| Functions | **${f}%** | ${fc} | ${bf}% | $(cov_delta "$f" "$bf") | $(cov_bar "$f" "$bf") |"
+              echo "| Lines | **${l}%** | ${lc} | ${bl}% | $(cov_delta "$l" "$bl") | $(cov_bar "$l" "$bl") |"
+              echo ""
+            }
+
+            if [ "$U_STMTS" != "N/A" ]; then
+              echo "### Code Coverage"
+              echo ""
+              cov_table "Tests" \
+                "$U_STMTS" "$U_BRANCH" "$U_FUNCS" "$U_LINES" \
+                "$U_STMTS_COV" "$U_BRANCH_COV" "$U_FUNCS_COV" "$U_LINES_COV" \
+                "$B_STMTS" "$B_BRANCH" "$B_FUNCS" "$B_LINES"
+            else
+              echo "### Code Coverage"
+              echo ""
+              echo "⚠️ Coverage data unavailable - check the [unit test job](${RUN_URL}) for details."
+              echo ""
+            fi
+
+            echo "---"
+            echo "<sub>📋 [View full run](${RUN_URL}) · Generated by CI</sub>"
+            echo "GITNEXUS_CI_REPORT_EOF_7f3a"
+          } >> "$GITHUB_OUTPUT"
+
+      - name: Comment on PR
+        if: steps.meta.outputs.skip != 'true'
+        uses: marocchino/sticky-pull-request-comment@773744901bac0e8cbb5a0dc842800d45e9b2b405 # v2
+        with:
+          header: ci-report
+          number: ${{ steps.meta.outputs.pr_number }}
+          message: ${{ steps.report.outputs.body }}

.github/workflows/ci-tests.yml

@@ -0,0 +1,71 @@
+name: Tests
+
+on:
+  workflow_call:
+
+jobs:
+  tests:
+    name: ubuntu / coverage
+    runs-on: ubuntu-latest
+    timeout-minutes: 25
+    steps:
+      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+      - uses: ./.github/actions/setup-gitnexus
+        with:
+          build: 'true'
+
+      - name: Run all tests with coverage
+        run: >-
+          npx vitest run
+          --reporter=default
+          --reporter=json
+          --outputFile=test-results.json
+          --coverage
+          --coverage.reporter=json-summary
+          --coverage.reporter=json
+          --coverage.reporter=text
+          --coverage.thresholdAutoUpdate=false
+          --coverage.reportOnFailure=true
+        working-directory: gitnexus
+
+      # gitnexus-shared already built by setup-gitnexus action above
+      - name: Install gitnexus-web dependencies
+        run: npm ci
+        working-directory: gitnexus-web
+
+      - name: Run gitnexus-web unit tests
+        run: >-
+          npx vitest run
+          --reporter=default
+          --reporter=json
+          --outputFile=web-test-results.json
+        working-directory: gitnexus-web
+
+      - name: Upload test reports
+        if: always()
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4
+        with:
+          name: test-reports
+          path: |
+            gitnexus/coverage/coverage-summary.json
+            gitnexus/coverage/coverage-final.json
+            gitnexus/test-results.json
+            gitnexus-web/web-test-results.json
+          retention-days: 5
+
+  cross-platform:
+    name: ${{ matrix.os }}
+    strategy:
+      fail-fast: false
+      matrix:
+        # Ubuntu already covered by the coverage job above
+        os: [windows-latest, macos-latest]
+    runs-on: ${{ matrix.os }}
+    timeout-minutes: 25
+    steps:
+      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+      - uses: ./.github/actions/setup-gitnexus
+        with:
+          build: 'true'
+      - run: npx vitest run
+        working-directory: gitnexus

.github/workflows/ci.yml

@@ -0,0 +1,110 @@
+name: CI
+
+on:
+  push:
+    branches: [main]
+    paths-ignore: ['**.md', 'docs/**', 'LICENSE']
+  pull_request:
+    branches: [main]
+    paths-ignore: ['**.md', 'docs/**', 'LICENSE']
+  workflow_call:
+
+concurrency:
+  group: ci-${{ github.ref }}
+  cancel-in-progress: true
+
+# ── Reusable workflow orchestration ─────────────────────────────────
+# Each concern lives in its own workflow file for maintainability:
+#   ci-quality.yml       — typecheck (tsc --noEmit)
+#   ci-tests.yml         — unit + integration tests with coverage + cross-platform
+#   ci-e2e.yml           — E2E tests (only when gitnexus-web/ changes)
+#
+# Shared setup is DRY via .github/actions/setup-gitnexus composite action.
+
+jobs:
+  quality:
+    uses: ./.github/workflows/ci-quality.yml
+    permissions:
+      contents: read
+
+  tests:
+    uses: ./.github/workflows/ci-tests.yml
+    permissions:
+      contents: read
+
+  e2e:
+    uses: ./.github/workflows/ci-e2e.yml
+    permissions:
+      contents: read
+
+  # ── Save PR metadata for the reporting workflow ─────────────────
+  # The ci-report.yml workflow (triggered by workflow_run) needs the
+  # PR number and job results to post a comment.  We save them as an
+  # artifact because workflow_run context doesn't reliably carry PR
+  # info for fork PRs.
+  save-pr-meta:
+    name: Save PR Metadata
+    if: always() && github.event_name == 'pull_request'
+    needs: [quality, tests, e2e]
+    runs-on: ubuntu-latest
+    timeout-minutes: 5
+    steps:
+      - name: Write metadata
+        shell: bash
+        env:
+          PR_NUMBER: ${{ github.event.number }}
+          QUALITY: ${{ needs.quality.result }}
+          TESTS: ${{ needs.tests.result }}
+          E2E: ${{ needs.e2e.result }}
+        run: |
+          mkdir -p pr-meta
+          echo "$PR_NUMBER"  > pr-meta/pr_number
+          echo "$QUALITY"    > pr-meta/quality_result
+          echo "$TESTS"      > pr-meta/tests_result
+          echo "$E2E"        > pr-meta/e2e_result
+          # TODO(post-merge): remove backward-compat copies once ci-report.yml
+          # on main reads underscore names.
+          # Backward-compat: ci-report.yml on main still reads hyphenated
+          # names.  workflow_run always executes from the default branch, so
+          # the main-branch reader won't find the underscore variants until
+          # this PR is merged.  Write both until then.
+          cp pr-meta/pr_number       pr-meta/pr-number
+          cp pr-meta/quality_result  pr-meta/quality-result
+          cp pr-meta/tests_result    pr-meta/tests-result
+          cp pr-meta/e2e_result      pr-meta/e2e-result
+
+      - name: Upload PR metadata
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4
+        with:
+          name: pr-meta
+          path: pr-meta/
+          retention-days: 1
+
+  # ── Unified CI gate ──────────────────────────────────────────────
+  # Single required check for branch protection.
+  ci-status:
+    name: CI Gate
+    needs: [quality, tests, e2e]
+    if: always()
+    runs-on: ubuntu-latest
+    timeout-minutes: 5
+    steps:
+      - name: Check all jobs passed
+        shell: bash
+        env:
+          QUALITY: ${{ needs.quality.result }}
+          TESTS: ${{ needs.tests.result }}
+          E2E: ${{ needs.e2e.result }}
+        run: |
+          echo "Quality:     $QUALITY"
+          echo "Tests:       $TESTS"
+          echo "E2E:         $E2E"
+          if [[ "$QUALITY" != "success" ]] ||
+             [[ "$TESTS" != "success" ]]; then
+            echo "::error::Quality or test jobs failed"
+            exit 1
+          fi
+          if [[ "$E2E" != "success" && "$E2E" != "skipped" ]]; then
+            echo "::error::E2E job failed"
+            exit 1
+          fi

.github/workflows/claude-code-review.yml

@@ -0,0 +1,95 @@
+name: Claude Code Review
+
+# Uses pull_request_target so the workflow runs as defined on the default branch,
+# which allows access to secrets for posting review comments on fork PRs.
+# SECURITY: The checkout pins the fork's HEAD SHA (not the branch name) to
+# prevent TOCTOU races (force-push between trigger and checkout). The
+# claude-code-action sandboxes execution — it does NOT run arbitrary code
+# from the checked-out source.
+
+on:
+  # Trigger only when explicitly requested:
+  #   - Add the "claude-review" label to a PR, OR
+  #   - Comment "@claude" or "/review" on a PR
+  pull_request_target:
+    types: [labeled]
+  issue_comment:
+    types: [created]
+
+# Serialize per-PR to avoid racing review comments.
+concurrency:
+  group: claude-review-${{ github.event.issue.number || github.event.pull_request.number }}
+  cancel-in-progress: false
+
+jobs:
+  claude-review:
+    # Run only when:
+    #   1. The "claude-review" label is added to a non-draft PR by a trusted contributor, OR
+    #   2. A trusted contributor comments "@claude" or "/review" on a PR
+    if: |
+      (
+        github.event_name == 'pull_request_target' &&
+        github.event.label.name == 'claude-review' &&
+        github.event.pull_request.draft == false &&
+        (github.event.pull_request.author_association == 'OWNER' ||
+         github.event.pull_request.author_association == 'MEMBER' ||
+         github.event.pull_request.author_association == 'COLLABORATOR')
+      ) ||
+      (
+        github.event_name == 'issue_comment' &&
+        github.event.issue.pull_request &&
+        (contains(github.event.comment.body, '@claude') ||
+         contains(github.event.comment.body, '/review')) &&
+        (github.event.comment.author_association == 'OWNER' ||
+         github.event.comment.author_association == 'MEMBER' ||
+         github.event.comment.author_association == 'COLLABORATOR')
+      )
+    runs-on: ubuntu-latest
+    timeout-minutes: 30
+    permissions:
+      contents: read
+      pull-requests: write
+      issues: read
+      id-token: write
+
+    steps:
+      # For issue_comment triggers, resolve the PR number, head SHA, and fork repo
+      - name: Resolve PR context
+        id: pr
+        uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7
+        with:
+          script: |
+            let pr;
+            if (context.eventName === 'issue_comment') {
+              const resp = await github.rest.pulls.get({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                pull_number: context.payload.issue.number,
+              });
+              pr = resp.data;
+            } else {
+              pr = context.payload.pull_request;
+            }
+            core.setOutput('number', pr.number);
+            core.setOutput('sha', pr.head.sha);
+            core.setOutput('repo', pr.head.repo.full_name);
+            core.setOutput('branch', pr.head.ref);
+
+      - name: Checkout PR head
+        uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+        with:
+          repository: ${{ steps.pr.outputs.repo }}
+          ref: ${{ steps.pr.outputs.sha }}
+          fetch-depth: 1
+
+      - name: Run Claude Code Review
+        id: claude-review
+        uses: anthropics/claude-code-action@9469d113c6afd29550c402740f22d1a97dd1209b # v1
+        with:
+          claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
+          github_token: ${{ secrets.GITHUB_TOKEN }}
+          allowed_non_write_users: '*'
+          show_full_output: true
+          plugin_marketplaces: 'https://github.com/anthropics/claude-code.git'
+          plugins: 'code-review@claude-code-plugins'
+          prompt: '/code-review:code-review ${{ github.repository }}/pull/${{ steps.pr.outputs.number }}'

.github/workflows/claude.yml

@@ -0,0 +1,110 @@
+name: Claude Code
+
+on:
+  issue_comment:
+    types: [created]
+  pull_request_review_comment:
+    types: [created]
+  issues:
+    types: [opened, assigned]
+  pull_request_review:
+    types: [submitted]
+
+# Serialize per-PR/issue to avoid racing comments.
+concurrency:
+  group: claude-code-${{ github.event.issue.number || github.event.pull_request.number || github.event.issue.id }}
+  cancel-in-progress: false
+
+jobs:
+  claude:
+    if: |
+      (
+        github.event_name == 'issue_comment' &&
+        contains(github.event.comment.body, '@claude') &&
+        (github.event.comment.author_association == 'OWNER' ||
+         github.event.comment.author_association == 'MEMBER' ||
+         github.event.comment.author_association == 'COLLABORATOR')
+      ) ||
+      (
+        github.event_name == 'pull_request_review_comment' &&
+        contains(github.event.comment.body, '@claude') &&
+        (github.event.comment.author_association == 'OWNER' ||
+         github.event.comment.author_association == 'MEMBER' ||
+         github.event.comment.author_association == 'COLLABORATOR')
+      ) ||
+      (
+        github.event_name == 'pull_request_review' &&
+        contains(github.event.review.body, '@claude') &&
+        (github.event.review.author_association == 'OWNER' ||
+         github.event.review.author_association == 'MEMBER' ||
+         github.event.review.author_association == 'COLLABORATOR')
+      ) ||
+      (
+        github.event_name == 'issues' &&
+        (contains(github.event.issue.body, '@claude') || contains(github.event.issue.title, '@claude')) &&
+        (github.event.issue.author_association == 'OWNER' ||
+         github.event.issue.author_association == 'MEMBER' ||
+         github.event.issue.author_association == 'COLLABORATOR')
+      )
+    runs-on: ubuntu-latest
+    timeout-minutes: 30
+    permissions:
+      contents: read
+      pull-requests: write
+      issues: write
+      id-token: write
+      actions: read # required for Claude to read CI results on PRs
+    steps:
+      # For PR-related triggers, resolve the fork repo so we can checkout correctly.
+      - name: Resolve PR context
+        id: pr
+        uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7
+        with:
+          script: |
+            // Determine if this event is PR-related
+            let prNumber = null;
+            if (context.eventName === 'issue_comment' && context.payload.issue.pull_request) {
+              prNumber = context.payload.issue.number;
+            } else if (context.eventName === 'pull_request_review_comment') {
+              prNumber = context.payload.pull_request.number;
+            } else if (context.eventName === 'pull_request_review') {
+              prNumber = context.payload.pull_request.number;
+            }
+
+            if (!prNumber) {
+              core.setOutput('is_pr', 'false');
+              return;
+            }
+
+            const resp = await github.rest.pulls.get({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              pull_number: prNumber,
+            });
+            const pr = resp.data;
+
+            core.setOutput('is_pr', 'true');
+            core.setOutput('number', String(prNumber));
+            core.setOutput('sha', pr.head.sha);
+            core.setOutput('repo', pr.head.repo.full_name);
+            core.setOutput('branch', pr.head.ref);
+
+      - name: Checkout repository
+        uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+        with:
+          repository: ${{ steps.pr.outputs.is_pr == 'true' && steps.pr.outputs.repo || github.repository }}
+          ref: ${{ steps.pr.outputs.is_pr == 'true' && steps.pr.outputs.sha || '' }}
+          fetch-depth: 1
+
+      - name: Run Claude Code
+        id: claude
+        uses: anthropics/claude-code-action@9469d113c6afd29550c402740f22d1a97dd1209b # v1
+        with:
+          claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
+          github_token: ${{ secrets.GITHUB_TOKEN }}
+          allowed_non_write_users: '*'
+          show_full_output: true
+
+          # This is an optional setting that allows Claude to read CI results on PRs
+          additional_permissions: |
+            actions: read

.github/workflows/pr-description-check.yml

@@ -0,0 +1,94 @@
+name: PR Description Check
+
+on:
+  pull_request:
+    types: [opened, edited, reopened]
+    branches: [main]
+
+permissions:
+  pull-requests: write
+
+concurrency:
+  group: pr-desc-${{ github.event.pull_request.number }}
+  cancel-in-progress: true
+
+jobs:
+  check-description:
+    runs-on: ubuntu-latest
+    timeout-minutes: 5
+    steps:
+      - name: Check PR description quality
+        uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd # v8
+        with:
+          script: |
+            const MIN_BODY_LENGTH = 50;
+            const LABEL = 'needs-description';
+
+            const pr = context.payload.pull_request;
+            const body = (pr.body || '').trim();
+            const owner = context.repo.owner;
+            const repo = context.repo.repo;
+            const number = pr.number;
+
+            const hasLabel = pr.labels.some(l => l.name === LABEL);
+
+            if (body.length < MIN_BODY_LENGTH) {
+              // Add label if not already present
+              if (!hasLabel) {
+                await github.rest.issues.addLabels({
+                  owner, repo, issue_number: number,
+                  labels: [LABEL],
+                });
+              }
+
+              // Post or update a comment
+              const marker = '<!-- pr-desc-check -->';
+              const message = [
+                marker,
+                `### PR description is too short`,
+                '',
+                `This PR's description is **${body.length}** characters, ` +
+                `but the minimum is **${MIN_BODY_LENGTH}**.`,
+                '',
+                'Please update the PR description to explain:',
+                '- **What** this PR changes',
+                '- **Why** the change is needed',
+                '',
+                'Use the PR template as a guide. This check will re-run when you edit the description.',
+              ].join('\n');
+
+              // Find existing bot comment to update (avoid spam)
+              const comments = await github.rest.issues.listComments({
+                owner, repo, issue_number: number,
+              });
+              const existing = comments.data.find(c =>
+                c.body && c.body.includes(marker)
+              );
+
+              if (existing) {
+                await github.rest.issues.updateComment({
+                  owner, repo, comment_id: existing.id,
+                  body: message,
+                });
+              } else {
+                await github.rest.issues.createComment({
+                  owner, repo, issue_number: number,
+                  body: message,
+                });
+              }
+
+              core.setFailed(
+                `PR description is ${body.length} chars (minimum: ${MIN_BODY_LENGTH})`
+              );
+            } else {
+              // Description is acceptable — remove the label if present
+              if (hasLabel) {
+                await github.rest.issues.removeLabel({
+                  owner, repo, issue_number: number,
+                  name: LABEL,
+                }).catch(() => {});
+                // .catch: label may have been removed manually
+              }
+
+              core.info(`PR description OK (${body.length} chars)`);
+            }

.github/workflows/publish.yml

@@ -0,0 +1,73 @@
+name: Publish to npm
+
+on:
+  push:
+    tags:
+      - 'v*'
+
+# No workflow-level permissions — scoped per job below.
+
+jobs:
+  ci:
+    uses: ./.github/workflows/ci.yml
+    permissions:
+      contents: read
+      actions: read
+      pull-requests: write
+
+  publish:
+    needs: ci
+    runs-on: ubuntu-latest
+    timeout-minutes: 15
+    permissions:
+      contents: write
+      id-token: write
+    steps:
+      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
+      - uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020 # v4
+        with:
+          node-version: 20
+          registry-url: https://registry.npmjs.org
+          cache: npm
+          cache-dependency-path: gitnexus/package-lock.json
+      - name: Build gitnexus-shared
+        run: npm install && npm run build
+        working-directory: gitnexus-shared
+
+      - run: npm ci
+        working-directory: gitnexus
+
+      - name: Verify version consistency
+        shell: bash
+        run: |
+          TAG_VERSION="${GITHUB_REF#refs/tags/v}"
+          if ! [[ "$TAG_VERSION" =~ ^[0-9]+\.[0-9]+\.[0-9]+(-[a-zA-Z0-9.]+)?$ ]]; then
+            echo "::error::Tag does not follow semver: v$TAG_VERSION"
+            exit 1
+          fi
+          PKG_VERSION=$(node -p "require('./package.json').version")
+          if [ "$TAG_VERSION" != "$PKG_VERSION" ]; then
+            echo "::error::Tag version (v$TAG_VERSION) does not match package.json version ($PKG_VERSION)"
+            exit 1
+          fi
+          echo "Version verified: $PKG_VERSION"
+        working-directory: gitnexus
+
+      - name: Build
+        run: npm run build
+        working-directory: gitnexus
+
+      - name: Dry-run publish
+        run: npm publish --dry-run
+        working-directory: gitnexus
+
+      - name: Publish to npm
+        run: npm publish --provenance --access public
+        working-directory: gitnexus
+        env:
+          NODE_AUTH_TOKEN: ${{ secrets.NPM_TOKEN }}
+
+      - name: Create GitHub Release
+        uses: softprops/action-gh-release@a06a81a03ee405af7f2048a818ed3f03bbf83c7b # v2
+        with:
+          generate_release_notes: true

.github/workflows/triage-sweep.yml

@@ -0,0 +1,103 @@
+name: Triage Sweep
+
+on:
+  workflow_dispatch:
+    inputs:
+      iqr_multiplier:
+        description: >-
+          IQR multiplier for outlier cutoff.
+          cutoff = Q75 + multiplier * IQR.
+          Higher = fewer outliers flagged.
+        type: number
+        default: 3.0
+      max_outlier_pct:
+        description: >-
+          Maximum fraction of items that can be flagged as outliers (0-1).
+          Hard cap to prevent over-flagging.
+        type: number
+        default: 0.05
+      contamination:
+        description: >-
+          Expected fraction of outliers in the data (0-0.5).
+          Controls how aggressively EllipticEnvelope downweights extremes.
+        type: number
+        default: 0.1
+      cosine_threshold:
+        description: >-
+          Cosine similarity threshold for duplicate detection.
+          Pairs with similarity above this are flagged as potential duplicates.
+          Higher = only very similar pairs flagged.
+        type: number
+        default: 0.92
+      max_items:
+        description: >-
+          Maximum number of open issues + PRs to process.
+          Hard cap to prevent runaway costs on very large repos.
+        type: number
+        default: 500
+      dry_run:
+        description: >-
+          Check this to only log results to the workflow summary.
+          Uncheck to create a GitHub issue with the report and apply labels.
+        type: boolean
+        default: false
+
+permissions:
+  contents: read
+  issues: write
+  pull-requests: write
+
+concurrency:
+  group: triage-sweep
+  cancel-in-progress: true
+
+jobs:
+  sweep:
+    runs-on: ubuntu-latest
+    timeout-minutes: 30
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6
+        with:
+          sparse-checkout: .github/scripts/triage
+          sparse-checkout-cone-mode: false
+          fetch-depth: 1
+
+      - name: Set up Python
+        uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6
+        with:
+          python-version: '3.12'
+          cache: pip
+          cache-dependency-path: .github/scripts/triage/requirements.txt
+
+      - name: Install dependencies
+        run: pip install -r .github/scripts/triage/requirements.txt
+
+      - name: Cache FastEmbed model weights
+        uses: actions/cache@668228422ae6a00e4ad889ee87cd7109ec5666a7 # v5
+        with:
+          path: ${{ github.workspace }}/.fastembed_cache
+          key: fastembed-bge-small-en-v1.5
+
+      - name: Run triage sweep
+        id: sweep
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          GITHUB_REPOSITORY: ${{ github.repository }}
+          FASTEMBED_CACHE_PATH: ${{ github.workspace }}/.fastembed_cache
+          INPUT_IQR_MULTIPLIER: ${{ inputs.iqr_multiplier }}
+          INPUT_MAX_OUTLIER_PCT: ${{ inputs.max_outlier_pct }}
+          INPUT_CONTAMINATION: ${{ inputs.contamination }}
+          INPUT_COSINE_THRESHOLD: ${{ inputs.cosine_threshold }}
+          INPUT_MAX_ITEMS: ${{ inputs.max_items }}
+          INPUT_DRY_RUN: ${{ inputs.dry_run }}
+        run: python .github/scripts/triage/sweep.py
+
+      - name: Post summary
+        if: always()
+        run: |
+          if [ -f /tmp/triage-report.md ]; then
+            cat /tmp/triage-report.md >> "$GITHUB_STEP_SUMMARY"
+          else
+            echo "No report generated." >> "$GITHUB_STEP_SUMMARY"
+          fi

.gitignore

@@ -1,46 +1,98 @@
+# Dependencies
+node_modules/
+
+# Build output
+dist/
+
+# TypeScript build info
+*.tsbuildinfo
+
+# IDE
+.vscode/
+.idea/
+*.swp
+*.swo
+
+# OS
+.DS_Store
+Thumbs.db
+
+.claude/settings.local.json
+
+# Environment variables
+.env
+.env.local
+.env.*.local
+
 # Logs
-logs
 *.log
 npm-debug.log*
-yarn-debug.log*
-yarn-error.log*
-pnpm-debug.log*
-lerna-debug.log*
 
-node_modules
-dist
-dist-ssr
+# Testing
+coverage/
+
+# Misc
 *.local
+HANDOFF.md
+HANDOFF*.md
 
-# Auto-generated files
-public/workers/compiled-queries.js
+.vercel
 
-# Editor directories and files
-.vscode/*
-!.vscode/extensions.json
-.idea
-.DS_Store
-*.suo
-*.ntvs*
-*.njsproj
-*.sln
-*.sw?
 
-# AI/Development tool directories
-.kilocode/
-.gemini/
-.cursor/
-.clinerules/
-.qoder/
 
-# Large JSON files
-gitnexus-project_*.json
-*-project_*.json
-.clinerules/byterover-rules.md
-.kilocode/rules/byterover-rules.md
-.roo/rules/byterover-rules.md
-.windsurf/rules/byterover-rules.md
-.cursor/rules/byterover-rules.mdc
-.kiro/steering/byterover-rules.md
-.qoder/rules/byterover-rules.md
-.augment/rules/byterover-rules.md
+
+
+
+.env*.local
+.gitnexus
+.claude/settings.local.json
+
+# Claude Code worktrees
+.claude/worktrees/
+
+# Claude code skills
+.claude/skills/generated/
+
+# Assets (screenshots, images)
+assets/
+
+# Generated files (should not be indexed)
+repomix-output*
+
+# Playwright artifacts
+gitnexus-web/playwright-report/
+gitnexus-web/test-results/
+
+# Python test artifacts
+eval/.coverage
+eval/.hypothesis/
+
+# Design docs (local only)
+docs/plans/
+
+gitnexus/test/fixtures/mini-repo/*.md
+gitnexus/test/fixtures/mini-repo/.claude
+gitnexus/test/fixtures/mini-repo/.gitignore
+
+# Ignore csharp generated obj and bin folders
+gitnexus/test/fixtures/lang-resolution/**/obj
+gitnexus/test/fixtures/lang-resolution/**/bin
+GitNexus.sln
+# Git worktrees
+.worktrees/
+
+/github/scripts/triage/__pycache__/
+
+.claude-flow/
+
+.claude/agents/
+.claude/commands/
+.claude/helpers
+.claude/skills/
+!.claude/skills/gitnexus/
+
+.history/
+
+.swarm/
+
+local_docs/

.history/gitnexus/vitest.config_20260317171253.ts

@@ -0,0 +1,33 @@
+import { defineConfig } from 'vitest/config';
+
+export default defineConfig({
+  test: {
+    globalSetup: ['test/global-setup.ts'],
+    include: ['test/**/*.test.ts'],
+    testTimeout: 30000,
+    hookTimeout: 120000,
+    pool: 'forks',
+    globals: true,
+    setupFiles: ['test/setup.ts'],
+    teardownTimeout: 3000,
+    dangerouslyIgnoreUnhandledErrors: true, // LadybugDB N-API destructor segfaults on fork exit — not a test failure
+    coverage: {
+      provider: 'v8',
+      include: ['src/**/*.ts'],
+      exclude: [
+        'src/cli/index.ts',          // CLI entry point (commander wiring)
+        'src/server/**',              // HTTP server (requires network)
+        'src/core/wiki/**',           // Wiki generation (requires LLM)
+      ],
+      // Auto-ratchet: vitest bumps thresholds when coverage exceeds them.
+      // CI will fail if a PR drops below these floors.
+      thresholds: {
+        statements: 26,
+        branches: 23,
+        functions: 28,
+        lines: 27,
+        autoUpdate: true,
+      },
+    },
+  },
+});

.husky/pre-commit

@@ -0,0 +1,26 @@
+#!/usr/bin/env bash
+# Pre-commit hook: format staged files + typecheck.
+# Tests run in CI (ci-tests.yml), not here.
+# Skip with: git commit --no-verify
+
+ROOT="$(git rev-parse --show-toplevel)"
+
+# 1. Format staged files with prettier via lint-staged
+echo "pre-commit: formatting staged files..."
+"$ROOT/node_modules/.bin/lint-staged" || exit 1
+
+# 2. Typecheck changed packages
+WEB_CHANGED=$(git diff --cached --name-only -- 'gitnexus-web/' | head -1)
+CLI_CHANGED=$(git diff --cached --name-only -- 'gitnexus/' | head -1)
+
+if [ -n "$WEB_CHANGED" ]; then
+  echo "pre-commit: typechecking gitnexus-web (tsc -b)..."
+  cd "$ROOT/gitnexus-web" && ./node_modules/.bin/tsc -b --noEmit || exit 1
+fi
+
+if [ -n "$CLI_CHANGED" ]; then
+  echo "pre-commit: typechecking gitnexus..."
+  cd "$ROOT/gitnexus" && ./node_modules/.bin/tsc --noEmit || exit 1
+fi
+
+echo "pre-commit: all checks passed"

.mcp.json

@@ -0,0 +1,9 @@
+{
+  "mcpServers": {
+    "gitnexus": {
+      "type": "stdio",
+      "command": "npx",
+      "args": ["-y", "gitnexus@latest", "mcp"]
+    }
+  }
+}

.prettierignore

@@ -0,0 +1,16 @@
+dist/
+coverage/
+gitnexus/vendor/
+gitnexus/test/fixtures/
+gitnexus-web/playwright-report/
+gitnexus-web/test-results/
+*.d.ts
+*.snap
+*.wasm
+*.md
+.gitnexus/
+.vercel/
+.claude-flow/
+.swarm/
+assets/
+repomix-output*

.prettierrc

@@ -0,0 +1,10 @@
+{
+  "semi": true,
+  "singleQuote": true,
+  "trailingComma": "all",
+  "printWidth": 100,
+  "tabWidth": 2,
+  "endOfLine": "lf",
+  "plugins": ["prettier-plugin-tailwindcss"],
+  "tailwindStylesheet": "./gitnexus-web/src/index.css"
+}

.sisyphus/drafts/gitnexus-brainstorming.md

@@ -0,0 +1,18 @@
+# Draft: Gitnexus Brainstorming - Clustering & Process Maps
+
+## Initial Context
+- Project: **GitnexusV2**
+- Structure: 
+    - `gitnexus/` (Likely the core application)
+    - `gitnexus-mcp/` (Likely a Model Context Protocol server)
+- Goal: Make it accurate and usable for smaller/dumber models.
+- Current Focus: Implementing **Clustering** and **Process Maps**.
+
+## Findings
+- **Clustering**: Found `gitnexus/src/core/ingestion/cluster-enricher.ts`.
+- **Process Maps**: No files matched `*process*map*` yet. Searching content next.
+
+## Open Questions
+- How is "process map" defined in this context? (Graph, mermaid diagram, flowchart?)
+- What is the input for clustering? (Code chunks, files, commits?)
+- What is the intended output for "smaller models"? (Simplified context, summaries?)

.sisyphus/drafts/noodlbox-comparison.md

@@ -0,0 +1,34 @@
+# Draft: Gitnexus vs Noodlbox Strategy
+
+## Objectives
+- Understand GitnexusV2 current state and goals.
+- Analyze Noodlbox capabilities from provided URL.
+- Compare features, architecture, and value proposition.
+- Provide strategic views and recommendations.
+
+## Research Findings
+- [GitnexusV2]: Zero-server, browser-native (WASM), KuzuDB based. Graph + Vector hybrid search.
+- [Noodlbox]: CLI-first, heavy install. Has "Session Hooks" and "Search Hooks" via plugins/CLI.
+
+## Comparison Points
+- **Core Philosophy**: Both bet on "Knowledge Graph + MCP" as the future. Noodlbox validates Gitnexus's direction.
+- **Architecture**:
+  - *Noodlbox*: CLI/Binary based. Likely local server management.
+  - *Gitnexus*: Zero-server, Browser-native (WASM). Lower friction, higher privacy.
+- **Features**:
+  - *Communities/Processes*: Both have them. Noodlbox uses them for "context injection". Gitnexus uses them for "visual exploration + query".
+  - *Impact Analysis*: Noodlbox has polished workflows (e.g., `detect_impact staged`). Gitnexus has the engine (`blastRadius`) but maybe not the specific workflow wrappers yet.
+- **UX/Integration**:
+  - *Noodlbox*: "Hooks" (Session/Search) are a killer feature. Proactively injecting context into the agent's session.
+  - *Gitnexus*: Powerful tools, but relies on agent *pulling* data?
+
+## Strategic Views
+1. **Validation**: The market direction is confirmed. You are building the right thing.
+2. **differentiation**: Lean into "Zero-Setup / Browser-Native". Noodlbox requires `noodl init` and CLI handling. Gitnexus could just *be*.
+3. **Opportunity**: Steal the "Session/Search Hooks" pattern. Make the agent smarter *automatically* without the user asking "check impact".
+4. **Workflow Polish**: Noodlbox's `/detect_impact staged` is a great specific use case. Gitnexus should wrap `blastRadius` into similar concrete workflows.
+
+## Technical Feasibility (Interception)
+- **Cursor**: Use `.cursorrules` to "shadow" default tools. Instruct agent to ALWAYS use `gitnexus_search` instead of `grep`.
+- **Claude Code**: Likely uses a private plugin API for `PreToolUse`. We can't match this exactly without an official plugin, but we can approximate it with strong prompt instructions in `AGENTS.md`.
+- **MCP Shadowing**: Define tools with names that conflict (e.g., `grep`)? No, unsafe. Better to use "Virtual Hooks" via system prompt instructions.

.windsurfrules

@@ -0,0 +1,5 @@
+# AI Agent Rules
+
+Follow .gitnexus/RULES.md for all project context and coding guidelines.
+
+This project uses GitNexus MCP for code intelligence. See .gitnexus/RULES.md for available tools and best practices.

AGENTS.md

@@ -0,0 +1,204 @@
+<!-- version: 1.2.0 -->
+<!--
+  Metadata: version, last reviewed, scope, model policy, reference docs, changelog.
+  Last updated: 2026-03-22
+-->
+
+Last reviewed: 2026-03-24
+
+**Project:** GitNexus · **Environment:** dev · **Maintainer:** repository maintainers (see GitHub)
+
+This file uses a standard agent header (version, scope, model policy, reference docs, changelog), adapted for this **TypeScript/JavaScript monorepo**.
+
+## Scope
+
+| | |
+|--|--|
+| **Reads** | Repository tree as needed for the task: `gitnexus/`, `gitnexus-web/`, `eval/`, plugin packages, `.github/`, `.gitnexus/` when present, and docs. |
+| **Writes** | Only paths required for the requested change; keep diffs minimal. Update lockfiles when dependencies change. |
+| **Executes** | `npm`, `npx`, `node` under `gitnexus/` and `gitnexus-web/`; `uv run` for Python under `eval/` when applicable; shell utilities for documented CI/dev workflows. |
+| **Off-limits** | User secrets (e.g. real `.env`), production deployment credentials, unrelated repositories, destructive git history operations without explicit human confirmation. |
+
+## Model Configuration
+
+- **Primary:** Pin in **Cursor** (Settings → model). Use a **named** model (e.g. GPT-5.2, Claude Sonnet 4.x). Avoid relying on **Auto** when reproducibility or audit trail matters.
+- **Fallback:** As configured in Cursor or your organization (do not encode `latest` or wildcards in automation configs).
+- **Notes:** The open-source GitNexus CLI indexer does not call an LLM. Optional Nexus AI in the web UI uses end-user provider keys and models.
+
+## Execution Sequence (complex tasks)
+
+Long sessions dilute instructions. For **multi-step** work, state up front:
+
+1. Which rules in this file and **[GUARDRAILS.md](GUARDRAILS.md)** apply (and any relevant Signs).
+2. Current **Scope** boundaries (Reads / Writes / Off-limits).
+3. Which **validation commands** you will run (e.g. `cd gitnexus && npm test`, `npx tsc --noEmit`).
+
+On very long threads, the human may add *“Remember: apply all AGENTS.md rules”* to re-weight rule tokens against context dilution.
+
+## Claude Code hooks
+
+Hooks enforce gates that prompts cannot. In **Claude Code**, **PreToolUse** hooks can block tools such as `git_commit` until checks pass. Adapt to this repo: e.g. `cd gitnexus && npm test` before commit.
+
+## Context budget (Cursor / standards)
+
+Generic “core standards” playbooks are often long and stack-specific. For this monorepo, commands and gotchas live under **Cursor Cloud specific instructions** below and in **[CONTRIBUTING.md](CONTRIBUTING.md)**. If always-on rules grow, split domain rules into **`.cursor/rules/*.mdc`** (globs). **Cursor:** project-wide rules live in **`.cursor/index.mdc`** (YAML frontmatter with `alwaysApply: true`). **Claude Code:** optionally load a **`STANDARDS.md`** only when needed (e.g. *“When writing new code, read STANDARDS.md”*) to save context.
+
+## Reference Documentation
+
+- **This repository:** **[ARCHITECTURE.md](ARCHITECTURE.md)**, **[CONTRIBUTING.md](CONTRIBUTING.md)**, **[GUARDRAILS.md](GUARDRAILS.md)**.
+- **Cursor:** `.cursor/index.mdc` (always-on rules); optional `.cursor/rules/*.mdc` (glob-scoped). Legacy `.cursorrules` is deprecated — see `.cursor/index.mdc`.
+- **Optional local files:** `NOTES.md` (short vendor-neutral project snapshot). For handoffs, keep notes local (e.g., a scratch file outside the repo) rather than committing `HANDOFF.md`.
+- **GitNexus:** skills under `.claude/skills/gitnexus/`; machine-oriented rules in the `gitnexus:start` … `gitnexus:end` block below.
+
+## Changelog
+
+| Date | Version | Change |
+|------|---------|--------|
+| 2026-03-24 | 1.2.0 | Fixed gitnexus:start block duplication (was inlined in Reference Docs bullet). |
+| 2026-03-23 | 1.1.0 | Updated agent instructions (sections, references, Cursor layout). |
+| 2026-03-22 | 1.0.0 | Added structured agent header and changelog. |
+
+---
+
+<!-- gitnexus:start -->
+# GitNexus — Code Intelligence
+
+This project is indexed by GitNexus as **GitNexus**. Use the GitNexus MCP tools to understand code, assess impact, and navigate safely. For current symbol stats, run `npx gitnexus analyze` and inspect `.gitnexus/meta.json`.
+
+> If any GitNexus tool warns the index is stale, run `npx gitnexus analyze` in terminal first.
+
+## Always Do
+
+- **MUST run impact analysis before editing any symbol.** Before modifying a function, class, or method, run `gitnexus_impact({target: "symbolName", direction: "upstream"})` and report the blast radius (direct callers, affected processes, risk level) to the user.
+- **MUST run `gitnexus_detect_changes()` before committing** to verify your changes only affect expected symbols and execution flows.
+- **MUST warn the user** if impact analysis returns HIGH or CRITICAL risk before proceeding with edits.
+- When exploring unfamiliar code, use `gitnexus_query({query: "concept"})` to find execution flows instead of grepping. It returns process-grouped results ranked by relevance.
+- When you need full context on a specific symbol — callers, callees, which execution flows it participates in — use `gitnexus_context({name: "symbolName"})`.
+
+## When Debugging
+
+1. `gitnexus_query({query: "<error or symptom>"})` — find execution flows related to the issue
+2. `gitnexus_context({name: "<suspect function>"})` — see all callers, callees, and process participation
+3. `READ gitnexus://repo/GitNexus/process/{processName}` — trace the full execution flow step by step
+4. For regressions: `gitnexus_detect_changes({scope: "compare", base_ref: "main"})` — see what your branch changed
+
+## When Refactoring
+
+- **Renaming**: MUST use `gitnexus_rename({symbol_name: "old", new_name: "new", dry_run: true})` first. Review the preview — graph edits are safe, text_search edits need manual review. Then run with `dry_run: false`.
+- **Extracting/Splitting**: MUST run `gitnexus_context({name: "target"})` to see all incoming/outgoing refs, then `gitnexus_impact({target: "target", direction: "upstream"})` to find all external callers before moving code.
+- After any refactor: run `gitnexus_detect_changes({scope: "all"})` to verify only expected files changed.
+
+## Never Do
+
+- NEVER edit a function, class, or method without first running `gitnexus_impact` on it.
+- NEVER ignore HIGH or CRITICAL risk warnings from impact analysis.
+- NEVER rename symbols with find-and-replace — use `gitnexus_rename` which understands the call graph.
+- NEVER commit changes without running `gitnexus_detect_changes()` to check affected scope.
+
+## Tools Quick Reference
+
+| Tool | When to use | Command |
+|------|-------------|---------|
+| `query` | Find code by concept | `gitnexus_query({query: "auth validation"})` |
+| `context` | 360-degree view of one symbol | `gitnexus_context({name: "validateUser"})` |
+| `impact` | Blast radius before editing | `gitnexus_impact({target: "X", direction: "upstream"})` |
+| `detect_changes` | Pre-commit scope check | `gitnexus_detect_changes({scope: "staged"})` |
+| `rename` | Safe multi-file rename | `gitnexus_rename({symbol_name: "old", new_name: "new", dry_run: true})` |
+| `cypher` | Custom graph queries | `gitnexus_cypher({query: "MATCH ..."})` |
+
+## Impact Risk Levels
+
+| Depth | Meaning | Action |
+|-------|---------|--------|
+| d=1 | WILL BREAK — direct callers/importers | MUST update these |
+| d=2 | LIKELY AFFECTED — indirect deps | Should test |
+| d=3 | MAY NEED TESTING — transitive | Test if critical path |
+
+## Resources
+
+| Resource | Use for |
+|----------|---------|
+| `gitnexus://repo/GitNexus/context` | Codebase overview, check index freshness |
+| `gitnexus://repo/GitNexus/clusters` | All functional areas |
+| `gitnexus://repo/GitNexus/processes` | All execution flows |
+| `gitnexus://repo/GitNexus/process/{name}` | Step-by-step execution trace |
+
+## Self-Check Before Finishing
+
+Before completing any code modification task, verify:
+1. `gitnexus_impact` was run for all modified symbols
+2. No HIGH/CRITICAL risk warnings were ignored
+3. `gitnexus_detect_changes()` confirms changes match expected scope
+4. All d=1 (WILL BREAK) dependents were updated
+
+## Keeping the Index Fresh
+
+After committing code changes, the GitNexus index becomes stale. Re-run analyze to update it:
+
+```bash
+npx gitnexus analyze
+```
+
+If the index previously included embeddings, preserve them by adding `--embeddings`:
+
+```bash
+npx gitnexus analyze --embeddings
+```
+
+To check whether embeddings exist, inspect `.gitnexus/meta.json` — the `stats.embeddings` field shows the count (0 means no embeddings). **Running analyze without `--embeddings` will delete any previously generated embeddings.**
+
+> Claude Code users: A PostToolUse hook handles this automatically after `git commit` and `git merge`.
+
+## CLI
+
+| Task | Read this skill file |
+|------|---------------------|
+| Understand architecture / "How does X work?" | `.claude/skills/gitnexus/gitnexus-exploring/SKILL.md` |
+| Blast radius / "What breaks if I change X?" | `.claude/skills/gitnexus/gitnexus-impact-analysis/SKILL.md` |
+| Trace bugs / "Why is X failing?" | `.claude/skills/gitnexus/gitnexus-debugging/SKILL.md` |
+| Rename / extract / split / refactor | `.claude/skills/gitnexus/gitnexus-refactoring/SKILL.md` |
+| Tools, resources, schema reference | `.claude/skills/gitnexus/gitnexus-guide/SKILL.md` |
+| Index, status, clean, wiki CLI commands | `.claude/skills/gitnexus/gitnexus-cli/SKILL.md` |
+
+<!-- gitnexus:end -->
+
+## Cursor Cloud specific instructions
+
+### Repository structure
+
+This is a monorepo with two main products and supporting config packages:
+
+| Component | Path | Purpose |
+|-----------|------|---------|
+| **GitNexus CLI/Core** | `gitnexus/` | Main product — TypeScript CLI, indexing pipeline, MCP server. Published to npm. |
+| **GitNexus Web UI** | `gitnexus-web/` | React/Vite browser app — graph explorer + AI chat. Runs entirely in WASM. |
+| Claude Plugin | `gitnexus-claude-plugin/` | Static config for Claude marketplace (no build). |
+| Cursor Integration | `gitnexus-cursor-integration/` | Static config for Cursor editor (no build). |
+| SWE-bench Eval | `eval/` | Python evaluation harness (optional; needs Docker + LLM API keys). |
+
+### Running services
+
+- **CLI/Core**: `cd gitnexus && npm run dev` (tsx watch mode) or `npm run build && node dist/cli/index.js <command>`
+- **Web UI**: `cd gitnexus-web && npm run dev` (Vite on port 5173)
+- **Backend mode**: `cd <indexed-repo> && node /workspace/gitnexus/dist/cli/index.js serve` (HTTP API on port 3741 by default)
+
+### Testing
+
+**CLI / Core (`gitnexus/`)**
+- **Unit tests**: `cd gitnexus && npm test` (vitest, ~2000 tests)
+- **Integration tests**: `cd gitnexus && npm run test:integration` (vitest, ~1850 tests). Two LadybugDB file-locking tests (`lbug-core-adapter`, `search-core`) may fail in containerized environments due to `/tmp` locking limitations — this is a known environment issue, not a code bug.
+- **TypeScript check**: `cd gitnexus && npx tsc --noEmit`
+
+**Web UI (`gitnexus-web/`)**
+- **Unit tests**: `cd gitnexus-web && npm test` (vitest, ~200 tests)
+- **E2E tests**: `cd gitnexus-web && E2E=1 npx playwright test` (Playwright, 5 tests — requires `gitnexus serve` + `npm run dev` running)
+- **TypeScript check**: `cd gitnexus-web && npx tsc -b --noEmit`
+
+No separate lint command is configured; TypeScript strict checking serves as the primary static analysis.
+
+### Gotchas
+
+- `npm install` in `gitnexus/` triggers `prepare` (builds via `tsc`) and `postinstall` (patches tree-sitter-swift). Native tree-sitter bindings require `python3`, `make`, and `g++` to be present.
+- `tree-sitter-kotlin` and `tree-sitter-swift` are optional dependencies — install warnings for these are expected and non-blocking.
+- The Web UI uses `vite-plugin-wasm` and requires `Cross-Origin-Opener-Policy`/`Cross-Origin-Embedder-Policy` headers for `SharedArrayBuffer` (handled automatically by Vite dev server).
+- There is no ESLint/Prettier configuration in this repo.

ARCHITECTURE.md

@@ -0,0 +1,66 @@
+# Architecture — GitNexus
+
+This repository is a **monorepo** with two main products: the **CLI / MCP package** (`gitnexus/`) and the **browser UI** (`gitnexus-web/`). Supporting folders ship editor integrations and plugins without changing the core graph engine.
+
+## Repository layout
+
+| Path | Role |
+|------|------|
+| `gitnexus/` | Published npm package `gitnexus`: CLI, MCP server (stdio), local HTTP API for bridge mode, ingestion pipeline, LadybugDB graph, embeddings (optional). |
+| `gitnexus-web/` | Vite + React UI: in-browser indexing (WASM), graph visualization, optional connection to `gitnexus serve`. |
+| `.claude/`, `gitnexus-claude-plugin/`, `gitnexus-cursor-integration/` | Packaged **skills** and plugin metadata so agents discover the same workflows as documented in `AGENTS.md`. |
+| `eval/` | Evaluation harnesses and docs for benchmarking tool usage. |
+| `.github/` | CI workflows (quality, unit, integration, E2E) and composite actions. |
+
+## End-to-end flow: index → graph → tools
+
+1. **Ingestion** (`gitnexus analyze`)  
+   - Entry: `gitnexus/src/cli/analyze.ts` → `runPipelineFromRepo` in `gitnexus/src/core/ingestion/pipeline.ts`.  
+   - Walks the git working tree, parses supported languages via **Tree-sitter**, resolves imports/calls/inheritance, detects **communities** and **processes** (execution flows), and builds an in-memory **knowledge graph** (`gitnexus/src/core/graph/`).  
+   - Output is loaded into **LadybugDB** under **`.gitnexus/`** at the repo root (`lbug/`, `meta.json`, etc.). Optional **FTS** indexes and **embeddings** attach to the same store.  
+   - The repo is registered in **`~/.gitnexus/registry.json`** so MCP can find it from any working directory.
+
+2. **Persistence & metadata**  
+   - `gitnexus/src/storage/repo-manager.ts` — paths, registry, cleanup of legacy Kuzu artifacts.  
+   - `gitnexus/src/core/lbug/lbug-adapter.ts` — graph load, queries, embedding restore batches.
+
+3. **Query & agents**  
+   - **MCP (stdio):** `gitnexus/src/cli/mcp.ts` → `startMCPServer` → `LocalBackend` (`gitnexus/src/mcp/local/local-backend.ts`) opens registered repos and serves **tools** from `gitnexus/src/mcp/tools.ts` and **resources** from `gitnexus/src/mcp/resources.ts`.  
+   - **Bridge HTTP:** `gitnexus/src/cli/serve.ts` → Express app in `gitnexus/src/server/api.ts` (CORS-limited) exposes REST + MCP-over-HTTP for the web UI.  
+   - **CLI tools (no MCP):** `gitnexus query`, `context`, `impact`, `cypher` in `gitnexus/src/cli/tool.ts` call the same backend for scripts and CI.
+
+4. **Staleness**  
+   - `gitnexus/src/mcp/staleness.ts` compares indexed `lastCommit` to `HEAD` and surfaces hints when the graph is behind git.
+
+## MCP tools (summary)
+
+| Tool | Purpose |
+|------|---------|
+| `list_repos` | Discover indexed repositories when more than one is registered. |
+| `query` | Natural-language / keyword search over the graph (hybrid BM25 + optional vectors). |
+| `cypher` | Ad hoc **Cypher** against the schema (see resource `gitnexus://repo/{name}/schema`). |
+| `context` | Callers, callees, processes for one symbol (with disambiguation). |
+| `impact` | Blast radius (upstream/downstream) with depth and risk summary. |
+| `detect_changes` | Map git diffs to affected symbols and processes. |
+| `rename` | Graph-assisted rename with `dry_run` preview (`graph` vs `text_search` confidence). |
+
+## Where to change what
+
+| If you are changing… | Start in… |
+|----------------------|-----------|
+| CLI commands / flags | `gitnexus/src/cli/` (`index.ts`, per-command modules). |
+| Parsing or graph construction | `gitnexus/src/core/ingestion/` (pipeline, processors, resolvers, type-extractors). |
+| Graph schema / DB access | `gitnexus/src/core/lbug/` (`schema.ts`, `lbug-adapter.ts`), `gitnexus/src/mcp/core/lbug-adapter.ts` if MCP-specific. |
+| MCP protocol, tools, resources | `gitnexus/src/mcp/server.ts`, `tools.ts`, `resources.ts`. |
+| Search ranking | `gitnexus/src/core/search/` (BM25, hybrid fusion). |
+| Embeddings | `gitnexus/src/core/embeddings/`, phases in `analyze.ts`. |
+| Wiki generation | `gitnexus/src/core/wiki/`. |
+| Web UI behavior | `gitnexus-web/src/` (components, workers, graph client). |
+| CI | `.github/workflows/*.yml`, `.github/actions/setup-gitnexus/`. |
+
+## Related docs
+
+- [RUNBOOK.md](RUNBOOK.md) — operational commands and recovery.  
+- [GUARDRAILS.md](GUARDRAILS.md) — safety boundaries for humans and agents.  
+- [TESTING.md](TESTING.md) — how to run tests.  
+- `AGENTS.md` / `CLAUDE.md` — agent workflows and tool usage expectations for **this** repo when indexed by GitNexus.

CHANGELOG.md

@@ -0,0 +1,109 @@
+# Changelog
+
+All notable changes to GitNexus will be documented in this file.
+
+## [Unreleased]
+
+### Changed
+- Migrated from KuzuDB to LadybugDB v0.15 (`@ladybugdb/core`, `@ladybugdb/wasm-core`)
+- Renamed all internal paths from `kuzu` to `lbug` (storage: `.gitnexus/kuzu` → `.gitnexus/lbug`)
+- Added automatic cleanup of stale KuzuDB index files
+- LadybugDB v0.15 requires explicit VECTOR extension loading for semantic search
+
+## [1.4.0] - 2026-03-13
+
+### Added
+
+- **Language-aware symbol resolution engine** with 3-tier resolver: exact FQN → scope-walk → guarded fuzzy fallback that refuses ambiguous matches (#238) — @magyargergo
+- **Method Resolution Order (MRO)** with 5 language-specific strategies: C++ leftmost-base, C#/Java class-over-interface, Python C3 linearization, Rust qualified syntax, default BFS (#238) — @magyargergo
+- **Constructor & struct literal resolution** across all languages — `new Foo()`, `User{...}`, C# primary constructors, target-typed new (#238) — @magyargergo
+- **Receiver-constrained resolution** using per-file TypeEnv — disambiguates `user.save()` vs `repo.save()` via `ownerId` matching (#238) — @magyargergo
+- **Heritage & ownership edges** — HAS_METHOD, OVERRIDES, Go struct embedding, Swift extension heritage, method signatures (`parameterCount`, `returnType`) (#238) — @magyargergo
+- **Language-specific resolver directory** (`resolvers/`) — extracted JVM, Go, C#, PHP, Rust resolvers from monolithic import-processor (#238) — @magyargergo
+- **Type extractor directory** (`type-extractors/`) — per-language type binding extraction with `Record<SupportedLanguages, Handler>` + `satisfies` dispatch (#238) — @magyargergo
+- **Export detection dispatch table** — compile-time exhaustive `Record` + `satisfies` pattern replacing switch/if chains (#238) — @magyargergo
+- **Language config module** (`language-config.ts`) — centralized tsconfig, go.mod, composer.json, .csproj, Swift package config loaders (#238) — @magyargergo
+- **Optional skill generation** via `npx gitnexus analyze --skills` — generates AI agent skills from KuzuDB knowledge graph (#171) — @zander-raycraft
+- **First-class C# support** — sibling-based modifier scanning, record/delegate/property/field/event declaration types (#163, #170, #178 via #237) — @Alice523, @benny-yamagata, @jnMetaCode
+- **C/C++ support fixes** — `.h` → C++ mapping, static-linkage export detection, qualified/parenthesized declarators, 48 entry point patterns (#163, #227 via #237) — @Alice523, @bitgineer
+- **Rust support fixes** — sibling-based `visibility_modifier` scanning for `pub` detection (#227 via #237) — @bitgineer
+- **Adaptive tree-sitter buffer sizing** — `Math.min(Math.max(contentLength * 2, 512KB), 32MB)` (#216 via #237) — @JasonOA888
+- **Call expression matching** in tree-sitter queries (#234 via #237) — @ex-nihilo-jg
+- **DeepSeek model configurations** (#217) — @JasonOA888
+- 282+ new unit tests, 178 integration resolver tests across 9 languages, 53 test files, 1146 total tests passing
+
+### Fixed
+
+- Skip unavailable native Swift parsers in sequential ingestion (#188) — @Gujiassh
+- Heritage heuristic language-gated — no longer applies class/interface rules to wrong languages (#238) — @magyargergo
+- C# `base_list` distinguishes EXTENDS vs IMPLEMENTS via symbol table + `I[A-Z]` heuristic (#238) — @magyargergo
+- Go `qualified_type` (`models.User`) correctly unwrapped in TypeEnv (#238) — @magyargergo
+- Global tier no longer blocks resolution when kind/arity filtering can narrow to 1 candidate (#238) — @magyargergo
+
+### Changed
+
+- `import-processor.ts` reduced from 1412 → 711 lines (50% reduction) via resolver and config extraction (#238) — @magyargergo
+- `type-env.ts` reduced from 635 → ~125 lines via type-extractor extraction (#238) — @magyargergo
+- CI/CD workflows hardened with security fixes and fork PR support (#222, #225) — @magyargergo
+
+## [1.3.11] - 2026-03-08
+
+### Security
+
+- Fix FTS Cypher injection by escaping backslashes in search queries (#209) — @magyargergo
+
+### Added
+
+- Auto-reindex hook that runs `gitnexus analyze` after commits and merges, with automatic embeddings preservation (#205) — @L1nusB
+- 968 integration tests (up from ~840) covering unhappy paths across search, enrichment, CLI, pipeline, worker pool, and KuzuDB (#209) — @magyargergo
+- Coverage auto-ratcheting so thresholds bump automatically on CI (#209) — @magyargergo
+- Rich CI PR report with coverage bars, test counts, and threshold tracking (#209) — @magyargergo
+- Modular CI workflow architecture with separate unit-test, integration-test, and orchestrator jobs (#209) — @magyargergo
+
+### Fixed
+
+- KuzuDB native addon crashes on Linux/macOS by running integration tests in isolated vitest processes with `--pool=forks` (#209) — @magyargergo
+- Worker pool `MODULE_NOT_FOUND` crash when script path is invalid (#209) — @magyargergo
+
+### Changed
+
+- Added macOS to the cross-platform CI test matrix (#208) — @magyargergo
+
+## [1.3.10] - 2026-03-07
+
+### Security
+
+- **MCP transport buffer cap**: Added 10 MB `MAX_BUFFER_SIZE` limit to prevent out-of-memory attacks via oversized `Content-Length` headers or unbounded newline-delimited input
+- **Content-Length validation**: Reject `Content-Length` values exceeding the buffer cap before allocating memory
+- **Stack overflow prevention**: Replaced recursive `readNewlineMessage` with iterative loop to prevent stack overflow from consecutive empty lines
+- **Ambiguous prefix hardening**: Tightened `looksLikeContentLength` to require 14+ bytes before matching, preventing false framing detection on short input
+- **Closed transport guard**: `send()` now rejects with a clear error when called after `close()`, with proper write-error propagation
+
+### Added
+
+- **Dual-framing MCP transport** (`CompatibleStdioServerTransport`): Auto-detects Content-Length (Codex/OpenCode) and newline-delimited JSON (Cursor/Claude Code) framing on the first message, responds in the same format (#207)
+- **Lazy CLI module loading**: All CLI subcommands now use `createLazyAction()` to defer heavy imports (tree-sitter, ONNX, KuzuDB) until invocation, significantly improving `gitnexus mcp` startup time (#207)
+- **Type-safe lazy actions**: `createLazyAction` uses constrained generics to validate export names against module types at compile time
+- **Regression test suite**: 13 unit tests covering transport framing, security hardening, buffer limits, and lazy action loading
+
+### Fixed
+
+- **CALLS edge sourceId alignment**: `findEnclosingFunctionId` now generates IDs with `:startLine` suffix matching node creation format, fixing process detector finding 0 entry points (#194)
+- **LRU cache zero maxSize crash**: Guard `createASTCache` against `maxSize=0` when repos have no parseable files (#144)
+
+### Changed
+
+- Transport constructor accepts `NodeJS.ReadableStream` / `NodeJS.WritableStream` (widened from concrete `ReadStream`/`WriteStream`)
+- `processReadBuffer` simplified to break on first error instead of stale-buffer retry loop
+
+## [1.3.9] - 2026-03-06
+
+### Fixed
+
+- Aligned CALLS edge sourceId with node ID format in parse worker (#194)
+
+## [1.3.8] - 2026-03-05
+
+### Fixed
+
+- Force-exit after analyze to prevent KuzuDB native cleanup hang (#192)

CLAUDE.md

@@ -0,0 +1,52 @@
+<!-- version: 1.2.0 -->
+<!--
+  Metadata: version, last reviewed, scope, model policy, reference docs, changelog.
+  Last updated: 2026-03-22
+-->
+
+Last reviewed: 2026-03-24
+
+**Project:** GitNexus · **Environment:** dev · **Maintainer:** repository maintainers (see GitHub)
+
+Follow **AGENTS.md** for the canonical rules; this file adds Claude Code–specific deltas. Cursor-specific notes live only in `AGENTS.md`.
+
+## Scope
+
+See the **Scope** table in [AGENTS.md](AGENTS.md) for read/write/execute/off-limits boundaries. Cursor-specific workflow notes also live only in AGENTS.md.
+
+## Model Configuration
+
+- **Primary:** Pin per **Claude Code** / Anthropic org policy (explicit model id). Do not rely on an unversioned `latest` alias for governed workflows.
+- **Fallback:** As configured in Claude Code (organization default or user override).
+- **Notes:** The GitNexus CLI analyzer does not call an LLM.
+
+## Execution Sequence (complex tasks)
+
+Same discipline as [AGENTS.md](AGENTS.md): before large multi-step work, state which **AGENTS.md** / **GUARDRAILS.md** rules apply, current **Scope**, and planned validation commands (`npm test`, `tsc`, etc.). When pausing, summarize progress in the chat or a **local** scratch file (do not add `HANDOFF.md` to the repo), then `/clear` and resume with that summary.
+
+## Claude Code hooks
+
+Prefer **PreToolUse** hooks for hard gates (e.g. tests before `git_commit`). Adapt hook commands to `gitnexus/` npm scripts.
+
+## Context budget
+
+If always-on instructions grow, load deep conventions via conditional reads (e.g. *“When writing new code, read STANDARDS.md”*) instead of pasting long blocks here. In Cursor, prefer `.cursor/index.mdc` plus optional `.cursor/rules/*.mdc` globs (see [AGENTS.md](AGENTS.md) § Context budget).
+
+## Reference Documentation
+
+- **This repository:** [AGENTS.md](AGENTS.md) (Cursor + monorepo notes), [ARCHITECTURE.md](ARCHITECTURE.md), [CONTRIBUTING.md](CONTRIBUTING.md), [GUARDRAILS.md](GUARDRAILS.md).
+- **GitNexus:** `.claude/skills/gitnexus/`; MCP and indexed-repo rules live only in [AGENTS.md](AGENTS.md) (`gitnexus:start` … `gitnexus:end`). See **GitNexus rules** below.
+
+## Changelog
+
+| Date | Version | Change |
+|------|---------|--------|
+| 2026-03-24 | 1.2.0 | Removed duplicated gitnexus:start block and scope table; replaced with pointers to AGENTS.md. |
+| 2026-03-23 | 1.1.0 | Updated agent instructions to match AGENTS.md. |
+| 2026-03-22 | 1.0.0 | Added structured header and changelog. |
+
+---
+
+## GitNexus rules
+
+GitNexus MCP rules are in the `<!-- gitnexus:start -->` … `<!-- gitnexus:end -->` block in **[AGENTS.md](AGENTS.md)** — load that section when working with MCP tools or the graph index.

COMPREHENSIVE_END_TO_END_VERIFICATION.md

@@ -1,207 +0,0 @@
-# 🔍 Comprehensive End-to-End Verification Report
-
-## Executive Summary
-✅ **ALL SYSTEMS VERIFIED** - The parallel processing implementation is now fully optimized with proper LRU caching, memory management, and cleanup mechanisms.
-
-## 🚨 Critical Issue Found & Fixed
-
-### **THE PROBLEM: LRU Cache Bypass in Parallel Mode**
-The parallel processing was **completely bypassing the LRU cache** for the main parsing logic, causing:
-- ❌ Every file parsed from scratch (no cache benefits)
-- ❌ Memory bloat and performance degradation
-- ❌ Sluggish behavior due to redundant processing
-
-### **THE FIX: Cache-First Processing**
-✅ Implemented **LRU cache-first processing** in `ParallelParsingProcessor.processFilesInParallel()`:
-```typescript
-// NEW: Check LRU cache before sending to workers
-for (const filePath of filePaths) {
-  const cacheKey = this.lruCache.generateFileCacheKey(filePath, contentHash);
-  const cachedResult = this.lruCache.getParsedFile(cacheKey);
-  if (cachedResult) {
-    // Use cached result ✅
-    cachedResults.push({...});
-  } else {
-    // Send to workers ✅
-    uncachedFiles.push(filePath);
-  }
-}
-```
-
----
-
-## 📋 Detailed Verification Results
-
-### 1. ✅ LRU Cache Integration
-**Status: VERIFIED & OPTIMIZED**
-
-#### Single-threaded Mode (`ParsingProcessor`):
-- ✅ File caching: `lruCache.getParsedFile()` / `setParsedFile()`
-- ✅ Query caching: `lruCache.getQueryResult()` / `setQueryResult()`
-- ✅ Parser caching: `lruCache.getParser()` / `setParser()`
-- ✅ Cache key generation: `generateFileCacheKey()` / `generateQueryCacheKey()`
-
-#### Parallel Mode (`ParallelParsingProcessor`):
-- ✅ **FIXED**: Now checks cache before worker processing
-- ✅ File caching: Same as single-threaded
-- ✅ Worker result caching: Results cached after processing
-- ✅ Cache statistics: `lruCache.getStats()` / `getCacheHitRate()`
-
-#### Expected Console Output:
-```
-ParallelParsingProcessor: Cache hits: X, Files to process: Y
-ParallelParsingProcessor: Cache hit for /path/to/file.ts
-ParallelParsingProcessor: Total results: Z (X cached, Y processed)
-```
-
-### 2. ✅ Memory Management & Cleanup
-**Status: VERIFIED & ROBUST**
-
-#### Memory Monitoring:
-- ✅ **30-second interval monitoring** in parallel mode
-- ✅ **Memory threshold triggers** (500MB → cleanup, 800MB → aggressive cleanup)
-- ✅ **AST map size limits** (1000 entries max with cleanup)
-- ✅ **LRU cache statistics** logging
-
-#### Cleanup Mechanisms:
-- ✅ **Memory Manager**: `memoryManager.clearCache()`
-- ✅ **LRU Cache**: `lruCache.clearAll()` / `clearFileCache()` / `clearQueryCache()`
-- ✅ **AST Map**: `cleanupASTMap()` with LRU-based eviction
-- ✅ **Duplicate Detector**: `duplicateDetector.clear()`
-
-#### Expected Console Output:
-```
-ParallelParsingProcessor Memory Stats:
-  - Memory Manager: XXXmb used, YYY files cached
-  - LRU File Cache: A/200 entries, B.XMB
-  - LRU Query Cache: C/100 entries, D.XMB
-  - Cache Hit Rates: File XX.X%, Query YY.Y%
-  - AST Map Size: ZZZ entries
-```
-
-### 3. ✅ Worker Pool Lifecycle & Cleanup
-**Status: VERIFIED & SECURE**
-
-#### Worker Pool Management:
-- ✅ **Proper initialization** with CPU-optimized settings
-- ✅ **Event listener cleanup** on task completion/error
-- ✅ **Worker termination** with Promise.all for parallel shutdown
-- ✅ **Singleton cleanup** via `FileProcessingPool.shutdownInstance()`
-
-#### Global Cleanup Handlers:
-- ✅ **Page unload**: `beforeunload` event → `cleanupAllPools()`
-- ✅ **Page hidden**: `visibilitychange` event → cleanup when hidden
-- ✅ **Memory pressure**: Automatic cleanup at 80% memory usage
-- ✅ **Manual cleanup**: `WebWorkerPoolUtils.cleanupAllPools()`
-
-#### Shutdown Sequence:
-```typescript
-// ParallelParsingProcessor.shutdown()
-1. Clear memory monitor interval
-2. Clear AST map & processed files
-3. Shutdown worker pool (terminate all workers)
-4. Clear LRU caches
-5. Clear language parsers
-```
-
-### 4. ✅ Cache Consistency Between Modes
-**Status: VERIFIED & IDENTICAL**
-
-#### Consistent Cache Key Generation:
-- ✅ **Same hash algorithm**: Both use identical `generateContentHash()`
-- ✅ **Same cache keys**: `lruCache.generateFileCacheKey(filePath, contentHash)`
-- ✅ **Same cache structure**: Identical cache data format
-- ✅ **Same LRU service**: Both use `LRUCacheService.getInstance()`
-
-#### Cache Data Format (Both Modes):
-```typescript
-{
-  ast: Parser.Tree,
-  definitions: ParsedDefinition[],
-  language: string,
-  lastModified: number,
-  fileSize: number
-}
-```
-
-### 5. ✅ Performance Monitoring & Logging
-**Status: VERIFIED & COMPREHENSIVE**
-
-#### Parallel Mode Logging:
-- ✅ **Cache hit rates**: Shows cached vs processed files
-- ✅ **Worker pool stats**: Active workers, completed tasks, errors
-- ✅ **Memory statistics**: Real-time memory usage monitoring
-- ✅ **Processing times**: Per-file and total processing duration
-- ✅ **Progress tracking**: Real-time progress updates
-
-#### Single-threaded Mode Logging:
-- ✅ **Memory statistics**: Memory manager stats
-- ✅ **Cache statistics**: LRU cache hit rates
-- ✅ **Processing stats**: File counts and definitions extracted
-
----
-
-## 🎯 Performance Improvements Expected
-
-### First Run (Cold Cache):
-- **Single-threaded**: Baseline performance
-- **Parallel**: Faster due to worker parallelization + caching setup
-
-### Subsequent Runs (Warm Cache):
-- **Both modes**: **Dramatically faster** due to cache hits
-- **Cache hit ratio**: Should be 80-95% for unchanged files
-- **Memory usage**: Stable and controlled via cleanup mechanisms
-
-### Memory Behavior:
-- **Before fix**: Unlimited growth → sluggishness
-- **After fix**: Controlled growth with automatic cleanup
-
----
-
-## 🔧 Verification Commands
-
-To verify the fixes are working, look for these console outputs:
-
-### Cache Verification:
-```bash
-# Should see cache hits on subsequent runs
-ParallelParsingProcessor: Cache hits: 150, Files to process: 50
-```
-
-### Memory Monitoring:
-```bash
-# Should see regular memory stats
-ParallelParsingProcessor Memory Stats:
-  - Memory Manager: 245MB used, 1250 files cached
-  - Cache Hit Rates: File 87.5%, Query 92.3%
-```
-
-### Worker Pool Stats:
-```bash
-# Should see worker efficiency
-ParallelParsingProcessor: Worker pool stats: {
-  activeWorkers: 4,
-  completedTasks: 200,
-  failedTasks: 0
-}
-```
-
----
-
-## ✅ Conclusion
-
-**ALL CRITICAL ISSUES RESOLVED:**
-
-1. **🚨 LRU Cache Bypass** → ✅ **Cache-first processing implemented**
-2. **🚨 Memory Leaks** → ✅ **Comprehensive cleanup mechanisms**
-3. **🚨 Worker Pool Leaks** → ✅ **Proper lifecycle management**
-4. **🚨 Inconsistent Caching** → ✅ **Identical cache behavior**
-5. **🚨 Poor Monitoring** → ✅ **Comprehensive performance logging**
-
-**The sluggishness should be significantly reduced** because:
-- ✅ **First run**: Files get cached after processing
-- ✅ **Subsequent runs**: Most files served from cache (near-instant)
-- ✅ **Memory management**: Automatic cleanup prevents bloat
-- ✅ **Worker efficiency**: Only uncached files sent to workers
-
-**🚀 Ready for production use with optimal performance!**

CONFIG.md

@@ -1,75 +0,0 @@
-# GitNexus Configuration
-
-## Feature Flags
-
-GitNexus uses feature flags to control the visibility of experimental and advanced features. By default, the UI is kept clean and simple for end users.
-
-### Available Feature Flags
-
-- `showEngineSelector` - Show engine selection dropdown
-- `showEnginePerformanceInfo` - Display performance comparison data
-- `showEngineCapabilities` - Show engine capabilities grid
-- `enableNextGenEngine` - Enable next-generation processing engine
-- `enableEngineComparison` - Run performance comparisons between engines
-- `enableDebugMode` - Show debugging information
-- `showProcessingDetails` - Display detailed processing information
-
-### Enabling Features for Development
-
-To enable advanced features during development, modify `/src/config/feature-flags.ts`:
-
-```typescript
-export const getFeatureFlags = (): FeatureFlags => {
-  if (import.meta.env.DEV) {
-    return {
-      ...defaultFeatureFlags,
-      // Enable engine features for development
-      showEngineSelector: true,
-      showEnginePerformanceInfo: true,
-      showEngineCapabilities: true,
-      enableDebugMode: true,
-      showProcessingDetails: true,
-    };
-  }
-  
-  return defaultFeatureFlags;
-};
-```
-
-### Production Configuration
-
-For production builds, keep feature flags disabled to maintain a clean user interface:
-
-```typescript
-export const defaultFeatureFlags: FeatureFlags = {
-  showEngineSelector: false,        // Hidden from end users
-  showEnginePerformanceInfo: false, // Hidden from end users
-  showEngineCapabilities: false,    // Hidden from end users
-  enableNextGenEngine: true,        // Enabled behind the scenes
-  enableEngineComparison: false,    // Disabled to save resources
-  enableDebugMode: false,
-  showProcessingDetails: false,
-};
-```
-
-### Environment Variables
-
-You can also control features via environment variables:
-
-```bash
-# Enable engine selector in development
-VITE_SHOW_ENGINE_SELECTOR=true npm run dev
-
-# Enable all debug features
-VITE_DEBUG_MODE=true npm run dev
-```
-
-## Layout Configuration
-
-The new layout prioritizes the graph visualization:
-
-- **Left Sidebar (400px)**: Processing status, chat interface, and actions
-- **Right Side (flex)**: Full graph visualization
-- **Mobile**: Responsive single-column layout
-
-This provides maximum space for the graph while keeping the chat interface easily accessible.

CONTRIBUTING.md

@@ -0,0 +1,50 @@
+# Contributing to GitNexus
+
+How to propose changes, run checks locally, and open pull requests.
+
+## License
+
+This project uses the [PolyForm Noncommercial License 1.0.0](https://polyformproject.org/licenses/noncommercial/1.0.0/). By contributing, you agree your contributions are licensed under the same terms unless stated otherwise.
+
+## Where to discuss
+
+- **Issues & feature ideas:** use [GitHub Issues](https://github.com/abhigyanpatwari/GitNexus/issues) for the upstream repo, or your fork’s tracker if you work from a fork.
+- **Community:** see the Discord link in the root [README.md](README.md).
+
+## Development setup
+
+1. Clone the repository.
+2. **CLI / MCP package:** `cd gitnexus && npm install && npm run build`
+3. **Web UI (if needed):** `cd gitnexus-web && npm install`
+4. Run tests as described in [TESTING.md](TESTING.md).
+
+## Branch and pull requests
+
+- Use short-lived branches off the default branch of the repo you are targeting.
+- Prefer **conventional commits** (short prefix + description), for example:
+
+  ```text
+  feat: add graph export option
+  fix: correct MCP tool schema for query
+  test: cover cluster merge edge case
+  docs: clarify analyze flags
+  ```
+
+- **PR title:** `[area] Short description` (e.g. `[cli] Fix index refresh race`).
+- **PR description:** what changed, why, how to verify (commands), and any risk or rollback notes.
+
+## Before you open a PR
+
+- [ ] Tests pass for the packages you touched (`gitnexus` and/or `gitnexus-web`).
+- [ ] Typecheck passes: `npx tsc --noEmit` in `gitnexus/` and `npx tsc -b --noEmit` in `gitnexus-web/`.
+- [ ] No secrets, tokens, or machine-specific paths committed.
+- [ ] Documentation updated if behavior or public CLI/MCP contract changes.
+- [ ] Pre-commit hook runs clean (`.husky/pre-commit` — typecheck + unit tests for staged packages).
+
+## Code review
+
+Maintainers may request changes for correctness, tests, performance, or consistency with existing patterns. Keeping diffs focused makes review faster.
+
+## AI-assisted contributions
+
+If you use coding agents, follow project context files (e.g. `AGENTS.md`, `CLAUDE.md`) and avoid drive-by refactors unrelated to the issue. Prefer incremental, test-backed changes.

FINAL_VERIFICATION_REPORT.md

@@ -1,148 +0,0 @@
-# 🔍 FINAL COMPREHENSIVE VERIFICATION REPORT
-
-## ✅ **VERIFICATION COMPLETE - ALL CRITICAL ISSUES RESOLVED**
-
-After an extremely thorough examination, both single-threaded and parallel processing modes are now **fully synchronized** and **memory-optimized**.
-
----
-
-## 🚨 **CRITICAL ISSUES FOUND AND FIXED**
-
-### **Issue #1: Wrong Pipeline Selection in Worker** 🔥 **CRITICAL**
-**Problem**: Worker was always using `GraphPipeline` instead of checking the feature flag.
-**Impact**: "Parallel processing" was actually running single-threaded code.
-**Fix**: ✅ Worker now correctly selects `ParallelGraphPipeline` when parallel mode is enabled.
-
-### **Issue #2: LRU Cache Completely Disabled in Single-Threaded Mode** 🔥 **CRITICAL**
-**Problem**: Single-threaded processor had all LRU caching commented out as "TEMPORARILY DISABLED FOR DEBUGGING".
-**Impact**: Single-threaded mode had no caching, parallel mode did - major performance inconsistency.
-**Fix**: ✅ Re-enabled all LRU cache operations in single-threaded processor.
-
-### **Issue #3: Incorrect Duplicate Detection** 🔥 **CRITICAL**
-**Problem**: 
-- Single-threaded: `checkAndMark()` (correct)
-- Parallel: `isDuplicate()` (wrong - doesn't mark as processed)
-**Fix**: ✅ Changed parallel processor to use `checkAndMark()`.
-
-### **Issue #4: Property Format Inconsistency** 🔶 **MEDIUM**
-**Problem**: Parallel processor stored arrays as comma-separated strings.
-**Fix**: ✅ Made both processors store identical array formats.
-
-### **Issue #5: Memory Leak Prevention** 🔶 **MEDIUM**
-**Problem**: Worker pools, event listeners, and AST maps could accumulate without cleanup.
-**Fix**: ✅ Comprehensive memory management implemented.
-
----
-
-## 📊 **FINAL VERIFICATION CHECKLIST**
-
-### **✅ Pipeline Architecture**
-- [x] Worker correctly selects `GraphPipeline` vs `ParallelGraphPipeline` based on feature flag
-- [x] Both pipelines use identical 4-pass structure (Structure → Parsing → Import → Call)
-- [x] Both pipelines use same processors (except parsing processor)
-- [x] Progress callbacks properly integrated
-
-### **✅ LRU Cache Consistency**
-- [x] Both modes use `LRUCacheService.getInstance()`
-- [x] File caching enabled in both modes (200 max, 1 hour TTL)
-- [x] Query caching enabled in both modes (1000 max, 15 min TTL)
-- [x] Parser caching enabled in both modes (10 max, 24 hours TTL)
-- [x] Cache hit rate tracking in both modes
-
-### **✅ Data Processing Consistency**
-- [x] Identical duplicate detection logic (`checkAndMark()`)
-- [x] Identical node ID generation
-- [x] Identical node label mapping
-- [x] Identical property formats (arrays as arrays, not strings)
-- [x] Identical relationship creation
-
-### **✅ Memory Management**
-- [x] Worker pools properly terminated with event listener cleanup
-- [x] AST maps size-limited (1000 entries) with automatic cleanup
-- [x] LRU caches automatically manage memory (100MB total limit)
-- [x] Singleton instances properly cleaned up
-- [x] Memory monitoring with automatic triggers (500MB threshold)
-- [x] Global cleanup handlers for page unload/visibility changes
-
-### **✅ Error Handling**
-- [x] Graceful worker termination on errors
-- [x] Proper resource cleanup in finally blocks
-- [x] Error recovery without resource leaks
-
----
-
-## 🎯 **EXPECTED BEHAVIOR AFTER FIXES**
-
-### **Single-Threaded Mode (`isParallelParsingEnabled() = false`)**:
-- Uses `GraphPipeline` with `ParsingProcessor`
-- Sequential file processing on main thread
-- Full LRU caching enabled
-- Direct Tree-sitter AST parsing
-
-### **Parallel Mode (`isParallelParsingEnabled() = true`)**:
-- Uses `ParallelGraphPipeline` with `ParallelParsingProcessor`  
-- Worker pool parallel processing (2-8 workers based on CPU cores)
-- Full LRU caching enabled
-- Worker-based parsing + main thread AST recreation
-
-### **Identical Output Guaranteed**:
-Both modes will now produce:
-- ✅ **Same node counts** by type (Function, Class, Variable, etc.)
-- ✅ **Same relationship counts** by type (CONTAINS, DEFINES, IMPORTS, CALLS)
-- ✅ **Same import relationships** - no more "missing import relationships"
-- ✅ **Same function call relationships** - no more "missing call relationships"
-- ✅ **Same graph connectivity** - proper relationships between files and definitions
-
----
-
-## 🚀 **PERFORMANCE IMPROVEMENTS**
-
-### **Memory Usage**:
-- **LRU Cache**: Automatic memory management with 100MB limit
-- **AST Maps**: Size-limited to 1000 entries with cleanup
-- **Worker Pools**: Proper termination prevents accumulation
-- **Global Monitoring**: Memory usage tracked every 30 seconds
-
-### **Processing Speed**:
-- **Single-threaded**: Now benefits from LRU caching (was disabled)
-- **Parallel**: 2-8x speedup on large codebases + LRU caching benefits
-- **Cache Hit Rates**: Both modes show file/query cache performance
-
-### **Resource Management**:
-- **No Memory Leaks**: All resources properly cleaned up
-- **Automatic Cleanup**: Triggers at 80% memory usage
-- **Graceful Shutdown**: Proper cleanup on page close/tab switch
-
----
-
-## 🧪 **TESTING VERIFICATION**
-
-To verify the fixes work:
-
-1. **Process the same codebase** with both modes
-2. **Compare console output** - should show identical relationship counts
-3. **Check for these success indicators**:
-   ```
-   ✅ Relationships by type: {CONTAINS: X, DEFINES: Y, IMPORTS: Z, CALLS: W}
-   ✅ No warnings about "missing import relationships"
-   ✅ No warnings about "missing function call relationships"
-   ✅ Memory usage stays stable during processing
-   ✅ LRU cache hit rates displayed in both modes
-   ```
-
-4. **Performance comparison**:
-   - Single-threaded: Should be faster than before (LRU cache now enabled)
-   - Parallel: Should be significantly faster on large codebases
-
----
-
-## 🎉 **CONCLUSION**
-
-The parallel processing implementation now produces **100% identical output** to single-threaded processing while maintaining all performance benefits:
-
-- **✅ Data Consistency**: Identical graph structure and relationships
-- **✅ Memory Efficiency**: Comprehensive leak prevention and monitoring  
-- **✅ Performance**: LRU caching enabled in both modes + parallel speedup
-- **✅ Reliability**: Proper error handling and resource cleanup
-
-**The system is now production-ready with both processing modes fully synchronized!** 🚀

GUARDRAILS.md

@@ -0,0 +1,88 @@
+# Guardrails — GitNexus (repo + agents)
+
+Rules for **human contributors** and **AI agents** working on this codebase or publishing artifacts. These complement `AGENTS.md` / `CLAUDE.md` (which focus on GitNexus-in-GitNexus workflows).
+
+## Scope (typical agent session)
+
+When automating changes in this repository, treat scope as **least privilege**:
+
+- **Read:** Source, tests, docs, public config as needed for the task.  
+- **Write:** Only files required for the requested fix or feature; avoid unrelated formatting or refactors.  
+- **Execute:** Tests, typecheck, and documented CLI commands; do not run destructive commands on user data outside the repo without explicit approval.  
+- **Off-limits:** Other people’s machines, production deployments you don’t own, and credentials you didn’t receive permission to use.
+
+Adjust explicitly if the maintainer defines a different scope for a task.
+
+---
+
+## Non-negotiables
+
+1. **Never commit secrets** — API keys, tokens, `.env` with real values, private URLs, or session cookies. Use `.env.example` with placeholders only.  
+2. **Never rename symbols with blind find-and-replace** when working in a GitNexus-indexed project — use the **`rename` MCP tool** with **`dry_run: true` first**, then review `graph` vs `text_search` edits. (There is no separate `gitnexus rename` CLI; renaming goes through MCP or editor integration.)  
+3. **Run impact analysis before editing shared symbols** — use **`impact`** (upstream) for functions/classes/methods others call; do not ignore **HIGH** / **CRITICAL** risk without maintainer sign-off.  
+4. **Prefer `detect_changes` before commit** — confirm diffs map to expected symbols/processes when the graph is available.  
+5. **Preserve embeddings** — if `.gitnexus/meta.json` shows embeddings, run `npx gitnexus analyze --embeddings` when refreshing the index; plain `analyze` can drop them.
+
+---
+
+## Signs (recurring failure patterns)
+
+Use this format: **Trigger → Instruction → Reason**.  
+Append new Signs here when the same mistake repeats (e.g. CI broken twice the same way).
+
+### Sign: Stale graph after edits
+
+- **Trigger:** MCP or resources warn the index is behind `HEAD`, or code search doesn’t match latest commit.  
+- **Instruction:** Run `npx gitnexus analyze` from the repo root (plus `--embeddings` if the project used them).  
+- **Reason:** Tools query LadybugDB built at last analyze; git changes are invisible until re-indexed.
+
+### Sign: Embeddings vanished after analyze
+
+- **Trigger:** Semantic search quality drops; `stats.embeddings` in `.gitnexus/meta.json` is 0 after a refresh.  
+- **Instruction:** Re-run `npx gitnexus analyze --embeddings` and confirm `meta.json` reflects stored embeddings.  
+- **Reason:** Embedding generation is opt-in; analyze without the flag does not preserve prior vectors.
+
+### Sign: MCP lists no repos
+
+- **Trigger:** MCP stderr says no indexed repos.  
+- **Instruction:** Run `npx gitnexus analyze` in the target repository; verify `npx gitnexus list` shows it.  
+- **Reason:** The MCP server discovers repos via `~/.gitnexus/registry.json`, populated by analyze.
+
+### Sign: Wrong repo in multi-repo setups
+
+- **Trigger:** Query/impact results clearly belong to another project.  
+- **Instruction:** Call `list_repos`, then pass **`repo`** on subsequent tools (or use per-workspace MCP config).  
+- **Reason:** Default target may be ambiguous when multiple repos are registered.
+
+### Sign: LadybugDB lock / “database busy”
+
+- **Trigger:** Errors opening `.gitnexus/lbug` while MCP and analyze both run.  
+- **Instruction:** Stop overlapping processes; one writer at a time. Retry analyze or restart MCP.  
+- **Reason:** Embedded DB expects single-process ownership of the store.
+
+---
+
+## Publishing & supply chain
+
+- **npm:** Do not publish from unreviewed automation; follow maintainer release process. Bump version intentionally; tag releases to match `package.json`.  
+- **Dependencies:** Prefer minimal, auditable changes to `package.json`; run tests and CI after lockfile updates.  
+- **License:** This project ships under **PolyForm Noncommercial 1.0.0** — do not relicense or imply a different license in docs or metadata without maintainer approval.
+
+---
+
+## Escalation
+
+Stop and ask a **human maintainer** when:
+
+- Impact analysis shows **HIGH** / **CRITICAL** risk and the task still requires the change.  
+- You need to alter **CI**, **release**, or **security-sensitive** config.  
+- Requirements conflict (e.g. “speed up analyze” vs “must keep all embeddings on huge repo”).  
+- You are unsure whether data loss is acceptable (`clean`, forced migrations, schema changes).
+
+---
+
+## Related docs
+
+- [ARCHITECTURE.md](ARCHITECTURE.md) — components and data flow.  
+- [RUNBOOK.md](RUNBOOK.md) — commands for recovery.  
+- [CONTRIBUTING.md](CONTRIBUTING.md) — PR and commit expectations.

KUZU_COPY_ENABLE_GUIDE.md

@@ -1,153 +0,0 @@
-# How to Enable KuzuDB COPY Feature
-
-## Quick Start
-
-To enable the new COPY-based bulk loading feature in GitNexus:
-
-### 1. Enable Feature Flag
-
-Edit your `gitnexus.config.ts` file and set:
-
-```typescript
-export default {
-  // ... other config
-  features: {
-    // ... other features
-    enableKuzuCopy: true,  // Enable COPY-based bulk loading
-    // ... other features
-  }
-}
-```
-
-### 2. Verify Configuration
-
-The feature flag can also be enabled via environment variables or runtime configuration. Check your current config with:
-
-```javascript
-// In browser console
-import { isKuzuCopyEnabled } from './src/config/features.ts';
-console.log('COPY enabled:', isKuzuCopyEnabled());
-```
-
-### 3. Monitor Performance
-
-Once enabled, you'll see different log messages in the browser console:
-
-**COPY Success:**
-```
-🚀 COPY: Starting COPY-based commit of 150 Function nodes
-📝 Written 12543 bytes to /temp_Function_nodes_1703123456789.csv
-✅ COPY: Successfully loaded 150 Function nodes via COPY
-```
-
-**COPY Fallback:**
-```
-⚠️ COPY failed for Function, falling back to MERGE: FS API not available
-🔄 BATCH: Committing 150 Function nodes in single query
-✅ BATCH: Successfully committed all nodes in batches
-```
-
-## Performance Expectations
-
-### Small Repositories (< 100 files)
-- **Improvement**: 2-3x faster
-- **COPY vs MERGE**: Minimal difference due to overhead
-
-### Medium Repositories (100-500 files)  
-- **Improvement**: 5-7x faster
-- **COPY vs MERGE**: Significant improvement in batch operations
-
-### Large Repositories (1000+ files)
-- **Improvement**: 10-15x faster
-- **COPY vs MERGE**: Dramatic improvement, especially for complex codebases
-
-## Troubleshooting
-
-### COPY Not Working
-
-1. **Check Feature Flag**: Ensure `enableKuzuCopy: true` in config
-2. **Browser Compatibility**: COPY requires Web Workers support
-3. **KuzuDB Version**: Ensure kuzu-wasm@0.11.1 or later
-4. **FS API**: Check browser console for FS API availability
-
-### Fallback to MERGE
-
-The system automatically falls back to MERGE if:
-- FS API is not available
-- COPY statement execution fails
-- CSV generation encounters errors
-- KuzuDB schema issues
-
-This ensures **zero downtime** and **no data loss**.
-
-### Common Error Messages
-
-| Error | Cause | Solution |
-|-------|-------|----------|
-| `FS API not available` | Browser/environment limitation | Normal fallback, no action needed |
-| `COPY failed: Table X does not exist` | Schema not initialized | Check KuzuDB schema initialization |
-| `CSV generation failed` | Data format issue | Check node/relationship properties |
-
-## Monitoring & Metrics
-
-### Success Indicators
-- ✅ COPY success messages in console
-- 📊 Faster ingestion times
-- 💾 Lower memory usage during batch operations
-
-### Performance Comparison
-```javascript
-// Before (MERGE): ~30 seconds for 1000 nodes
-🔄 BATCH: Committing 1000 Function nodes in single query
-✅ BATCH: Successfully committed all nodes in batches (29.8s)
-
-// After (COPY): ~3 seconds for 1000 nodes  
-🚀 COPY: Starting COPY-based commit of 1000 Function nodes
-✅ COPY: Successfully loaded 1000 Function nodes via COPY (2.9s)
-```
-
-## Rollback Plan
-
-To disable COPY and revert to MERGE:
-
-```typescript
-export default {
-  features: {
-    enableKuzuCopy: false,  // Disable COPY feature
-  }
-}
-```
-
-Changes take effect immediately on next repository ingestion.
-
-## Advanced Configuration
-
-### Batch Size Optimization
-
-The system automatically calculates optimal batch sizes, but you can tune performance:
-
-```typescript
-// In KuzuKnowledgeGraph initialization
-const kuzuGraph = new KuzuKnowledgeGraph(queryEngine, {
-  batchSize: 200,        // Increase for better COPY performance
-  autoCommit: true,      // Keep enabled for COPY
-  enableCache: true      // Recommended for performance
-});
-```
-
-### Memory Management
-
-For very large repositories, the system uses chunked processing:
-- **< 1000 items**: Single CSV generation
-- **1000-5000 items**: 1000-item chunks  
-- **> 5000 items**: 1500-item chunks
-
-## Next Steps
-
-1. **Enable Feature**: Set `enableKuzuCopy: true`
-2. **Test Small Repository**: Verify functionality with a small codebase
-3. **Monitor Performance**: Check console logs for COPY success
-4. **Scale Up**: Test with larger repositories
-5. **Report Issues**: Document any fallback scenarios or performance issues
-
-The COPY feature is designed to be **safe**, **fast**, and **transparent** - it should work seamlessly with your existing GitNexus workflow while providing significant performance improvements.

KUZU_COPY_IMPLEMENTATION_GUIDE.md

@@ -1,392 +0,0 @@
-# KuzuDB COPY Implementation Guide for GitNexus
-
-## Executive Summary
-
-**Status**: ✅ **PROVEN WORKING** - COPY approach successfully tested and verified
-**Performance**: 5-10x faster than current MERGE batch operations
-**Recommendation**: Implement with fallback to current MERGE approach
-
-## Test Results Summary
-
-| Component         | Status     | Details                                         |
-| ----------------- | ---------- | ----------------------------------------------- |
-| KuzuDB WASM Init  | ✅ Working | Initializes successfully in browser environment |
-| FS.writeFile      | ✅ Working | Successfully writes CSV to WASM filesystem      |
-| FS.readFile       | ✅ Working | Reads back data (fixed data type handling)      |
-| COPY Statements   | ✅ Working | Bulk loads data from CSV files                  |
-| Data Verification | ✅ Working | All data queryable after COPY operations        |
-
-## Technical Implementation Details
-
-### 1. Environment Requirements
-
-**Working Environment**: Browser with Web Workers support
-**Failed Environment**: Node.js (Worker2 constructor not available)
-**KuzuDB Version**: kuzu-wasm@0.11.1
-
-```javascript
-// Confirmed working initialization
-await kuzu.init();
-const db = new kuzu.Database('');  // In-memory database
-const conn = new kuzu.Connection(db);
-```
-
-### 2. FS API Implementation
-
-**Key Finding**: FS API is available and functional in browser environment
-
-```javascript
-// Verified working pattern
-await kuzu.FS.writeFile('/path/file.csv', csvData);
-const readData = await kuzu.FS.readFile('/path/file.csv');
-```
-
-**Critical Issue Solved**: FS.readFile data type handling
-
-- **Problem**: `readData.substring is not a function`
-- **Cause**: FS.readFile returns Buffer/Uint8Array, not string
-- **Solution**: Proper data type conversion
-
-```javascript
-// Fixed data handling
-let dataStr;
-if (typeof readData === 'string') {
-    dataStr = readData;
-} else if (readData instanceof Uint8Array || readData instanceof ArrayBuffer) {
-    dataStr = new TextDecoder().decode(readData);
-} else if (readData && readData.toString) {
-    dataStr = readData.toString();
-} else {
-    dataStr = String(readData);
-}
-```
-
-### 3. COPY Statement Implementation
-
-**Verified Working Pattern**:
-
-```javascript
-// 1. Write CSV to WASM filesystem
-await kuzu.FS.writeFile('/users.csv', csvData);
-
-// 2. Execute COPY statement
-const result = await conn.query("COPY User FROM '/users.csv'");
-await result.close();
-
-// 3. Data is immediately available for queries
-const verifyResult = await conn.query('MATCH (u:User) RETURN count(u)');
-```
-
-**CSV Format Requirements**:
-
-- Standard CSV format (comma-separated)
-- Header row with column names matching schema
-- Proper escaping for special characters
-- No additional formatting needed
-
-### 4. Schema Management
-
-**Critical Issue Solved**: Table existence conflicts
-
-- **Problem**: `Binder exception: User already exists in catalog`
-- **Cause**: Multiple test runs without cleanup
-- **Solution**: Drop tables before creation
-
-```javascript
-// Required cleanup pattern
-try {
-    await conn.query('DROP TABLE User IF EXISTS');
-    await conn.query('DROP TABLE City IF EXISTS');
-} catch (cleanupError) {
-    // Tables might not exist, ignore errors
-}
-
-// Then create fresh schema
-await conn.query('CREATE NODE TABLE User(name STRING, age INT64, PRIMARY KEY (name))');
-```
-
-## GitNexus Integration Strategy
-
-### 1. CSV Generator Service
-
-**Location**: `src/core/kuzu/csv-generator.ts`
-
-```typescript
-export class GitNexusCSVGenerator {
-    static generateNodeCSV(nodes: GraphNode[], label: string): string {
-        const filteredNodes = nodes.filter(node => node.label === label);
-        if (filteredNodes.length === 0) return '';
-      
-        // Get all unique properties for schema
-        const allProps = new Set(['id']);
-        filteredNodes.forEach(node => {
-            Object.keys(node.properties).forEach(key => allProps.add(key));
-        });
-      
-        const columns = Array.from(allProps);
-        const header = columns.join(',');
-      
-        const rows = filteredNodes.map(node => {
-            return columns.map(col => {
-                if (col === 'id') return this.escapeCSV(node.id);
-                const value = node.properties[col];
-                return value !== undefined ? this.escapeCSV(value) : '';
-            }).join(',');
-        });
-      
-        return [header, ...rows].join('\n');
-    }
-  
-    static generateRelationshipCSV(relationships: GraphRelationship[], type: string): string {
-        // Similar implementation for relationships
-        // Include source, target, and properties
-    }
-  
-    static escapeCSV(value: any): string {
-        if (value === null || value === undefined) return '';
-        const str = String(value);
-      
-        if (str.includes(',') || str.includes('"') || str.includes('\n')) {
-            return '"' + str.replace(/"/g, '""') + '"';
-        }
-        return str;
-    }
-}
-```
-
-### 2. Enhanced KuzuKnowledgeGraph
-
-**Location**: `src/core/graph/kuzu-knowledge-graph.ts`
-
-**Replace current batch methods**:
-
-```typescript
-// Current method (keep as fallback)
-private async commitNodesBatchWithMERGE(label: string, nodes: GraphNode[]): Promise<void> {
-    // Existing MERGE implementation
-}
-
-// New COPY method
-private async commitNodesBatchWithCOPY(label: string, nodes: GraphNode[]): Promise<void> {
-    try {
-        // Generate CSV
-        const csvData = GitNexusCSVGenerator.generateNodeCSV(nodes, label);
-      
-        // Write to WASM filesystem
-        const csvPath = `/temp_${label}_${Date.now()}.csv`;
-        await this.kuzuModule.FS.writeFile(csvPath, csvData);
-      
-        // Execute COPY statement
-        const result = await this.queryEngine.executeQuery(`COPY ${label} FROM '${csvPath}'`);
-        await result.close();
-      
-        console.log(`✅ COPY: Successfully loaded ${nodes.length} ${label} nodes`);
-      
-    } catch (error) {
-        console.error(`❌ COPY failed for ${label}:`, error);
-        throw error;
-    }
-}
-
-// Main batch method with fallback
-private async commitNodesBatch(label: string, nodes: GraphNode[]): Promise<void> {
-    try {
-        await this.commitNodesBatchWithCOPY(label, nodes);
-    } catch (copyError) {
-        console.warn(`⚠️ COPY failed, falling back to MERGE for ${label}:`, copyError.message);
-        await this.commitNodesBatchWithMERGE(label, nodes);
-    }
-}
-```
-
-### 3. KuzuDB Module Access
-
-**Location**: `src/core/kuzu/kuzu-npm-integration.ts`
-
-**Add FS API access**:
-
-```typescript
-// Expose FS API in KuzuInstance interface
-export interface KuzuInstance {
-    // ... existing methods
-    getFS(): any; // Access to FS API
-}
-
-// In createKuzuInstance()
-return {
-    // ... existing methods
-    getFS(): any {
-        return kuzuModule.default.FS;
-    }
-};
-```
-
-### 4. Integration Points
-
-**Files to modify**:
-
-1. `src/core/graph/kuzu-knowledge-graph.ts` - Add COPY batch methods
-2. `src/core/kuzu/kuzu-npm-integration.ts` - Expose FS API
-3. `src/core/kuzu/csv-generator.ts` - New CSV generation service
-4. `src/config/features.ts` - Add COPY feature flag
-
-**Feature Flag**:
-
-```typescript
-export function isKuzuCopyEnabled(): boolean {
-    return cachedConfig?.features.enableKuzuCopy ?? false;
-}
-```
-
-## Performance Characteristics
-
-### Current MERGE Approach
-
-- **Operations**: N individual MERGE statements per batch
-- **Memory**: String concatenation for large queries
-- **Database Load**: N query parsing operations
-- **Scalability**: Linear degradation with batch size
-
-### COPY Approach
-
-- **Operations**: 1 FS write + 1 COPY statement per batch
-- **Memory**: Streaming CSV generation
-- **Database Load**: 1 optimized bulk operation
-- **Scalability**: Constant time regardless of batch size
-
-### Expected Performance Gains
-
-- **Small batches (10-50 items)**: 2-3x improvement
-- **Medium batches (100-500 items)**: 5-7x improvement
-- **Large batches (1000+ items)**: 10-15x improvement
-
-## Error Handling Strategy
-
-### 1. Environment Detection
-
-```typescript
-function isCopySupported(): boolean {
-    return !!(kuzu.FS && kuzu.FS.writeFile && typeof kuzu.FS.writeFile === 'function');
-}
-```
-
-### 2. Graceful Degradation
-
-```typescript
-if (isCopySupported() && isKuzuCopyEnabled()) {
-    try {
-        await commitNodesBatchWithCOPY(label, nodes);
-    } catch (copyError) {
-        await commitNodesBatchWithMERGE(label, nodes);
-    }
-} else {
-    await commitNodesBatchWithMERGE(label, nodes);
-}
-```
-
-### 3. Error Categories
-
-- **FS Errors**: File system operations (writeFile/readFile)
-- **COPY Errors**: SQL execution errors (syntax, schema mismatch)
-- **Data Errors**: CSV format or encoding issues
-
-## Testing Strategy
-
-### 1. Unit Tests
-
-- CSV generation with various data types
-- Error handling for malformed data
-- Schema compatibility validation
-
-### 2. Integration Tests
-
-- End-to-end COPY workflow
-- Fallback mechanism verification
-- Performance benchmarking
-
-### 3. Browser Compatibility
-
-- Test across different browsers
-- Verify Web Worker support
-- Memory usage monitoring
-
-## Deployment Considerations
-
-### 1. Feature Flag Rollout
-
-- **Phase 1**: Internal testing with feature flag disabled
-- **Phase 2**: Gradual rollout to subset of users
-- **Phase 3**: Full deployment with monitoring
-
-### 2. Monitoring Metrics
-
-- COPY success/failure rates
-- Performance improvement measurements
-- Memory usage comparison
-- Error frequency and types
-
-### 3. Rollback Strategy
-
-- Feature flag can instantly disable COPY approach
-- Automatic fallback ensures no service disruption
-- Existing MERGE approach remains fully functional
-
-## Known Limitations
-
-### 1. Environment Constraints
-
-- **Browser Only**: COPY approach requires browser environment
-- **Web Workers**: Depends on Web Worker support
-- **Memory**: WASM filesystem is in-memory only
-
-### 2. Data Constraints
-
-- **CSV Format**: Data must be CSV-compatible
-- **File Paths**: Limited to WASM filesystem paths
-- **Encoding**: UTF-8 encoding required
-
-### 3. Schema Constraints
-
-- **Table Existence**: Tables must exist before COPY
-- **Column Matching**: CSV columns must match schema
-- **Data Types**: Proper type conversion required
-
-## Future Enhancements
-
-### 1. Streaming CSV Generation
-
-- Process large datasets without loading into memory
-- Incremental file writing for very large batches
-
-### 2. Parallel COPY Operations
-
-- Multiple concurrent COPY statements
-- Batch processing optimization
-
-### 3. Advanced Error Recovery
-
-- Partial batch recovery on COPY failures
-- Detailed error reporting and diagnostics
-
-## Implementation Checklist
-
-- [ ] Create CSV generator service
-- [ ] Implement COPY-based bulk loader
-- [ ] Add FS API access to KuzuInstance
-- [ ] Implement fallback strategy
-- [ ] Add feature flag support
-- [ ] Create comprehensive tests
-- [ ] Performance benchmarking
-- [ ] Documentation updates
-- [ ] Gradual rollout plan
-- [ ] Monitoring and alerting setup
-
-## Conclusion
-
-The COPY approach has been **proven to work** through comprehensive testing. Implementation should proceed with:
-
-1. **Immediate**: CSV generator and COPY bulk loader
-2. **Short-term**: Feature flag and fallback mechanism
-3. **Long-term**: Performance optimization and monitoring
-
-This implementation will provide significant performance improvements for GitNexus, especially for large repository ingestion scenarios.

KUZU_DB_INTEGRATION_STATUS.md

@@ -1,213 +0,0 @@
-# KuzuDB Integration Status Report
-
-## 🎉 **IMPLEMENTATION COMPLETE!**
-
-The full KuzuDB integration has been successfully implemented according to the implementation plan. The system now supports **dual-write functionality** where data is written to both JSON (primary) and KuzuDB (secondary) storage systems simultaneously.
-
----
-
-## ✅ **What's Been Implemented**
-
-### **Phase 1: Foundation Setup - COMPLETE**
-- ✅ **KuzuDB WASM Loader** (`src/core/kuzu/kuzu-loader.ts`) - Full implementation
-- ✅ **KuzuDB Query Engine** (`src/core/graph/kuzu-query-engine.ts`) - Complete with caching, transactions, performance monitoring
-- ✅ **KuzuDB Knowledge Graph** (`src/core/graph/kuzu-knowledge-graph.ts`) - Full implementation with batching and caching
-- ✅ **KuzuDB Schema Manager** (`src/core/kuzu/kuzu-schema.ts`) - Complete schema definitions for all node and relationship types
-- ✅ **Feature Flag Integration** - Full KuzuDB feature flag support
-
-### **Phase 2: Parallel Storage Implementation - COMPLETE**
-- ✅ **KuzuProcessorBase** - Abstract base class with dual-write pattern, transaction management, and statistics
-- ✅ **Enhanced StructureProcessor** - Dual-write support for Project, Folder, File nodes and CONTAINS relationships
-- ✅ **Enhanced ParsingProcessor** - Dual-write support for all definition nodes and relationships
-- ✅ **Enhanced ImportProcessor** - Dual-write support for IMPORTS relationships
-- ✅ **Enhanced CallProcessor** - Dual-write support for CALLS relationships
-
-### **Core Features Implemented**
-- ✅ **Dual-Write Pattern** - Data written to both JSON and KuzuDB simultaneously
-- ✅ **Transaction Management** - Begin, commit, rollback support
-- ✅ **Error Handling** - Graceful degradation to JSON-only mode
-- ✅ **Performance Monitoring** - Comprehensive statistics and timing metrics
-- ✅ **Batch Processing** - Optimized batch operations for better performance
-- ✅ **Caching System** - LRU cache for improved query performance
-- ✅ **Schema Validation** - Complete schema definitions for all node and relationship types
-
----
-
-## 🏗️ **Architecture Overview**
-
-```
-┌─────────────────────────────────────────────────────────────────┐
-│                    GitNexus KuzuDB Integration                  │
-├─────────────────────────────────────────────────────────────────┤
-│                                                                 │
-│  ┌─────────────────┐    ┌─────────────────┐    ┌──────────────┐ │
-│  │ JSON Storage    │    │ KuzuDB Storage  │    │ Feature Flags│ │
-│  │ (Primary)       │    │ (Secondary)     │    │ (Control)    │ │
-│  └─────────────────┘    └─────────────────┘    └──────────────┘ │
-│           │                       │                      │      │
-│           └───────────────────────┼──────────────────────┘      │
-│                                   │                             │
-│  ┌─────────────────────────────────────────────────────────────┐ │
-│  │              KuzuProcessorBase                              │ │
-│  │   • Dual-write pattern                                     │ │
-│  │   • Transaction management                                 │ │
-│  │   • Error handling & graceful degradation                 │ │
-│  │   • Performance monitoring & statistics                   │ │
-│  └─────────────────────────────────────────────────────────────┘ │
-│           │              │              │              │        │
-│  ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ ┌──────────┐ │
-│  │ Structure    │ │ Parsing      │ │ Import       │ │ Call     │ │
-│  │ Processor    │ │ Processor    │ │ Processor    │ │Processor │ │
-│  └──────────────┘ └──────────────┘ └──────────────┘ └──────────┘ │
-│                                                                 │
-└─────────────────────────────────────────────────────────────────┘
-```
-
----
-
-## 🚀 **How to Use KuzuDB Integration**
-
-### **1. Enable KuzuDB (Currently Disabled by Default)**
-
-```typescript
-import { featureFlags } from './src/config/feature-flags';
-
-// Enable KuzuDB integration
-featureFlags.enableKuzuDB();
-
-// Check status
-console.log('KuzuDB enabled:', featureFlags.getFlag('enableKuzuDB'));
-```
-
-### **2. Current Storage Behavior**
-
-**With KuzuDB Disabled (Default):**
-- ✅ Data stored in JSON format (existing functionality)
-- ✅ All processors work as before
-- ✅ No performance impact
-
-**With KuzuDB Enabled:**
-- ✅ Data written to **both** JSON and KuzuDB simultaneously
-- ✅ JSON remains primary storage (no breaking changes)
-- ✅ KuzuDB failures gracefully degrade to JSON-only mode
-- ✅ Enhanced logging and statistics available
-
-### **3. Enhanced Console Output**
-
-When KuzuDB is enabled, you'll see enhanced logging:
-
-```
-📁 Processing structure for MyProject with 150 paths...
-🚀 Initializing KuzuDB integration...
-✅ KuzuDB integration initialized successfully.
-✅ Structure processing completed. Hidden 45 items from display.
-
-📊 StructureProcessor Statistics:
-  Total Nodes Processed: 105
-  Total Relationships Processed: 104
-  KuzuDB Nodes Written: 105
-  KuzuDB Relationships Written: 104
-  KuzuDB Errors: 0
-  Processing Time: 1,234.56ms
-```
-
----
-
-## 📊 **Current Status**
-
-| Component | Status | Notes |
-|-----------|--------|-------|
-| **KuzuDB WASM Loader** | ✅ Complete | Ready for WASM binary integration |
-| **Query Engine** | ✅ Complete | Full Cypher query support, caching, transactions |
-| **Knowledge Graph** | ✅ Complete | Drop-in replacement for SimpleKnowledgeGraph |
-| **Schema Manager** | ✅ Complete | All node and relationship types defined |
-| **Dual-Write Pattern** | ✅ Complete | All 4 processors support dual-write |
-| **Feature Flags** | ✅ Complete | Full control over KuzuDB integration |
-| **Error Handling** | ✅ Complete | Graceful degradation to JSON-only mode |
-| **Performance Monitoring** | ✅ Complete | Comprehensive statistics and timing |
-| **Transaction Management** | ✅ Complete | ACID compliance with rollback support |
-
----
-
-## 🔧 **What's Missing (Optional Enhancements)**
-
-1. **KuzuDB WASM Binary**: Need to add the actual KuzuDB WASM file to `public/kuzu/`
-2. **Query Migration**: Phase 3 implementation (read operations from KuzuDB)
-3. **UI Integration**: Update UI components to use KuzuDB queries
-4. **Advanced Analytics**: Graph algorithms and complex queries
-
----
-
-## 🎯 **Key Benefits Achieved**
-
-### **1. Zero Breaking Changes**
-- All existing functionality preserved
-- JSON storage remains primary
-- Backward compatibility maintained
-
-### **2. Production-Ready Error Handling**
-- KuzuDB failures don't break the system
-- Graceful degradation to JSON-only mode
-- Comprehensive error logging
-
-### **3. Performance & Monitoring**
-- Detailed statistics for all operations
-- Performance timing and success rates
-- Transaction management with rollback
-
-### **4. Scalable Architecture**
-- Dual-write pattern supports gradual migration
-- Feature flags enable controlled rollout
-- Extensible base classes for future enhancements
-
----
-
-## 🧪 **Testing the Integration**
-
-### **Current Compilation Status**
-- ✅ **Core KuzuDB components compile successfully**
-- ✅ **All processors extend KuzuProcessorBase properly**
-- ✅ **Feature flags work correctly**
-- ⚠️ **Some test files need updates** (non-critical)
-- ⚠️ **Some UI components need interface updates** (non-critical)
-
-### **What You Can Test Now**
-1. **Enable KuzuDB via feature flags**
-2. **Run the ingestion pipeline** - it will attempt dual-write
-3. **Observe enhanced logging and statistics**
-4. **Verify graceful degradation** when KuzuDB WASM is not available
-
----
-
-## 📋 **Next Steps (Optional)**
-
-### **Phase 3: Query Migration** (Future Enhancement)
-1. Replace `graph.nodes.filter()` with KuzuDB queries
-2. Update UI components to use KuzuDB query results
-3. Implement query performance comparisons
-
-### **Phase 4: JSON Deprecation** (Future Enhancement)
-1. Remove dual-write pattern
-2. Make KuzuDB the primary storage
-3. Implement advanced graph analytics
-
----
-
-## 🎉 **Conclusion**
-
-**The KuzuDB integration is FULLY IMPLEMENTED and ready for use!** 
-
-The system now supports:
-- ✅ **Dual-write functionality** (JSON + KuzuDB)
-- ✅ **Complete error handling** and graceful degradation
-- ✅ **Production-ready architecture** with monitoring and statistics
-- ✅ **Feature flag control** for safe deployment
-- ✅ **Zero breaking changes** to existing functionality
-
-You can now:
-1. **Enable KuzuDB** via feature flags
-2. **Test the dual-write system** with any repository
-3. **Observe enhanced logging** and performance metrics
-4. **Add the KuzuDB WASM binary** when ready for full functionality
-
-The foundation is solid and ready for the next phases of the migration plan! 🚀

LICENSE

@@ -0,0 +1,73 @@
+PolyForm Noncommercial License 1.0.0
+
+<https://polyformproject.org/licenses/noncommercial/1.0.0>
+
+## Acceptance
+
+In order to get any license under these terms, you must agree to them as both strict obligations and conditions to all your licenses.
+
+## Copyright License
+
+The licensor grants you a copyright license for the software to do everything you might do with the software that would otherwise infringe the licensor's copyright in it for any permitted purpose.  However, you may only distribute the software according to [Distribution License](#distribution-license) and make changes or new works based on the software according to [Changes and New Works License](#changes-and-new-works-license).
+
+## Distribution License
+
+The licensor grants you an additional copyright license to distribute copies of the software.  Your license to distribute covers distributing the software with changes and new works permitted by [Changes and New Works License](#changes-and-new-works-license).
+
+## Notices
+
+You must ensure that anyone who gets a copy of any part of the software from you also gets a copy of these terms or the URL for them above, as well as copies of any plain-text lines beginning with `Required Notice:` that the licensor provided with the software.  For example:
+
+> Required Notice: Copyright Abhigyan Patwari (https://github.com/abhigyanpatwari/GitNexus)
+
+## Changes and New Works License
+
+The licensor grants you an additional copyright license to make changes and new works based on the software for any permitted purpose.
+
+## Patent License
+
+The licensor grants you a patent license for the software that covers patent claims the licensor can license, or becomes able to license, that you would infringe by using the software.
+
+## Noncommercial Purposes
+
+Any noncommercial purpose is a permitted purpose.
+
+## Personal Uses
+
+Personal use for research, experiment, and testing for the benefit of public knowledge, personal study, private entertainment, hobby projects, amateur pursuits, or religious observance, without any anticipated commercial application, is use for a permitted purpose.
+
+## Noncommercial Organizations
+
+Use by any charitable organization, educational institution, public research organization, public safety or health organization, environmental protection organization, or government institution is use for a permitted purpose regardless of the source of funding or obligations resulting from the funding.
+
+## Fair Use
+
+You may have "fair use" rights for the software under the law. These terms do not limit them.
+
+## No Other Rights
+
+These terms do not allow you to sublicense or transfer any of your licenses to anyone else, or prevent the licensor from granting licenses to anyone else.  These terms do not imply any other licenses.
+
+## Patent Defense
+
+If you make any written claim that the software infringes or contributes to infringement of any patent, your patent license for the software granted under these terms ends immediately. If your company makes such a claim, your patent license ends immediately for work on behalf of your company.
+
+## Violations
+
+The first time you are notified in writing that you have violated any of these terms, or done anything with the software not covered by your licenses, your licenses can nonetheless continue if you come into full compliance with these terms, and take practical steps to correct past violations, within 32 days of receiving notice.  Otherwise, all your licenses end immediately.
+
+## No Liability
+
+***As far as the law allows, the software comes as is, without any warranty or condition, and the licensor will not be liable to you for any damages arising out of these terms or the use or nature of the software, under any kind of legal claim.***
+
+## Definitions
+
+The **licensor** is the individual or entity offering these terms, and the **software** is the software the licensor makes available under these terms.
+
+**You** refers to the individual or entity agreeing to these terms.
+
+**Your company** is any legal entity, sole proprietorship, or other kind of organization that you work for, plus all organizations that have control over, are under the control of, or are under common control with that organization.  **Control** means ownership of substantially all the assets of an entity, or the power to direct its management and policies by vote, contract, or otherwise.  Control can be direct or indirect.
+
+**Your licenses** are all the licenses granted to you for the software under these terms.
+
+**Use** means anything you do with the software requiring one of your licenses.

MEMORY_LEAK_FIXES.md

@@ -1,163 +0,0 @@
-# Memory Leak Fixes for Worker Pool Parallel Processing
-
-## Issues Identified and Fixed
-
-### 1. **Event Listener Memory Leaks** ✅ FIXED
-**Problem**: WebWorkerPool wasn't properly cleaning up event listeners during shutdown, causing memory leaks.
-
-**Fix Applied**:
-- Added proper event listener cleanup in `shutdown()` method
-- Set `worker.onmessage = null`, `worker.onerror = null`, `worker.onmessageerror = null` before terminating workers
-- Clear the `eventListeners` Map during shutdown
-
-**Files Modified**: `src/lib/web-worker-pool.ts`
-
-### 2. **Singleton Worker Pool Issues** ✅ FIXED  
-**Problem**: FileProcessingPool singleton instances were never cleaned up, accumulating memory over time.
-
-**Fix Applied**:
-- Added `shutdownInstance()` static method to properly cleanup singleton instances
-- Added `hasInstance()` method to check if instance exists
-- Added `cleanupAllPools()` utility method in WebWorkerPoolUtils
-
-**Files Modified**: `src/lib/web-worker-pool.ts`
-
-### 3. **Worker Error Handling** ✅ FIXED
-**Problem**: Errors in worker termination could leave resources hanging.
-
-**Fix Applied**:
-- Improved `handleWorkerError()` method to properly cleanup worker event listeners
-- Added try-catch around worker termination
-- Ensure workers are removed from pools even on errors
-
-**Files Modified**: `src/lib/web-worker-pool.ts`
-
-### 4. **Missing Memory Monitoring** ✅ FIXED
-**Problem**: No memory usage tracking or automatic cleanup triggers.
-
-**Fix Applied**:
-- Added `monitorMemoryUsage()` method with automatic cleanup triggers
-- Added memory usage estimation in `getStats()` method
-- Created periodic memory monitoring in ParallelParsingProcessor
-- Added global cleanup handlers for page unload and visibility changes
-
-**Files Modified**: `src/lib/web-worker-pool.ts`, `src/core/ingestion/parallel-parsing-processor.ts`
-
-### 5. **AST Map Accumulation** ✅ FIXED
-**Problem**: Large AST maps and function tries were kept in memory without limits.
-
-**Fix Applied**:
-- Added `MAX_AST_MAP_SIZE` constant (1000 entries)
-- Implemented `cleanupASTMap()` method to remove old entries when limit is exceeded
-- Added proper cleanup of AST maps, processed files, and function tries in shutdown
-- Added Tree-sitter parser cleanup with `parser.delete()`
-
-**Files Modified**: `src/core/ingestion/parallel-parsing-processor.ts`
-
-### 6. **Pipeline Cleanup Issues** ✅ FIXED
-**Problem**: Parallel pipeline wasn't ensuring proper cleanup on errors.
-
-**Fix Applied**:
-- Enhanced cleanup method to call `WebWorkerPoolUtils.cleanupAllPools()`
-- Added cleanup in both error and finally blocks
-- Ensured cleanup happens even when errors occur
-
-**Files Modified**: `src/core/ingestion/parallel-pipeline.ts`
-
-## New Features Added
-
-### Memory Monitoring System
-- **Automatic Memory Monitoring**: Checks memory usage every 30 seconds
-- **Threshold-based Cleanup**: Triggers cleanup when memory usage exceeds 500MB
-- **AST Map Size Limiting**: Automatically cleans up old AST entries when limit exceeded
-- **Global Memory Monitoring**: Monitors overall browser memory usage
-
-### Global Cleanup Handlers
-- **Page Unload Cleanup**: Automatically cleans up when user closes/refreshes page
-- **Tab Visibility Cleanup**: Cleans up when user switches tabs (page becomes hidden)
-- **Manual Cleanup Functions**: Utilities for forcing cleanup when needed
-
-### Enhanced Error Handling
-- **Graceful Worker Termination**: Proper cleanup even when workers fail
-- **Resource Leak Prevention**: Ensures all event listeners and references are cleared
-- **Error Recovery**: System continues working even if some workers fail
-
-## Usage Instructions
-
-### 1. Initialize Cleanup Handlers (RECOMMENDED)
-```typescript
-import { initializeWorkerPoolCleanup } from './src/lib/worker-pool-init.js';
-
-// Call this once when your app starts
-initializeWorkerPoolCleanup();
-```
-
-### 2. Manual Cleanup (if needed)
-```typescript
-import { cleanupWorkerPools } from './src/lib/worker-pool-init.js';
-
-// Force cleanup when needed
-await cleanupWorkerPools();
-```
-
-### 3. Monitor Memory Usage
-```typescript
-import { getMemoryInfo } from './src/lib/worker-pool-init.js';
-
-const memInfo = getMemoryInfo();
-if (memInfo) {
-  console.log(`Memory: ${memInfo.usedMB}MB / ${memInfo.totalMB}MB (${memInfo.percentage}%)`);
-}
-```
-
-## Performance Improvements
-
-### Before Fixes:
-- Worker pools accumulated without cleanup
-- Event listeners remained attached after worker termination
-- AST maps grew unbounded causing memory bloat
-- No automatic memory management
-
-### After Fixes:
-- **Automatic Resource Cleanup**: All resources properly cleaned up
-- **Memory Usage Monitoring**: Real-time monitoring with automatic cleanup triggers
-- **Bounded Memory Growth**: AST maps and other data structures have size limits
-- **Graceful Shutdown**: Proper cleanup on app/tab close
-
-## Monitoring and Debugging
-
-The system now provides detailed logging for:
-- Memory usage statistics
-- Worker pool status
-- Cleanup operations
-- Error conditions
-
-Check browser console for messages like:
-```
-ParallelParsingProcessor Memory Stats:
-- Memory Manager: 245.67MB used, 150 files cached
-- AST Map: 750 entries
-- Processed Files: 890 entries
-
-Memory usage: 245.67MB / 512.00MB (47.98%)
-```
-
-## Files Created/Modified
-
-### Modified Files:
-- `src/lib/web-worker-pool.ts` - Enhanced with memory leak fixes
-- `src/core/ingestion/parallel-parsing-processor.ts` - Added memory monitoring and cleanup
-- `src/core/ingestion/parallel-pipeline.ts` - Enhanced cleanup handling
-
-### New Files:
-- `src/lib/worker-pool-init.ts` - Initialization and utility functions
-- `MEMORY_LEAK_FIXES.md` - This documentation
-
-## Testing Recommendations
-
-1. **Monitor Memory Usage**: Watch browser's Task Manager during large codebase processing
-2. **Test Tab Switching**: Switch tabs during processing to verify cleanup triggers
-3. **Test Page Refresh**: Refresh page during processing to ensure proper cleanup
-4. **Long-running Tests**: Process multiple large codebases to verify no memory accumulation
-
-The system should now maintain stable memory usage even during intensive parallel processing operations.

PARALLEL_PROCESSING_VERIFICATION.md

@@ -1,154 +0,0 @@
-# Parallel Processing Verification & Fixes
-
-## 🎯 **Goal: Ensure Parallel Processing Produces Identical Output to Single-Threaded**
-
-## ❌ **Critical Issues Found and Fixed**
-
-### **Issue #1: Wrong Pipeline Class in Worker** 🚨 **CRITICAL**
-**Problem**: The `IngestionWorker` was always using `GraphPipeline` (single-threaded) instead of `ParallelGraphPipeline` when parallel processing was enabled.
-
-**Impact**: Even when "parallel processing" was enabled, it was actually running single-threaded processing in the worker, just with the parallel flag set.
-
-**Fix Applied**:
-```typescript
-// BEFORE (BROKEN)
-export class IngestionWorker {
-  private pipeline: GraphPipeline; // Always single-threaded!
-  constructor() {
-    this.pipeline = new GraphPipeline(); // Wrong!
-  }
-}
-
-// AFTER (FIXED)
-export class IngestionWorker {
-  private pipeline: GraphPipeline | ParallelGraphPipeline;
-  constructor() {
-    if (isParallelParsingEnabled()) {
-      this.pipeline = new ParallelGraphPipeline(); // Correct!
-    } else {
-      this.pipeline = new GraphPipeline();
-    }
-  }
-}
-```
-
-### **Issue #2: Incorrect Duplicate Detection** 🚨 **CRITICAL**
-**Problem**: Different duplicate detection logic between processors.
-
-**Single-threaded (CORRECT)**:
-```typescript
-if (this.duplicateDetector.checkAndMark(nodeId)) continue; // ✅ Checks AND marks
-```
-
-**Parallel (BROKEN)**:
-```typescript
-if (this.duplicateDetector.isDuplicate(nodeId)) return; // ❌ Only checks, doesn't mark
-```
-
-**Impact**: Parallel processor could create duplicate nodes because it wasn't marking them as processed.
-
-**Fix Applied**: Changed parallel processor to use `checkAndMark()`.
-
-### **Issue #3: Property Format Inconsistency** 🚨 **MEDIUM**
-**Problem**: Node properties stored in different formats.
-
-**Single-threaded**:
-```typescript
-decorators: def.decorators,        // Array format
-extends: def.extends,              // Array format  
-implements: def.implements,        // Array format
-```
-
-**Parallel (BROKEN)**:
-```typescript
-decorators: definition.decorators?.join(', '),   // String format ❌
-extends: definition.extends?.join(', '),         // String format ❌
-implements: definition.implements?.join(', '),   // String format ❌
-```
-
-**Impact**: Import/call processors expecting arrays would fail or produce different results.
-
-**Fix Applied**: Made parallel processor store arrays to match single-threaded.
-
-### **Issue #4: Progress Callback Integration** ✅ **ENHANCEMENT**
-**Problem**: Worker wasn't properly forwarding progress updates from `ParallelGraphPipeline`.
-
-**Fix Applied**: Added proper progress callback integration.
-
-## ✅ **Verification Checklist**
-
-### **Pipeline Selection** ✅
-- [x] Worker uses correct pipeline class based on feature flag
-- [x] `ParallelGraphPipeline` used when `isParallelParsingEnabled() === true`
-- [x] `GraphPipeline` used when `isParallelParsingEnabled() === false`
-
-### **Data Processing** ✅
-- [x] Duplicate detection logic identical (`checkAndMark()`)
-- [x] Node property formats identical (arrays not strings)
-- [x] Node ID generation identical
-- [x] Node label mapping identical
-
-### **Graph Structure** ✅
-- [x] Same 4-pass pipeline structure
-- [x] Same processor sequence (Structure → Parsing → Import → Call)
-- [x] Same AST map and function registry handling
-- [x] Same relationship creation logic
-
-### **Memory Management** ✅
-- [x] Both use LRU cache service
-- [x] Both maintain AST maps for compatibility
-- [x] Proper cleanup in both modes
-
-## 🔍 **Expected Behavior After Fixes**
-
-### **Single-threaded Mode**:
-- Uses `GraphPipeline` 
-- Uses `ParsingProcessor`
-- Sequential file processing
-- Direct definition extraction
-
-### **Parallel Mode**:
-- Uses `ParallelGraphPipeline`
-- Uses `ParallelParsingProcessor` 
-- Worker pool parallel processing
-- Worker-extracted definitions + main thread AST recreation
-
-### **Identical Output**:
-Both modes should now produce:
-- ✅ Same node counts by type
-- ✅ Same relationship counts by type  
-- ✅ Same import relationships (IMPORTS, DEPENDS_ON)
-- ✅ Same function call relationships (CALLS)
-- ✅ Same definition nodes with identical properties
-- ✅ Same graph connectivity
-
-## 🧪 **Testing Recommendations**
-
-1. **Process the same codebase** with both modes enabled/disabled
-2. **Compare graph statistics** - nodes by type, relationships by type
-3. **Verify specific relationships** - check for import and call relationships
-4. **Check isolated nodes** - should be minimal in both modes
-5. **Performance comparison** - parallel should be faster on large codebases
-
-## 📊 **Success Metrics**
-
-The parallel processing should now show:
-```
-✅ Relationships by type: {CONTAINS: X, DEFINES: Y, IMPORTS: Z, CALLS: W}
-✅ No "missing import relationships" warnings
-✅ No "missing function call relationships" warnings  
-✅ Same graph node/relationship counts as single-threaded
-```
-
-Instead of the previous broken output:
-```
-❌ Relationships by type: {CONTAINS: 103, DEFINES: 1971} // Missing IMPORTS/CALLS!
-❌ "No import relationships found between files"
-❌ "No function call relationships found"
-```
-
-## 🎯 **Conclusion**
-
-The parallel processing implementation now uses the correct pipeline classes and processing logic to produce **identical output** to single-threaded mode, while maintaining the performance benefits of parallel worker pool processing.
-
-**The root cause was using the wrong pipeline class in the worker** - a simple but critical configuration issue that made "parallel processing" actually run single-threaded code with inconsistent data structures.

PARSING_AND_STORAGE_TECHNICAL_DOCUMENTATION.md

@@ -1,623 +0,0 @@
-# GitNexus Parsing and Storage Technical Documentation
-## Complete Data Flow Analysis for Kuzu DB Migration
-
-> **Purpose**: This document provides an extremely detailed, line-by-line analysis of GitNexus's current parsing and storage implementation to facilitate the migration from JSON-based storage to Kuzu DB.
-
----
-
-## Executive Summary
-
-GitNexus uses a **4-pass ingestion pipeline** that processes code repositories into a knowledge graph stored in JSON format. The system employs in-memory data structures (`SimpleKnowledgeGraph`) with JSON serialization for persistence. This document traces every step of the data transformation process to enable precise Kuzu DB migration.
-
-### Key Storage Points Identified:
-1. **In-Memory Graph Storage**: `SimpleKnowledgeGraph` class with arrays
-2. **JSON Export/Import**: Via `src/lib/export.ts` functions
-3. **LRU Cache Storage**: For AST and parsing results
-4. **LocalStorage**: For settings, feature flags, and chat history
-5. **IndexedDB**: Planned for KuzuDB persistence (WIP)
-
----
-
-## 1. Core Data Structures
-
-### 1.1 Knowledge Graph Structure
-
-**File**: `src/core/graph/types.ts`
-
-```typescript
-// Primary graph interface - this is what gets stored
-export interface KnowledgeGraph {
-  nodes: GraphNode[];           // Array of all nodes
-  relationships: GraphRelationship[];  // Array of all relationships
-}
-
-// Node structure - every entity in the system
-export interface GraphNode {
-  id: string;                   // Unique identifier (generated)
-  label: NodeLabel;             // Type classification
-  properties: NodeProperties;   // All metadata as key-value pairs
-}
-
-// Relationship structure - connections between nodes
-export interface GraphRelationship {
-  id: string;                   // Unique identifier (generated)
-  type: RelationshipType;       // Relationship classification
-  source: string;               // Source node ID
-  target: string;               // Target node ID  
-  properties: RelationshipProperties; // Metadata as key-value pairs
-}
-```
-
-**Node Types** (`NodeLabel`):
-- `'Project'` - Repository root
-- `'Folder'` - Directory nodes
-- `'File'` - Source files
-- `'Function'` - Function definitions
-- `'Class'` - Class definitions
-- `'Method'` - Class methods
-- `'Variable'` - Variable declarations
-- `'Interface'` - TypeScript interfaces
-- `'Decorator'` - Python/TS decorators
-- `'Import'` - Import statements
-- `'Type'` - Type definitions
-- `'CodeElement'` - Generic code elements
-
-**Relationship Types** (`RelationshipType`):
-- `'CONTAINS'` - Hierarchical containment (folder → file, file → function)
-- `'CALLS'` - Function/method calls
-- `'INHERITS'` - Class inheritance
-- `'OVERRIDES'` - Method overrides
-- `'IMPORTS'` - Module imports
-- `'IMPLEMENTS'` - Interface implementations
-- `'DECORATES'` - Decorator applications
-
-### 1.2 Implementation Class
-
-**File**: `src/core/graph/graph.ts`
-
-```typescript
-export class SimpleKnowledgeGraph implements KnowledgeGraph {
-  nodes: GraphNode[] = [];              // Simple array storage
-  relationships: GraphRelationship[] = []; // Simple array storage
-
-  addNode(node: GraphNode): void {
-    this.nodes.push(node);              // Direct array append
-  }
-
-  addRelationship(relationship: GraphRelationship): void {
-    this.relationships.push(relationship); // Direct array append
-  }
-}
-```
-
-**Critical Storage Characteristics**:
-- **No indexing**: Linear search for node/relationship lookups
-- **No constraints**: No validation of referential integrity
-- **Memory-only**: No built-in persistence
-- **Simple append**: No deduplication or conflict resolution
-
----
-
-## 2. Four-Pass Ingestion Pipeline
-
-The pipeline transforms raw repository data through four distinct phases, each building upon the previous:
-
-### Pass 1: Structure Analysis (`StructureProcessor`)
-
-**File**: `src/core/ingestion/structure-processor.ts`
-
-**Input**: 
-- `projectRoot: string` - Repository path
-- `projectName: string` - Repository name  
-- `filePaths: string[]` - All discovered file paths
-
-**Process**:
-1. **Project Node Creation** (Lines 58-61):
-   ```typescript
-   const projectNode = this.createProjectNode(projectName, projectRoot);
-   graph.addNode(projectNode);  // STORAGE POINT 1
-   ```
-
-2. **Path Categorization** (Lines 87-127):
-   ```typescript
-   const { directories, files } = this.categorizePaths(filePaths);
-   // Separates files from directories using path analysis
-   ```
-
-3. **Directory Node Creation** (Lines 147-174):
-   ```typescript
-   const directoryNodes = this.createDirectoryNodes(visibleDirectories);
-   directoryNodes.forEach(node => graph.addNode(node)); // STORAGE POINT 2
-   ```
-
-4. **File Node Creation** (Lines 179-208):
-   ```typescript
-   const fileNodes = this.createFileNodes(visibleFiles);
-   fileNodes.forEach(node => graph.addNode(node)); // STORAGE POINT 3
-   ```
-
-5. **CONTAINS Relationship Creation** (Lines 213-255):
-   ```typescript
-   this.createContainsRelationships(graph, projectNode.id, visibleDirectories, visibleFiles);
-   // Creates hierarchical relationships - STORAGE POINT 4
-   ```
-
-**Storage Pattern**: Direct `graph.addNode()` and `graph.addRelationship()` calls append to arrays.
-
-### Pass 2: Code Parsing (`ParsingProcessor` / `ParallelParsingProcessor`)
-
-**Files**: 
-- `src/core/ingestion/parsing-processor.ts`
-- `src/core/ingestion/parallel-parsing-processor.ts`
-
-**Input**:
-- `filePaths: string[]` - Files to parse
-- `fileContents: Map<string, string>` - File content mapping
-- `options?: ParsingOptions` - Filtering options
-
-**Critical Data Structures**:
-
-1. **AST Storage** (Line 55 in `parsing-processor.ts`):
-   ```typescript
-   private astMap: Map<string, ParsedAST> = new Map();
-   // STORAGE POINT 5 - AST trees indexed by file path
-   ```
-
-2. **Function Registry** (Line 56):
-   ```typescript
-   private functionTrie: FunctionRegistryTrie = new FunctionRegistryTrie();
-   // STORAGE POINT 6 - Searchable function definitions
-   ```
-
-**Process Flow**:
-
-1. **File Filtering** (Lines 116-152):
-   ```typescript
-   const filteredFiles = this.applyFiltering(filePaths, fileContents, options);
-   // Applies directory and extension filters
-   ```
-
-2. **Tree-sitter Initialization** (Lines 84, 245):
-   ```typescript
-   await this.initializeParser();
-   // Loads WASM parsers for each language
-   ```
-
-3. **Batch Processing** (Lines 86-108):
-   ```typescript
-   const batchProcessor = new BatchProcessor<string, void>(BATCH_SIZE, async (filePaths: string[]) => {
-     for (const filePath of filePaths) {
-       await this.parseFile(graph, filePath, content); // CRITICAL PARSING
-       this.processedFiles.add(filePath);
-     }
-   });
-   ```
-
-4. **Definition Extraction** (`parseFile` method):
-   ```typescript
-   // Extracts: functions, classes, methods, variables, interfaces, types
-   // Each creates nodes via: graph.addNode(definitionNode) - STORAGE POINT 7
-   ```
-
-**Parallel Processing Variant**:
-- Uses Web Workers for CPU-intensive parsing
-- Results aggregated in main thread
-- Same storage patterns but with worker coordination
-
-### Pass 3: Import Resolution (`ImportProcessor`)
-
-**File**: `src/core/ingestion/import-processor.ts`
-
-**Input**:
-- `graph: KnowledgeGraph` - Current graph state
-- `astMap: Map<string, ParsedAST>` - Parsed ASTs
-- `fileContents: Map<string, string>` - File contents
-
-**Critical Data Structure**:
-```typescript
-interface ImportMap {
-  [importingFile: string]: {
-    [localName: string]: {
-      targetFile: string;
-      exportedName: string;
-      importType: 'default' | 'named' | 'namespace' | 'dynamic';
-    }
-  }
-}
-private importMap: ImportMap = {}; // STORAGE POINT 8
-```
-
-**Process Flow**:
-
-1. **Import Extraction** (Lines 84-88):
-   ```typescript
-   for (const [filePath, ast] of astMap) {
-     const fileImports = await this.processFileImports(filePath, ast, graph);
-     // Extracts import statements from AST
-   }
-   ```
-
-2. **Language-Specific Processing**:
-   - **JavaScript/TypeScript** (Lines 231-324): Handles ES6 imports, CommonJS requires
-   - **Python** (Lines 160-226): Handles `import` and `from...import` statements
-
-3. **Module Path Resolution** (Lines 522-606):
-   ```typescript
-   private resolveModulePath(moduleName: string, importingFile: string, language: string): string
-   // Resolves relative and absolute imports to actual file paths
-   ```
-
-4. **Relationship Creation** (Lines 611-645):
-   ```typescript
-   private createImportRelationship(graph: KnowledgeGraph, importInfo: ImportInfo): void {
-     // Creates IMPORTS relationships - STORAGE POINT 9
-     graph.relationships.push(relationship);
-   }
-   ```
-
-### Pass 4: Call Resolution (`CallProcessor`)
-
-**File**: `src/core/ingestion/call-processor.ts`
-
-**Input**:
-- `graph: KnowledgeGraph` - Current graph
-- `astMap: Map<string, ParsedAST>` - ASTs for call extraction
-- `importMap: ImportMap` - Import resolution data
-
-**Process Flow**:
-
-1. **Call Extraction** (Lines 86-120):
-   ```typescript
-   private async processFileCalls(filePath: string, ast: ParsedAST, graph: KnowledgeGraph) {
-     const calls = this.extractFunctionCalls(ast.tree!.rootNode, filePath);
-     // Extracts function/method call sites from AST
-   }
-   ```
-
-2. **3-Stage Resolution Strategy** (Lines 146-162):
-   ```typescript
-   // Stage 1: Exact Match using ImportMap (High Confidence)
-   // Stage 2: Same-Module Match (Medium Confidence)  
-   // Stage 3: Heuristic Fallback (Low Confidence)
-   ```
-
-3. **Call Relationship Creation** (Lines 99-101):
-   ```typescript
-   if (resolution.success && resolution.targetNodeId) {
-     this.createCallRelationship(graph, call, resolution.targetNodeId);
-     // Creates CALLS relationships - STORAGE POINT 10
-   }
-   ```
-
----
-
-## 3. JSON Storage Implementation
-
-### 3.1 Export Functions
-
-**File**: `src/lib/export.ts`
-
-**Primary Export Function** (Lines 34-68):
-```typescript
-export function exportGraphToJSON(
-  graph: KnowledgeGraph,
-  options: ExportOptions = {},
-  fileContents?: Map<string, string>,
-  processingStats?: { duration: number }
-): string {
-  
-  // Metadata wrapper structure
-  if (includeMetadata) {
-    const metadata: ExportMetadata = {
-      exportedAt: includeTimestamp ? new Date().toISOString() : '',
-      version: '1.0.0',
-      nodeCount: graph.nodes.length,
-      relationshipCount: graph.relationships.length,
-      fileCount: fileContents?.size,
-      processingDuration: processingStats?.duration
-    };
-
-    exportData = {
-      metadata,
-      graph,  // CRITICAL: Raw graph object serialization
-      ...(fileContents && { fileContents: Object.fromEntries(fileContents) })
-    };
-  } else {
-    exportData = graph; // Direct graph serialization
-  }
-
-  return JSON.stringify(exportData, null, prettyPrint ? 2 : 0);
-  // STORAGE POINT 11 - JSON string generation
-}
-```
-
-**JSON Structure**:
-```json
-{
-  "metadata": {
-    "exportedAt": "2024-01-01T00:00:00.000Z",
-    "version": "1.0.0", 
-    "nodeCount": 1250,
-    "relationshipCount": 3400,
-    "fileCount": 45,
-    "processingDuration": 2500
-  },
-  "graph": {
-    "nodes": [
-      {
-        "id": "node_project_abc123",
-        "label": "Project",
-        "properties": {
-          "name": "MyProject",
-          "path": "/path/to/project",
-          "createdAt": "2024-01-01T00:00:00.000Z"
-        }
-      }
-      // ... more nodes
-    ],
-    "relationships": [
-      {
-        "id": "rel_contains_def456", 
-        "type": "CONTAINS",
-        "source": "node_project_abc123",
-        "target": "node_folder_ghi789",
-        "properties": {}
-      }
-      // ... more relationships
-    ]
-  },
-  "fileContents": {
-    "src/main.ts": "export function main() { ... }",
-    // ... more file contents
-  }
-}
-```
-
-### 3.2 Import Functions
-
-**Import Function** (Lines 411-443):
-```typescript
-export function importGraphFromJSON(jsonString: string): {
-  graph: KnowledgeGraph;
-  metadata?: ExportMetadata;
-  fileContents?: Map<string, string>;
-} {
-  try {
-    const parsed = JSON.parse(jsonString); // DESERIALIZATION POINT
-    
-    if (parsed.metadata && parsed.graph) {
-      // Handle wrapped format
-      const result = {
-        graph: parsed.graph, // Direct object assignment
-        metadata: parsed.metadata
-      };
-      
-      if (parsed.fileContents) {
-        result.fileContents = new Map(Object.entries(parsed.fileContents));
-        // Convert plain object back to Map
-      }
-      
-      return result;
-    }
-    
-    return { graph: parsed }; // Direct graph object
-  } catch (error) {
-    throw new Error(`Failed to import graph: ${error.message}`);
-  }
-}
-```
-
-### 3.3 Download Implementation
-
-**File Download** (Lines 182-207):
-```typescript
-export function downloadJSON(content: string, filename: string): void {
-  // Create blob with JSON content
-  const blob = new Blob([content], { type: 'application/json' });
-  
-  // Browser download mechanism
-  const url = URL.createObjectURL(blob);
-  const link = document.createElement('a');
-  link.href = url;
-  link.download = filename;
-  link.click(); // Triggers download - PERSISTENCE POINT
-}
-```
-
----
-
-## 4. Additional Storage Mechanisms
-
-### 4.1 LRU Cache Storage
-
-**Files**: 
-- `src/services/memory-manager.ts` (Memory management)
-- Various processor classes (Cache usage)
-
-**Purpose**: Caches parsed ASTs and query results for performance
-
-**Implementation Pattern**:
-```typescript
-// Cache key generation
-const cacheKey = this.lruCache.generateFileCacheKey(filePath, contentHash);
-
-// Cache retrieval
-const cachedResult = this.lruCache.getParsedFile(cacheKey);
-
-// Cache storage  
-this.lruCache.setParsedFile(cacheKey, parseResult); // STORAGE POINT 12
-```
-
-### 4.2 LocalStorage Usage
-
-**Settings Storage** (`src/config/feature-flags.ts`, Lines 104-110):
-```typescript
-private saveFlags(): void {
-  try {
-    localStorage.setItem('gitnexus_feature_flags', JSON.stringify(this.flags));
-    // STORAGE POINT 13 - Browser localStorage
-  } catch (error) {
-    console.warn('Failed to save feature flags to localStorage:', error);
-  }
-}
-```
-
-**Chat History** (`src/lib/chat-history.ts`, Lines 275-291):
-```typescript
-private saveSession(session: ChatSession): void {
-  try {
-    localStorage.setItem(this.storageKey, JSON.stringify(session));
-    // STORAGE POINT 14 - Chat persistence
-  } catch (error) {
-    // Handle quota exceeded errors
-  }
-}
-```
-
-### 4.3 IndexedDB (Planned for KuzuDB)
-
-**Current Status**: Implementation exists but not fully integrated
-**Files**: `src/core/kuzu/` directory contains KuzuDB integration code
-**Purpose**: Will replace JSON storage with embedded graph database
-
----
-
-## 5. Data Flow Summary
-
-```mermaid
-flowchart TD
-    START([Repository Input]) --> STRUCT[Structure Processor]
-    
-    STRUCT --> |Creates nodes/relationships| GRAPH1[In-Memory Graph]
-    
-    GRAPH1 --> PARSE[Parsing Processor]
-    PARSE --> |AST Storage| AST_MAP[AST Map]
-    PARSE --> |Function Registry| FUNC_TRIE[Function Trie]
-    PARSE --> |Adds definition nodes| GRAPH2[Enhanced Graph]
-    
-    GRAPH2 --> IMPORT[Import Processor]
-    AST_MAP --> IMPORT
-    IMPORT --> |Import Map| IMP_MAP[Import Map]
-    IMPORT --> |Adds IMPORTS relationships| GRAPH3[Graph + Imports]
-    
-    GRAPH3 --> CALLS[Call Processor]
-    AST_MAP --> CALLS
-    IMP_MAP --> CALLS
-    FUNC_TRIE --> CALLS
-    CALLS --> |Adds CALLS relationships| FINAL_GRAPH[Final Knowledge Graph]
-    
-    FINAL_GRAPH --> JSON_EXPORT[JSON Export]
-    JSON_EXPORT --> |JSON.stringify| JSON_STRING[JSON String]
-    JSON_STRING --> |Browser Download| FILE_SYSTEM[File System]
-    
-    FINAL_GRAPH --> |Direct object reference| UI[UI Components]
-    
-    subgraph "Storage Points"
-        GRAPH1
-        GRAPH2  
-        GRAPH3
-        FINAL_GRAPH
-        AST_MAP
-        FUNC_TRIE
-        IMP_MAP
-        JSON_STRING
-    end
-    
-    subgraph "Cache Layer"
-        LRU_CACHE[LRU Cache]
-        LOCAL_STORAGE[LocalStorage]
-    end
-    
-    PARSE -.-> LRU_CACHE
-    LOCAL_STORAGE -.-> SETTINGS[Settings/Flags]
-```
-
----
-
-## 6. Critical Migration Points for Kuzu DB
-
-### 6.1 Schema Mapping Requirements
-
-**Current JSON Structure → Kuzu Schema**:
-
-1. **Nodes Table**:
-   ```sql
-   CREATE NODE TABLE IF NOT EXISTS nodes (
-     id STRING PRIMARY KEY,
-     label STRING NOT NULL,
-     properties MAP(STRING, STRING)
-   );
-   ```
-
-2. **Relationships Table**:
-   ```sql
-   CREATE REL TABLE IF NOT EXISTS relationships (
-     FROM nodes TO nodes,
-     id STRING,
-     type STRING NOT NULL, 
-     properties MAP(STRING, STRING)
-   );
-   ```
-
-### 6.2 Data Transformation Points
-
-**Every `graph.addNode()` call** → **Kuzu INSERT statement**
-**Every `graph.addRelationship()` call** → **Kuzu MATCH/CREATE statement**
-
-### 6.3 Query Transformation Requirements
-
-**Current**: Linear array searches in `SimpleKnowledgeGraph`
-**Target**: Cypher queries in KuzuDB
-
-**Example Transformations**:
-```typescript
-// Current: Find nodes by label
-graph.nodes.filter(n => n.label === 'Function')
-
-// Target: Cypher query
-MATCH (n:Function) RETURN n
-```
-
-### 6.4 Persistence Layer Changes
-
-**Current Flow**:
-1. Build `SimpleKnowledgeGraph` in memory
-2. Export to JSON string
-3. Download as file
-
-**Target Flow**:
-1. Stream data directly to KuzuDB during processing
-2. Persist to IndexedDB automatically
-3. Export via Cypher queries
-
----
-
-## 7. Implementation Recommendations
-
-### 7.1 Migration Strategy
-
-1. **Phase 1**: Create parallel KuzuDB storage alongside existing JSON
-2. **Phase 2**: Implement streaming ingestion (write to Kuzu during pipeline)
-3. **Phase 3**: Replace SimpleKnowledgeGraph with KuzuDB queries
-4. **Phase 4**: Remove JSON export/import (keep as backup option)
-
-### 7.2 Critical Considerations
-
-1. **Referential Integrity**: Kuzu enforces relationships, JSON doesn't
-2. **Transaction Boundaries**: Kuzu needs explicit transactions
-3. **Query Performance**: Index strategy for common access patterns
-4. **Memory Management**: Kuzu handles memory, current system uses manual arrays
-5. **Concurrent Access**: Kuzu supports concurrent reads, current system is single-threaded
-
-### 7.3 Testing Strategy
-
-1. **Data Integrity**: Compare JSON export with Kuzu export for identical results
-2. **Performance**: Benchmark ingestion and query performance
-3. **Memory Usage**: Monitor memory consumption during large repository processing
-4. **Error Handling**: Test transaction rollback and recovery scenarios
-
----
-
-## Conclusion
-
-This document provides the complete technical foundation for migrating GitNexus from JSON-based storage to KuzuDB. Every storage point, data transformation, and persistence mechanism has been identified and documented. The migration should focus on replacing the `SimpleKnowledgeGraph` implementation while maintaining the exact same data structures and relationships in the new KuzuDB schema.

PHASE_2_IMPLEMENTATION_SUMMARY.md

@@ -1,253 +0,0 @@
-# Phase 2: Parallel Storage Implementation - Complete! 🎉
-
-## Overview
-
-Successfully implemented the **Parallel Storage** phase of the KuzuDB migration plan, enabling dual-write functionality where data is written to both JSON (primary) and KuzuDB (secondary) storage systems simultaneously.
-
-## ✅ **Components Implemented**
-
-### 1. **KuzuProcessorBase** (`src/core/ingestion/kuzu-processor-base.ts`)
-
-**Core Features:**
-- **Dual-Write Pattern**: Seamless writes to both JSON and KuzuDB
-- **Transaction Management**: Begin, commit, and rollback transaction support
-- **Error Handling**: Graceful degradation when KuzuDB fails
-- **Performance Monitoring**: Detailed statistics and timing metrics
-- **Data Validation**: Consistency checks between storage systems
-- **Feature Flag Integration**: Respects `isKuzuDBEnabled()` settings
-
-**Key Methods:**
-- `addNodeDualWrite()` - Writes nodes to both storages
-- `addRelationshipDualWrite()` - Writes relationships to both storages
-- `beginTransaction()` / `commitTransaction()` / `rollbackTransaction()`
-- `initializeKuzuDB()` - Sets up KuzuDB connection
-- `validateNodeConsistency()` / `validateRelationshipConsistency()`
-
-### 2. **Enhanced StructureProcessor** 
-
-**Modifications:**
-- ✅ Extends `KuzuProcessorBase` for dual-write capability
-- ✅ Async `process()` method with KuzuDB initialization
-- ✅ Dual-write support for Project, Folder, and File nodes
-- ✅ Dual-write support for CONTAINS relationships
-- ✅ Transaction boundaries with commit/rollback
-- ✅ Comprehensive error handling and statistics
-
-**Dual-Write Flow:**
-1. Initialize KuzuDB connection
-2. Create project node → write to JSON + KuzuDB
-3. Create directory nodes → write to JSON + KuzuDB  
-4. Create file nodes → write to JSON + KuzuDB
-5. Create CONTAINS relationships → write to JSON + KuzuDB
-6. Commit KuzuDB transaction
-7. Log detailed statistics
-
-### 3. **Enhanced ParsingProcessor**
-
-**Modifications:**
-- ✅ Extends `KuzuProcessorBase` for dual-write capability
-- ✅ Async definition processing with KuzuDB writes
-- ✅ Dual-write support for Function, Class, Method, Variable, Interface, Type nodes
-- ✅ Dual-write support for INHERITS, IMPLEMENTS, IMPORTS relationships
-- ✅ Transaction boundaries with automatic commit
-- ✅ Batch processing optimization
-
-**Dual-Write Flow:**
-1. Initialize KuzuDB connection
-2. Process each file's definitions
-3. Create definition nodes → write to JSON + KuzuDB
-4. Create containment relationships → write to JSON + KuzuDB
-5. Create inheritance/implementation relationships → write to JSON + KuzuDB
-6. Commit KuzuDB transaction
-7. Log processing statistics
-
-### 4. **Enhanced ImportProcessor**
-
-**Modifications:**
-- ✅ Extends `KuzuProcessorBase` for dual-write capability
-- ✅ Async import relationship creation
-- ✅ Dual-write support for IMPORTS relationships
-- ✅ Transaction management with rollback support
-- ✅ Enhanced error handling and progress tracking
-
-**Dual-Write Flow:**
-1. Initialize KuzuDB connection
-2. Process imports for each file
-3. Create IMPORTS relationships → write to JSON + KuzuDB
-4. Commit KuzuDB transaction
-5. Log import resolution statistics
-
-### 5. **Enhanced CallProcessor**
-
-**Modifications:**
-- ✅ Extends `KuzuProcessorBase` for dual-write capability
-- ✅ Async call relationship creation
-- ✅ Dual-write support for CALLS relationships
-- ✅ 3-stage resolution strategy maintained
-- ✅ Transaction boundaries and error handling
-
-**Dual-Write Flow:**
-1. Initialize KuzuDB connection
-2. Extract function calls from AST
-3. Resolve calls using 3-stage strategy
-4. Create CALLS relationships → write to JSON + KuzuDB
-5. Commit KuzuDB transaction
-6. Log call resolution statistics
-
-## 🏗️ **Architecture Highlights**
-
-### **Dual-Write Pattern Implementation**
-```typescript
-// JSON write (primary - always succeeds)
-jsonGraph.addNode(node);
-
-// KuzuDB write (secondary - graceful failure)
-if (this.kuzuGraph) {
-  try {
-    this.kuzuGraph.addNode(node);
-  } catch (kuzuError) {
-    console.warn('KuzuDB write failed:', kuzuError);
-    // Continue processing - JSON is primary storage
-  }
-}
-```
-
-### **Transaction Management**
-```typescript
-// Begin transaction
-await this.beginTransaction();
-
-try {
-  // Perform operations
-  await this.addNodeDualWrite(graph, node);
-  await this.addRelationshipDualWrite(graph, relationship);
-  
-  // Commit transaction
-  await this.commitTransaction();
-} catch (error) {
-  // Rollback on failure
-  await this.rollbackTransaction();
-  throw error;
-}
-```
-
-### **Statistics and Monitoring**
-- **Nodes processed**: Total nodes written to JSON
-- **KuzuDB nodes written**: Successful KuzuDB writes
-- **KuzuDB errors**: Failed KuzuDB operations
-- **Success rate**: Percentage of successful dual-writes
-- **Processing time**: Total time spent on operations
-- **Validation errors**: Data consistency issues detected
-
-## 📊 **Key Benefits Achieved**
-
-### **1. Zero Breaking Changes**
-- All existing processors maintain their original interfaces
-- JSON storage remains primary - system continues working even if KuzuDB fails
-- Backward compatibility with all existing code
-
-### **2. Production-Ready Error Handling**
-- KuzuDB failures don't break the ingestion pipeline
-- Graceful degradation to JSON-only mode
-- Comprehensive error logging and categorization
-- Transaction rollback on critical failures
-
-### **3. Performance Optimization**
-- Batch processing for optimal KuzuDB performance
-- Async operations with proper error boundaries
-- Transaction boundaries reduce database overhead
-- Detailed performance monitoring and statistics
-
-### **4. Data Consistency**
-- Dual-write ensures both storages have the same data
-- Transaction management prevents partial writes
-- Validation hooks for consistency checking
-- Rollback capabilities for data integrity
-
-### **5. Feature Flag Integration**
-- Respects `isKuzuDBEnabled()` configuration
-- Can be enabled/disabled without code changes
-- Gradual rollout capabilities
-- A/B testing support
-
-## 🔄 **Integration Points**
-
-### **Pipeline Integration**
-All processors now support the enhanced dual-write pattern:
-
-```typescript
-// Structure Phase
-const structureProcessor = new StructureProcessor({ enableKuzuDB: true });
-await structureProcessor.process(graph, structureInput);
-
-// Parsing Phase  
-const parsingProcessor = new ParsingProcessor({ enableKuzuDB: true });
-await parsingProcessor.process(graph, parsingInput);
-
-// Import Phase
-const importProcessor = new ImportProcessor({ enableKuzuDB: true });
-await importProcessor.process(graph, astMap, fileContents);
-
-// Call Phase
-const callProcessor = new CallProcessor(functionTrie, { enableKuzuDB: true });
-await callProcessor.process(graph, astMap, importMap);
-```
-
-### **Configuration Options**
-```typescript
-interface KuzuProcessorOptions {
-  enableKuzuDB?: boolean;        // Enable/disable KuzuDB integration
-  batchSize?: number;            // Batch size for optimal performance
-  autoCommit?: boolean;          // Automatic transaction commits
-  enableValidation?: boolean;    // Data consistency validation
-}
-```
-
-## 📈 **Performance Expectations**
-
-### **Memory Usage**
-- Minimal additional memory overhead (~5-10%)
-- Transaction batching prevents memory bloat
-- Graceful handling of large codebases
-
-### **Processing Time**
-- Expected 10-20% increase in processing time
-- Batch operations optimize KuzuDB performance
-- Async operations prevent blocking
-
-### **Error Resilience**
-- 100% reliability for JSON storage (primary)
-- Graceful degradation for KuzuDB failures
-- No data loss even with KuzuDB issues
-
-## 🚀 **Ready for Phase 3**
-
-The parallel storage implementation provides a solid foundation for **Phase 3: Query Migration**, where we'll:
-
-1. **Implement Query Abstraction Layer**: Create unified query interface
-2. **Add Query Routing Logic**: Route queries to appropriate storage
-3. **Performance Comparison Tools**: A/B test JSON vs KuzuDB queries
-4. **Query Result Validation**: Ensure consistent results between storages
-
-## 📁 **Files Modified/Created**
-
-### **New Files**
-- `src/core/ingestion/kuzu-processor-base.ts` - Base class for dual-write pattern
-
-### **Modified Files**
-- `src/core/ingestion/structure-processor.ts` - Added KuzuDB dual-write support
-- `src/core/ingestion/parsing-processor.ts` - Added KuzuDB dual-write support  
-- `src/core/ingestion/import-processor.ts` - Added KuzuDB dual-write support
-- `src/core/ingestion/call-processor.ts` - Added KuzuDB dual-write support
-
-## 🎯 **Success Metrics**
-
-- ✅ **100% Backward Compatibility**: All existing functionality preserved
-- ✅ **Graceful Error Handling**: KuzuDB failures don't break the system
-- ✅ **Transaction Safety**: Data integrity maintained with rollback support
-- ✅ **Performance Monitoring**: Comprehensive statistics and metrics
-- ✅ **Feature Flag Ready**: Can be enabled/disabled via configuration
-- ✅ **Production Quality**: Error handling, logging, and monitoring
-
-The dual-write pattern is now fully implemented and ready for production deployment! 🚀
-

README.md

@@ -1,359 +1,589 @@
-# GitNexus - Fully Client sided Knowledge Graph Generator and Graph RAG Agent
+# GitNexus
+⚠️ Important Notice:** GitNexus has NO official cryptocurrency, token, or coin. Any token/coin using the GitNexus name on Pump.fun or any other platform is **not affiliated with, endorsed by, or created by** this project or its maintainers. Do not purchase any cryptocurrency claiming association with GitNexus.
+
+<div align="center">
+
+  <a href="https://trendshift.io/repositories/19809" target="_blank">
+    <img src="https://trendshift.io/api/badge/repositories/19809" alt="abhigyanpatwari%2FGitNexus | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/>
+  </a>
+
+  <h2>Join the official Discord to discuss ideas, issues etc!</h2>
+
+  <a href="https://discord.gg/AAsRVT6fGb">
+    <img src="https://img.shields.io/discord/1477255801545429032?color=5865F2&logo=discord&logoColor=white" alt="Discord"/>
+  </a>
+  <a href="https://www.npmjs.com/package/gitnexus">
+    <img src="https://img.shields.io/npm/v/gitnexus.svg" alt="npm version"/>
+  </a>
+  <a href="https://polyformproject.org/licenses/noncommercial/1.0.0/">
+    <img src="https://img.shields.io/badge/License-PolyForm%20Noncommercial-blue.svg" alt="License: PolyForm Noncommercial"/>
+  </a>
+
+  <p><strong>Enterprise (SaaS & Self-hosted)</strong> - <a href="https://akonlabs.com">akonlabs.com</a></p>
+
+</div>
+
+**Building nervous system for agent context.**
+
+Indexes any codebase into a knowledge graph — every dependency, call chain, cluster, and execution flow — then exposes it through smart tools so AI agents never miss code.
 
 
-GitNexus is a privacy-focused, zero-server knowledge graph generator that runs entirely in your browser. It transforms codebases into interactive knowledge graphs using advanced AST parsing, multi-threaded Web Workers, and an embedded KuzuDB WASM database. Features a Graph RAG agent for intelligent code exploration through natural language queries using cypher queries executed directly against the in-browser graph database.
-
-https://github.com/user-attachments/assets/6f13bd45-d6e9-4f4e-a360-ceb66f41c741
-
-## Current Work in Progress:
-
-- Ollama support
-- Export as csv ( for both node and relation table )
 
 
-## Features
+https://github.com/user-attachments/assets/172685ba-8e54-4ea7-9ad1-e31a3398da72
 
-**Code Analysis**
 
-- Analyze GitHub repositories or ZIP files
-- Support for TypeScript, JavaScript, Python
-- Interactive graph visualization with D3.js
-- File filtering and directory selection
-- Export results as JSON/CSV
 
-**AI Chat**
+> *Like DeepWiki, but deeper.* DeepWiki helps you *understand* code. GitNexus lets you *analyze* it — because a knowledge graph tracks every relationship, not just descriptions.
 
-- Multiple LLM providers (OpenAI, Anthropic, Gemini, Azure)
-- Query code structure and relationships
-- Context-aware conversations
-- Graph-based code search
+**TL;DR:** The **Web UI** is a quick way to chat with any repo. The **CLI + MCP** is how you make your AI agent actually reliable — it gives Cursor, Claude Code, Codex, and friends a deep architectural view of your codebase so they stop missing dependencies, breaking call chains, and shipping blind edits. Even smaller models get full architectural clarity, making it compete with goliath models.
 
-**Processing**
+---
 
-- Four-pass analysis: structure → parsing → imports → calls
-- Parallel processing with Web Workers
-- AST-based code extraction using Tree-sitter
-- Memory-efficient caching
+## Star History
 
-## Architecture
+[![Star History Chart](https://api.star-history.com/svg?repos=abhigyanpatwari/GitNexus&type=date&legend=top-left)](https://www.star-history.com/#abhigyanpatwari/GitNexus&type=date&legend=top-left)
 
-```mermaid
-graph TB
-    UI[React UI Layer] --> EM[Engine Manager]
-    EM --> LEG[Legacy Engine]
-    EM --> NG[Next-Gen Engine - WIP]
-  
-    subgraph "Legacy Engine (Production Ready)"
-        LEG --> GP[Sequential Pipeline]
-        GP --> SP[Single-threaded Parser]
-        GP --> MEM[In-Memory Graph Store]
-        MEM --> JSON[JSON Export]
-    end
-  
-    subgraph "Next-Gen Engine (Work in Progress)"
-        NG --> PP[Parallel Pipeline]
-        PP --> WP[Web Worker Pool]
-        PP --> KDB[KuzuDB WASM]
-        KDB --> CYP[Cypher Queries]
-        CYP --> RAG[Graph RAG Agent - WIP]
-    end
-  
-    subgraph "Core Technologies"
-        TS[Tree-sitter WASM]
-        D3[D3.js Force Simulation]
-        LC[LangChain ReAct Agents]
-        IDB[IndexedDB Persistence]
-    end
+
+## Two Ways to Use GitNexus
+
+|                   | **CLI + MCP**                                            | **Web UI**                                             |
+| ----------------- | -------------------------------------------------------------- | ------------------------------------------------------------ |
+| **What**    | Index repos locally, connect AI agents via MCP                 | Visual graph explorer + AI chat in browser                   |
+| **For**     | Daily development with Cursor, Claude Code, Codex, Windsurf, OpenCode | Quick exploration, demos, one-off analysis                   |
+| **Scale**   | Full repos, any size                                           | Limited by browser memory (~5k files), or unlimited via backend mode |
+| **Install** | `npm install -g gitnexus`                                    | No install —[gitnexus.vercel.app](https://gitnexus.vercel.app) |
+| **Storage** | LadybugDB native (fast, persistent)                               | LadybugDB WASM (in-memory, per session)                         |
+| **Parsing** | Tree-sitter native bindings                                    | Tree-sitter WASM                                             |
+| **Privacy** | Everything local, no network                                   | Everything in-browser, no server                             |
+
+> **Bridge mode:** `gitnexus serve` connects the two — the web UI auto-detects the local server and can browse all your CLI-indexed repos without re-uploading or re-indexing.
+
+---
+
+## Enterprise
+
+GitNexus is available as an **enterprise offering** - either as a fully managed **SaaS** or a **self-hosted** deployment. Also available for **commercial use** of the OSS version with proper licensing.
+
+Enterprise includes:
+- **PR Review** - automated blast radius analysis on pull requests
+- **Auto-updating Code Wiki** - always up-to-date documentation (Code Wiki is also available in OSS)
+- **Auto-reindexing** - knowledge graph stays fresh automatically
+- **Multi-repo support** - unified graph across repositories
+- **OCaml support** - additional language coverage
+- **Priority feature/language support** - request new languages or features
+
+**Upcoming:**
+- Auto regression forensics
+- End-to-end test generation
+
+👉 Learn more at [akonlabs.com](https://akonlabs.com)
+
+💬 For commercial licensing or enterprise inquiries, ping us on [Discord](https://discord.gg/AAsRVT6fGb) or drop an email at founders@akonlabs.com
+
+---
+
+## Development
+
+- [ARCHITECTURE.md](ARCHITECTURE.md) — packages, index → graph → MCP flow, where to change code
+- [RUNBOOK.md](RUNBOOK.md) — analyze, embeddings, stale index, MCP recovery, CI snippets
+- [GUARDRAILS.md](GUARDRAILS.md) — safety rules and operational “Signs” for contributors and agents
+- [CONTRIBUTING.md](CONTRIBUTING.md) — license, setup, commits, and pull requests
+- [TESTING.md](TESTING.md) — test commands for `gitnexus` and `gitnexus-web`
+
+## CLI + MCP (recommended)
+
+The CLI indexes your repository and runs an MCP server that gives AI agents deep codebase awareness.
+
+### Quick Start
+
+```bash
+# Index your repo (run from repo root)
+npx gitnexus analyze
 ```
 
-**Tech Stack**:
+That's it. This indexes the codebase, installs agent skills, registers Claude Code hooks, and creates `AGENTS.md` / `CLAUDE.md` context files — all in one command.
 
-- **Frontend**: React 18 + TypeScript + Vite + D3.js force simulation
-- **Parsing**: Tree-sitter WASM parsers (TypeScript, JavaScript, Python)
-- **Concurrency**: Web Worker Pool with Comlink for thread-safe communication
-- **Caching**: LRU-based AST cache with memory management and eviction policies
-- **AI**: LangChain.js ReAct agents with tool-augmented reasoning
-- **Database**: KuzuDB WASM integration (WIP) + IndexedDB persistence
-- **Graph RAG**: Cypher query generation for knowledge graph reasoning (WIP)
+To configure MCP for your editor, run `npx gitnexus setup` once — or set it up manually below.
 
-## Four-Pass Ingestion Pipeline
+### MCP Setup
+
+`gitnexus setup` auto-detects your editors and writes the correct global MCP config. You only need to run it once.
+
+### Editor Support
+
+| Editor                | MCP | Skills | Hooks (auto-augment) | Support        |
+| --------------------- | --- | ------ | -------------------- | -------------- |
+| **Claude Code** | Yes | Yes    | Yes (PreToolUse + PostToolUse) | **Full** |
+| **Cursor**      | Yes | Yes    | —                   | MCP + Skills   |
+| **Codex**       | Yes | Yes    | —                   | MCP + Skills   |
+| **Windsurf**    | Yes | —     | —                   | MCP            |
+| **OpenCode**    | Yes | Yes    | —                   | MCP + Skills   |
+| **Codex**       | Yes | —     | —                   | MCP            |
+
+> **Claude Code** gets the deepest integration: MCP tools + agent skills + PreToolUse hooks that enrich searches with graph context + PostToolUse hooks that auto-reindex after commits.
+
+## Community Integrations
+
+Built by the community — not officially maintained, but worth checking out.
+
+| Project | Author | Description |
+|---------|--------|-------------|
+| [pi-gitnexus](https://github.com/tintinweb/pi-gitnexus) | [@tintinweb](https://github.com/tintinweb) | GitNexus plugin for [pi](https://pi.dev) — `pi install npm:pi-gitnexus` |
+| [gitnexus-stable-ops](https://github.com/ShunsukeHayashi/gitnexus-stable-ops) | [@ShunsukeHayashi](https://github.com/ShunsukeHayashi) | Stable ops & deployment workflows (Miyabi ecosystem) |
+
+> Have a project built on GitNexus? Open a PR to add it here!
+
+If you prefer manual configuration:
+
+**Claude Code** (full support — MCP + skills + hooks):
+
+```bash
+# macOS / Linux
+claude mcp add gitnexus -- npx -y gitnexus@latest mcp
+
+# Windows
+claude mcp add gitnexus -- cmd /c npx -y gitnexus@latest mcp
+```
+
+**Codex** (full support — MCP + skills):
+
+```bash
+codex mcp add gitnexus -- npx -y gitnexus@latest mcp
+```
+
+**Cursor** (`~/.cursor/mcp.json` — global, works for all projects):
+
+```json
+{
+  "mcpServers": {
+    "gitnexus": {
+      "command": "npx",
+      "args": ["-y", "gitnexus@latest", "mcp"]
+    }
+  }
+}
+```
+
+**OpenCode** (`~/.config/opencode/config.json`):
+
+```json
+{
+  "mcp": {
+    "gitnexus": {
+      "type": "local",
+      "command": ["gitnexus", "mcp"]
+    }
+  }
+}
+```
+
+**Codex** (`~/.codex/config.toml` for system scope, or `.codex/config.toml` for project scope):
+
+```toml
+[mcp_servers.gitnexus]
+command = "npx"
+args = ["-y", "gitnexus@latest", "mcp"]
+```
+
+### CLI Commands
+
+```bash
+gitnexus setup                   # Configure MCP for your editors (one-time)
+gitnexus analyze [path]          # Index a repository (or update stale index)
+gitnexus analyze --force         # Force full re-index
+gitnexus analyze --skills        # Generate repo-specific skill files from detected communities
+gitnexus analyze --skip-embeddings  # Skip embedding generation (faster)
+gitnexus analyze --skip-agents-md  # Preserve custom AGENTS.md/CLAUDE.md gitnexus section edits
+gitnexus analyze --embeddings    # Enable embedding generation (slower, better search)
+gitnexus analyze --verbose       # Log skipped files when parsers are unavailable
+gitnexus mcp                     # Start MCP server (stdio) — serves all indexed repos
+gitnexus serve                   # Start local HTTP server (multi-repo) for web UI connection
+gitnexus list                    # List all indexed repositories
+gitnexus status                  # Show index status for current repo
+gitnexus clean                   # Delete index for current repo
+gitnexus clean --all --force     # Delete all indexes
+gitnexus wiki [path]             # Generate repository wiki from knowledge graph
+gitnexus wiki --model <model>    # Wiki with custom LLM model (default: gpt-4o-mini)
+gitnexus wiki --base-url <url>   # Wiki with custom LLM API base URL
+```
+
+### What Your AI Agent Gets
+
+**7 tools** exposed via MCP:
+
+| Tool               | What It Does                                                      | `repo` Param |
+| ------------------ | ----------------------------------------------------------------- | -------------- |
+| `list_repos`     | Discover all indexed repositories                                 | —             |
+| `query`          | Process-grouped hybrid search (BM25 + semantic + RRF)             | Optional       |
+| `context`        | 360-degree symbol view — categorized refs, process participation | Optional       |
+| `impact`         | Blast radius analysis with depth grouping and confidence          | Optional       |
+| `detect_changes` | Git-diff impact — maps changed lines to affected processes       | Optional       |
+| `rename`         | Multi-file coordinated rename with graph + text search            | Optional       |
+| `cypher`         | Raw Cypher graph queries                                          | Optional       |
+
+> When only one repo is indexed, the `repo` parameter is optional. With multiple repos, specify which one: `query({query: "auth", repo: "my-app"})`.
+
+**Resources** for instant context:
+
+| Resource                                  | Purpose                                              |
+| ----------------------------------------- | ---------------------------------------------------- |
+| `gitnexus://repos`                      | List all indexed repositories (read this first)      |
+| `gitnexus://repo/{name}/context`        | Codebase stats, staleness check, and available tools |
+| `gitnexus://repo/{name}/clusters`       | All functional clusters with cohesion scores         |
+| `gitnexus://repo/{name}/cluster/{name}` | Cluster members and details                          |
+| `gitnexus://repo/{name}/processes`      | All execution flows                                  |
+| `gitnexus://repo/{name}/process/{name}` | Full process trace with steps                        |
+| `gitnexus://repo/{name}/schema`         | Graph schema for Cypher queries                      |
+
+**2 MCP prompts** for guided workflows:
+
+| Prompt            | What It Does                                                              |
+| ----------------- | ------------------------------------------------------------------------- |
+| `detect_impact` | Pre-commit change analysis — scope, affected processes, risk level       |
+| `generate_map`  | Architecture documentation from the knowledge graph with mermaid diagrams |
+
+**4 agent skills** installed to `.claude/skills/` automatically:
+
+- **Exploring** — Navigate unfamiliar code using the knowledge graph
+- **Debugging** — Trace bugs through call chains
+- **Impact Analysis** — Analyze blast radius before changes
+- **Refactoring** — Plan safe refactors using dependency mapping
+
+**Repo-specific skills** generated with `--skills`:
+
+When you run `gitnexus analyze --skills`, GitNexus detects the functional areas of your codebase (via Leiden community detection) and generates a `SKILL.md` file for each one under `.claude/skills/generated/`. Each skill describes a module's key files, entry points, execution flows, and cross-area connections — so your AI agent gets targeted context for the exact area of code you're working in. Skills are regenerated on each `--skills` run to stay current with the codebase.
+
+---
+
+## Multi-Repo MCP Architecture
+
+GitNexus uses a **global registry** so one MCP server can serve multiple indexed repos. No per-project MCP config needed — set it up once and it works everywhere.
 
 ```mermaid
 flowchart TD
-    START([Repository Input]) --> PASS1
-  
-    subgraph PASS1 ["Pass 1: Structure Analysis"]
-        P1A[Recursive Directory Traversal] --> P1B[File Type Classification]
-        P1B --> P1C[Project/Folder/File Nodes]
-        P1C --> P1D[CONTAINS Relationships]
+    subgraph CLI [CLI Commands]
+        Setup["gitnexus setup"]
+        Analyze["gitnexus analyze"]
+        Clean["gitnexus clean"]
+        List["gitnexus list"]
     end
-  
-    subgraph PASS2 ["Pass 2: Code Parsing & AST"]
-        P2A[Tree-sitter WASM Init] --> P2B[Grammar Loading]
-        P2B --> P2C[AST Generation]
-        P2C --> P2D[Symbol Extraction]
-        P2D --> P2E[LRU Cache Storage]
+
+    subgraph Registry ["~/.gitnexus/"]
+        RegFile["registry.json"]
     end
-  
-    subgraph PASS3 ["Pass 3: Import Resolution"]
-        P3A[Import Statement Extraction] --> P3B[Module Path Resolution]
-        P3B --> P3C[Cross-Reference Tables]
-        P3C --> P3D[IMPORTS Relationships]
+
+    subgraph Repos [Project Repos]
+        RepoA[".gitnexus/ in repo A"]
+        RepoB[".gitnexus/ in repo B"]
     end
-  
-    subgraph PASS4 ["Pass 4: Call Graph Analysis"]
-        P4A[Function Call Pattern Matching] --> P4B[Exact Match via Import Map]
-        P4B --> P4C[Fuzzy Match + Levenshtein]
-        P4C --> P4D[CALLS Relationships]
+
+    subgraph MCP [MCP Server]
+        Server["server.ts"]
+        Backend["LocalBackend"]
+        Pool["Connection Pool"]
+        ConnA["LadybugDB conn A"]
+        ConnB["LadybugDB conn B"]
     end
-  
-    PASS1 --> PASS2
-    PASS2 --> PASS3
-    PASS3 --> PASS4
-    PASS4 --> END([Knowledge Graph])
-  
-    classDef passBox fill:#e1f5fe,stroke:#01579b,stroke-width:2px,color:#000
-    classDef startEnd fill:#c8e6c9,stroke:#2e7d32,stroke-width:3px,color:#000
-    classDef step fill:#fff3e0,stroke:#ef6c00,stroke-width:1px,color:#000
-  
-    class PASS1,PASS2,PASS3,PASS4 passBox
-    class START,END startEnd
-    class P1A,P1B,P1C,P1D,P2A,P2B,P2C,P2D,P2E,P3A,P3B,P3C,P3D,P4A,P4B,P4C,P4D step
+
+    Setup -->|"writes global MCP config"| CursorConfig["~/.cursor/mcp.json"]
+    Analyze -->|"registers repo"| RegFile
+    Analyze -->|"stores index"| RepoA
+    Clean -->|"unregisters repo"| RegFile
+    List -->|"reads"| RegFile
+    Server -->|"reads registry"| RegFile
+    Server --> Backend
+    Backend --> Pool
+    Pool -->|"lazy open"| ConnA
+    Pool -->|"lazy open"| ConnB
+    ConnA -->|"queries"| RepoA
+    ConnB -->|"queries"| RepoB
 ```
 
-### Data Flow & Storage Architecture
+**How it works:** Each `gitnexus analyze` stores the index in `.gitnexus/` inside the repo (portable, gitignored) and registers a pointer in `~/.gitnexus/registry.json`. When an AI agent starts, the MCP server reads the registry and can serve any indexed repo. LadybugDB connections are opened lazily on first query and evicted after 5 minutes of inactivity (max 5 concurrent). If only one repo is indexed, the `repo` parameter is optional on all tools — agents don't need to change anything.
 
-```mermaid
-flowchart TD
-    START([Repository Input]) --> STRUCT[Structure Processor]
-    
-    STRUCT --> |Creates nodes/relationships| GRAPH1[In-Memory Graph]
-    
-    GRAPH1 --> PARSE[Parsing Processor]
-    PARSE --> |AST Storage| AST_MAP[AST Map]
-    PARSE --> |Function Registry| FUNC_TRIE[Function Trie]
-    PARSE --> |Adds definition nodes| GRAPH2[Enhanced Graph]
-    
-    GRAPH2 --> IMPORT[Import Processor]
-    AST_MAP --> IMPORT
-    IMPORT --> |Import Map| IMP_MAP[Import Map]
-    IMPORT --> |Adds IMPORTS relationships| GRAPH3[Graph + Imports]
-    
-    GRAPH3 --> CALLS[Call Processor]
-    AST_MAP --> CALLS
-    IMP_MAP --> CALLS
-    FUNC_TRIE --> CALLS
-    CALLS --> |Adds CALLS relationships| FINAL_GRAPH[Final Knowledge Graph]
-    
-    FINAL_GRAPH --> JSON_EXPORT[JSON Export]
-    JSON_EXPORT --> |JSON.stringify| JSON_STRING[JSON String]
-    JSON_STRING --> |Browser Download| FILE_SYSTEM[File System]
-    
-    FINAL_GRAPH --> |Direct object reference| UI[UI Components]
-    
-    subgraph "Storage Points"
-        GRAPH1
-        GRAPH2  
-        GRAPH3
-        FINAL_GRAPH
-        AST_MAP
-        FUNC_TRIE
-        IMP_MAP
-        JSON_STRING
-    end
-    
-    subgraph "Cache Layer"
-        LRU_CACHE[LRU Cache]
-        LOCAL_STORAGE[LocalStorage]
-    end
-    
-    PARSE -.-> LRU_CACHE
-    LOCAL_STORAGE -.-> SETTINGS[Settings/Flags]
-```
+---
 
-### Technical Implementation Details
+## Web UI (browser-based)
 
-**Pass 1: Structure Analysis**
+A fully client-side graph explorer and AI chat. No server, no install — your code never leaves the browser.
 
-- Implements recursive directory traversal with configurable depth limits
-- File type detection using MIME types and extension mapping
-- Creates hierarchical node structure with parent-child relationships
-- Establishes CONTAINS relationships for project organization
+**Try it now:** [gitnexus.vercel.app](https://gitnexus.vercel.app) — drag & drop a ZIP and start exploring.
 
-**Pass 2: Code Parsing & AST Extraction**
+<img width="2550" height="1343" alt="gitnexus_img" src="https://github.com/user-attachments/assets/cc5d637d-e0e5-48e6-93ff-5bcfdb929285" />
 
-- Initializes Tree-sitter WASM parsers with language-specific grammars
-- Generates Abstract Syntax Trees for each source file
-- Implements AST traversal algorithms to extract code symbols
-- **LRU Cache System**: Memory-efficient AST storage with configurable eviction policies
-- **Parallel Processing**: Web Worker Pool distributes parsing across multiple threads
-- **Memory Management**: Automatic cleanup and garbage collection for large codebases
-
-**Pass 3: Import Resolution**
-
-- Extracts import/require statements using AST pattern matching
-- Implements module resolution algorithms (Node.js, ES6, Python)
-- Builds cross-reference tables for dependency mapping
-- Handles relative/absolute path resolution with fallback strategies
-
-**Pass 4: Call Graph Analysis**
-
-- **Stage 1**: Exact function call matching using import resolution data
-- **Stage 2**: Fuzzy matching with Levenshtein distance for unresolved calls
-- **Stage 3**: Heuristic-based matching for dynamic calls and method chaining
-- Creates CALLS relationships with confidence scoring
-
-## Getting Started
-
-**Prerequisites**: Node.js 18+, API keys for AI features
+Or run locally:
 
 ```bash
-   git clone <repository-url>
-   cd gitnexus
-   npm install
-   npm run dev
+git clone https://github.com/abhigyanpatwari/gitnexus.git
+cd gitnexus/gitnexus-shared && npm install && npm run build
+cd ../gitnexus-web && npm install
+npm run dev
 ```
 
-Open http://localhost:5173
+The web UI uses the same indexing pipeline as the CLI but runs entirely in WebAssembly (Tree-sitter WASM, LadybugDB WASM, in-browser embeddings). It's great for quick exploration but limited by browser memory for larger repos.
 
-**Configuration**
+**Local Backend Mode:** Run `gitnexus serve` and open the web UI locally — it auto-detects the server and shows all your indexed repos, with full AI chat support. No need to re-upload or re-index. The agent's tools (Cypher queries, search, code navigation) route through the backend HTTP API automatically.
 
-- GitHub token (optional): Increases rate limit to 5,000/hour
-- AI API keys: OpenAI, Anthropic, Gemini, or Azure OpenAI
-- Performance: Set file limits and directory filters
+---
 
-## Usage
+## The Problem GitNexus Solves
 
-**Analyze Repository**
+Tools like **Cursor**, **Claude Code**, **Codex**, **Cline**, **Roo Code**, and **Windsurf** are powerful — but they don't truly know your codebase structure.
 
-1. Enter GitHub URL or upload ZIP file
-2. Set filters (optional): directories, file patterns, size limits
-3. Click "Analyze" and wait for processing
-4. Explore the interactive graph
+**What happens:**
 
-**AI Chat**
+1. AI edits `UserService.validate()`
+2. Doesn't know 47 functions depend on its return type
+3. **Breaking changes ship**
 
-1. Configure API key in settings
-2. Ask questions about the codebase:
-   - "What functions are in main.py?"
-   - "Show classes that inherit from BaseClass"
-   - "How does authentication work?"
+### Traditional Graph RAG vs GitNexus
 
-**Export Data**
-
-- Click Export button to download graph as JSON/CSV
-
-## Advanced Features & Work in Progress
-
-### Web Worker Pool Architecture
+Traditional approaches give the LLM raw graph edges and hope it explores enough. GitNexus **precomputes structure at index time** — clustering, tracing, scoring — so tools return complete context in one call:
 
 ```mermaid
-graph LR
-    MT[Main Thread] --> WM[Worker Manager]
-    WM --> W1[Worker 1<br/>Tree-sitter Parser]
-    WM --> W2[Worker 2<br/>Tree-sitter Parser]
-    WM --> W3[Worker N<br/>Tree-sitter Parser]
-  
-    W1 --> AST1[AST Cache]
-    W2 --> AST2[AST Cache]
-    W3 --> AST3[AST Cache]
-  
-    AST1 --> LRU[LRU Eviction Policy]
-    AST2 --> LRU
-    AST3 --> LRU
-```
+flowchart TB
+    subgraph Traditional["Traditional Graph RAG"]
+        direction TB
+        U1["User: What depends on UserService?"]
+        U1 --> LLM1["LLM receives raw graph"]
+        LLM1 --> Q1["Query 1: Find callers"]
+        Q1 --> Q2["Query 2: What files?"]
+        Q2 --> Q3["Query 3: Filter tests?"]
+        Q3 --> Q4["Query 4: High-risk?"]
+        Q4 --> OUT1["Answer after 4+ queries"]
+    end
 
-### LRU Cache Implementation
-
-- **Memory-bounded AST storage** with configurable size limits (default: 1000 entries)
-- **Automatic eviction policies** based on access patterns and memory pressure
-- **Thread-safe operations** across Web Worker boundaries using Comlink
-- **Cache hit optimization** for repeated file analysis and import resolution
-- **Garbage collection integration** with browser memory management APIs
-
-### KuzuDB Integration Status (Work in Progress)
-
-```mermaid
-graph TD
-    APP[Application Layer] --> RAG[Graph RAG Agent]
-    RAG --> CYP[Cypher Query Generator]
-    CYP --> KDB[KuzuDB WASM Engine]
-    KDB --> IDB[IndexedDB Persistence]
-  
-    subgraph STATUS ["Current Status"]
-        IMPL[KuzuDB WASM Integration - Complete]
-        PERS[IndexedDB Persistence - Complete]
-        SCHEMA[Graph Schema Definition - Complete]
-        QUERY[Cypher Query Execution - WIP]
-        AGENT[Graph RAG Agent - WIP]
+    subgraph GN["GitNexus Smart Tools"]
+        direction TB
+        U2["User: What depends on UserService?"]
+        U2 --> TOOL["impact UserService upstream"]
+        TOOL --> PRECOMP["Pre-structured response:
+        8 callers, 3 clusters, all 90%+ confidence"]
+        PRECOMP --> OUT2["Complete answer, 1 query"]
     end
-  
-    classDef complete fill:#c8e6c9,stroke:#2e7d32,stroke-width:2px
-    classDef wip fill:#fff3e0,stroke:#f57c00,stroke-width:2px
-    classDef main fill:#e3f2fd,stroke:#1976d2,stroke-width:2px
-  
-    class IMPL,PERS,SCHEMA complete
-    class QUERY,AGENT wip
-    class APP,RAG,CYP,KDB,IDB main
 ```
 
-**Implementation Status**:
+**Core innovation: Precomputed Relational Intelligence**
 
-- ✅ **KuzuDB WASM Engine**: Fully integrated embedded graph database
-- ✅ **Graph Schema**: Node and relationship type definitions implemented
-- ✅ **Data Ingestion**: Knowledge graph storage in KuzuDB format
-- 🚧 **Cypher Query Engine**: Query execution layer under development
-- 🚧 **Graph RAG Agent**: AI agent with graph querying capabilities (blocked by Cypher integration)
+- **Reliability** — LLM can't miss context, it's already in the tool response
+- **Token efficiency** — No 10-query chains to understand one function
+- **Model democratization** — Smaller LLMs work because tools do the heavy lifting
 
-**Current Limitation**: The Graph RAG agent cannot execute sophisticated graph queries because the Cypher query execution layer is still being implemented. Basic AI chat works with in-memory graph traversal, but advanced graph reasoning requires the KuzuDB Cypher integration to be completed.
+---
 
-### Dual-Engine Architecture
+## How It Works
 
-- **Legacy Engine**: Production-ready single-threaded processing with JSON storage
-- **Next-Gen Engine**: Parallel processing with KuzuDB persistence (4-8x performance improvement)
-- **Automatic Fallback**: System gracefully degrades to legacy engine if next-gen fails
-- **Runtime Switching**: Users can toggle between engines without data loss
+GitNexus builds a complete knowledge graph of your codebase through a multi-phase indexing pipeline:
 
-## Deployment
+1. **Structure** — Walks the file tree and maps folder/file relationships
+2. **Parsing** — Extracts functions, classes, methods, and interfaces using Tree-sitter ASTs
+3. **Resolution** — Resolves imports, function calls, heritage, constructor inference, and `self`/`this` receiver types across files with language-aware logic
+4. **Clustering** — Groups related symbols into functional communities
+5. **Processes** — Traces execution flows from entry points through call chains
+6. **Search** — Builds hybrid search indexes for fast retrieval
+
+### Supported Languages
+
+| Language | Imports | Named Bindings | Exports | Heritage | Type Annotations | Constructor Inference | Config | Frameworks | Entry Points |
+|----------|---------|----------------|---------|----------|-----------------|---------------------|--------|------------|-------------|
+| TypeScript | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
+| JavaScript | ✓ | ✓ | ✓ | ✓ | — | ✓ | ✓ | ✓ | ✓ |
+| Python | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
+| Java | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | — | ✓ | ✓ |
+| Kotlin | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | — | ✓ | ✓ |
+| C# | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
+| Go | ✓ | — | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
+| Rust | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | — | ✓ | ✓ |
+| PHP | ✓ | ✓ | ✓ | — | ✓ | ✓ | ✓ | ✓ | ✓ |
+| Ruby | ✓ | — | ✓ | ✓ | — | ✓ | — | ✓ | ✓ |
+| Swift | — | — | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
+| C | — | — | ✓ | — | ✓ | ✓ | — | ✓ | ✓ |
+| C++ | — | — | ✓ | ✓ | ✓ | ✓ | — | ✓ | ✓ |
+| Dart | ✓ | — | ✓ | ✓ | ✓ | ✓ | — | ✓ | ✓ |
+
+**Imports** — cross-file import resolution · **Named Bindings** — `import { X as Y }` / re-export tracking · **Exports** — public/exported symbol detection · **Heritage** — class inheritance, interfaces, mixins · **Type Annotations** — explicit type extraction for receiver resolution · **Constructor Inference** — infer receiver type from constructor calls (`self`/`this` resolution included for all languages) · **Config** — language toolchain config parsing (tsconfig, go.mod, etc.) · **Frameworks** — AST-based framework pattern detection · **Entry Points** — entry point scoring heuristics
+
+---
+
+## Tool Examples
+
+### Impact Analysis
+
+```
+impact({target: "UserService", direction: "upstream", minConfidence: 0.8})
+
+TARGET: Class UserService (src/services/user.ts)
+
+UPSTREAM (what depends on this):
+  Depth 1 (WILL BREAK):
+    handleLogin [CALLS 90%] -> src/api/auth.ts:45
+    handleRegister [CALLS 90%] -> src/api/auth.ts:78
+    UserController [CALLS 85%] -> src/controllers/user.ts:12
+  Depth 2 (LIKELY AFFECTED):
+    authRouter [IMPORTS] -> src/routes/auth.ts
+```
+
+Options: `maxDepth`, `minConfidence`, `relationTypes` (`CALLS`, `IMPORTS`, `EXTENDS`, `IMPLEMENTS`), `includeTests`
+
+### Process-Grouped Search
+
+```
+query({query: "authentication middleware"})
+
+processes:
+  - summary: "LoginFlow"
+    priority: 0.042
+    symbol_count: 4
+    process_type: cross_community
+    step_count: 7
+
+process_symbols:
+  - name: validateUser
+    type: Function
+    filePath: src/auth/validate.ts
+    process_id: proc_login
+    step_index: 2
+
+definitions:
+  - name: AuthConfig
+    type: Interface
+    filePath: src/types/auth.ts
+```
+
+### Context (360-degree Symbol View)
+
+```
+context({name: "validateUser"})
+
+symbol:
+  uid: "Function:validateUser"
+  kind: Function
+  filePath: src/auth/validate.ts
+  startLine: 15
+
+incoming:
+  calls: [handleLogin, handleRegister, UserController]
+  imports: [authRouter]
+
+outgoing:
+  calls: [checkPassword, createSession]
+
+processes:
+  - name: LoginFlow (step 2/7)
+  - name: RegistrationFlow (step 3/5)
+```
+
+### Detect Changes (Pre-Commit)
+
+```
+detect_changes({scope: "all"})
+
+summary:
+  changed_count: 12
+  affected_count: 3
+  changed_files: 4
+  risk_level: medium
+
+changed_symbols: [validateUser, AuthService, ...]
+affected_processes: [LoginFlow, RegistrationFlow, ...]
+```
+
+### Rename (Multi-File)
+
+```
+rename({symbol_name: "validateUser", new_name: "verifyUser", dry_run: true})
+
+status: success
+files_affected: 5
+total_edits: 8
+graph_edits: 6     (high confidence)
+text_search_edits: 2  (review carefully)
+changes: [...]
+```
+
+### Cypher Queries
+
+```cypher
+-- Find what calls auth functions with high confidence
+MATCH (c:Community {heuristicLabel: 'Authentication'})<-[:CodeRelation {type: 'MEMBER_OF'}]-(fn)
+MATCH (caller)-[r:CodeRelation {type: 'CALLS'}]->(fn)
+WHERE r.confidence > 0.8
+RETURN caller.name, fn.name, r.confidence
+ORDER BY r.confidence DESC
+```
+
+---
+
+## Wiki Generation
+
+Generate LLM-powered documentation from your knowledge graph:
 
 ```bash
-npm run build
-npm run preview
+# Requires an LLM API key (OPENAI_API_KEY, etc.)
+gitnexus wiki
+
+# Use a custom model or provider
+gitnexus wiki --model gpt-4o
+gitnexus wiki --base-url https://api.anthropic.com/v1
+
+# Force full regeneration
+gitnexus wiki --force
 ```
 
-**Environment Variables**
+The wiki generator reads the indexed graph structure, groups files into modules via LLM, generates per-module documentation pages, and creates an overview page — all with cross-references to the knowledge graph.
 
-```env
-VITE_OPENAI_API_KEY=sk-...
-VITE_DEFAULT_MAX_FILES=500
-VITE_ENABLE_DEBUG_LOGGING=false
-```
+---
+
+## Tech Stack
+
+| Layer                     | CLI                                   | Web                                     |
+| ------------------------- | ------------------------------------- | --------------------------------------- |
+| **Runtime**         | Node.js (native)                      | Browser (WASM)                          |
+| **Parsing**         | Tree-sitter native bindings           | Tree-sitter WASM                        |
+| **Database**        | LadybugDB native                         | LadybugDB WASM                             |
+| **Embeddings**      | HuggingFace transformers.js (GPU/CPU) | transformers.js (WebGPU/WASM)           |
+| **Search**          | BM25 + semantic + RRF                 | BM25 + semantic + RRF                   |
+| **Agent Interface** | MCP (stdio)                           | LangChain ReAct agent                   |
+| **Visualization**   | —                                    | Sigma.js + Graphology (WebGL)           |
+| **Frontend**        | —                                    | React 18, TypeScript, Vite, Tailwind v4 |
+| **Clustering**      | Graphology                            | Graphology                              |
+| **Concurrency**     | Worker threads + async                | Web Workers + Comlink                   |
+
+---
+
+## Roadmap
+
+### Actively Building
+
+- [ ] **LLM Cluster Enrichment** — Semantic cluster names via LLM API
+- [ ] **AST Decorator Detection** — Parse @Controller, @Get, etc.
+- [ ] **Incremental Indexing** — Only re-index changed files
+
+### Recently Completed
+
+- [X] Constructor-Inferred Type Resolution, `self`/`this` Receiver Mapping
+- [X] Wiki Generation, Multi-File Rename, Git-Diff Impact Analysis
+- [X] Process-Grouped Search, 360-Degree Context, Claude Code Hooks
+- [X] Multi-Repo MCP, Zero-Config Setup, 14 Language Support
+- [X] Community Detection, Process Detection, Confidence Scoring
+- [X] Hybrid Search, Vector Index
+
+---
 
 ## Security & Privacy
 
-- All processing happens in your browser
-- API keys stored locally, never transmitted
-- No code or results stored remotely
-- Uses GitHub public API only
+- **CLI**: Everything runs locally on your machine. No network calls. Index stored in `.gitnexus/` (gitignored). Global registry at `~/.gitnexus/` stores only paths and metadata.
+- **Web**: Everything runs in your browser. No code uploaded to any server. API keys stored in localStorage only.
+- Open source — audit the code yourself.
 
-## Contributing
-
-1. Fork the repository
-2. Create feature branch: `git checkout -b feature/name`
-3. Make changes and test
-4. Commit: `git commit -m 'Add feature'`
-5. Push and open Pull Request
-
-**Code Style**: TypeScript strict mode, ESLint rules, minimal comments
-
-## License
-
-MIT License - see [LICENSE](LICENSE) file
+---
 
 ## Acknowledgments
 
-- Tree-sitter for syntax parsing
-- LangChain.js for AI agents
-- D3.js for graph visualization
-- KuzuDB for embedded database
-- [code-graph-rag](https://github.com/vitali87/code-graph-rag) for reference implementation
+- [Tree-sitter](https://tree-sitter.github.io/) — AST parsing
+- [LadybugDB](https://ladybugdb.com/) — Embedded graph database with vector support (formerly KuzuDB)
+- [Sigma.js](https://www.sigmajs.org/) — WebGL graph rendering
+- [transformers.js](https://huggingface.co/docs/transformers.js) — Browser ML
+- [Graphology](https://graphology.github.io/) — Graph data structures
+- [MCP](https://modelcontextprotocol.io/) — Model Context Protocol

RUNBOOK.md

@@ -0,0 +1,163 @@
+# Runbook — GitNexus
+
+Short, copy-paste operations for **local development**, **MCP**, and **CI**. Commands assume a Unix shell; on Windows use Git Bash or equivalent paths.
+
+## Prerequisites
+
+- **Node.js** ≥ 20 (`gitnexus-web/package.json` `engines`).  
+- **Git** (analyze requires a git repository).  
+- From repo root, install and build the CLI package:
+
+```bash
+cd gitnexus
+npm install
+npm run build
+```
+
+Use `npx gitnexus …` from any path after global/published install, or `node dist/cli/index.js …` when developing from `gitnexus/` with a local build.
+
+---
+
+## Index out of date / “stale” tools
+
+**Symptom:** MCP or resources warn the index is behind `HEAD`, or results don’t reflect recent commits.
+
+**Fix (from the target repo root):**
+
+```bash
+npx gitnexus analyze
+```
+
+**Force full rebuild** (same commit but suspect corruption or changed ignore rules):
+
+```bash
+npx gitnexus analyze --force
+```
+
+**Check status:**
+
+```bash
+npx gitnexus status
+```
+
+**List what MCP knows about:**
+
+```bash
+npx gitnexus list
+```
+
+---
+
+## Embeddings
+
+**First time with vectors** (slower, more disk/RAM):
+
+```bash
+npx gitnexus analyze --embeddings
+```
+
+**Important:** If you already had embeddings, **always** pass `--embeddings` on later analyzes, or they can be dropped. See `stats.embeddings` in `.gitnexus/meta.json` (0 means none).
+
+**Large repos:** Analyze may skip or limit embedding work when node counts are very high; watch CLI output.
+
+---
+
+## MCP: no repos / empty tools
+
+**Symptom:** `GitNexus: No indexed repos yet` on stderr when starting MCP.
+
+**Fix:** In each project you want indexed:
+
+```bash
+cd /path/to/repo
+npx gitnexus analyze
+```
+
+Restart the editor MCP session if needed. The server **refreshes the registry lazily**; new analyzes are picked up without necessarily reinstalling MCP.
+
+**Symptom:** Wrong repo when multiple are indexed — pass `repo` on tools or use `list_repos` first.
+
+---
+
+## Clean slate (corrupt or huge `.gitnexus`)
+
+**Current repo only** (prompts for confirmation):
+
+```bash
+npx gitnexus clean
+```
+
+**Skip confirmation:**
+
+```bash
+npx gitnexus clean --force
+```
+
+**All registered repos:**
+
+```bash
+npx gitnexus clean --all --force
+```
+
+Then re-run `npx gitnexus analyze` (and `--embeddings` if you need vectors).
+
+---
+
+## Local bridge for the web UI
+
+```bash
+cd gitnexus
+npx gitnexus serve
+# default http://127.0.0.1:4747 — see serve --help for port/host
+```
+
+Use when the browser UI should talk to **local** indexed repos instead of WASM-only mode.
+
+---
+
+## CLI equivalents of MCP tools
+
+Useful for debugging without an editor:
+
+```bash
+cd gitnexus
+npx gitnexus query "authentication flow" --repo MyRepo
+npx gitnexus context SomeSymbol --repo MyRepo
+npx gitnexus impact SomeSymbol --direction upstream --repo MyRepo
+npx gitnexus cypher "MATCH (n) RETURN count(n) LIMIT 1" --repo MyRepo
+```
+
+---
+
+## CI failures (contributors)
+
+Orchestrator: `.github/workflows/ci.yml`.
+
+| Job | Typical local repro |
+|-----|---------------------|
+| **quality** | `cd gitnexus && npx tsc --noEmit` |
+| **unit-tests** | `cd gitnexus && npx vitest run test/unit` |
+| **integration** | `cd gitnexus && npx vitest run test/integration` (see workflow matrix for groups) |
+| **e2e** | Triggered when `gitnexus-web/` changes; `cd gitnexus-web && E2E=1 npx playwright test` (requires `gitnexus serve` + `npm run dev`) |
+
+**Note:** Pushes that touch only certain markdown paths may be skipped by `paths-ignore` in CI — see workflow file for exact patterns.
+
+---
+
+## Memory / analyze crashes
+
+Analyze re-execs Node with a **large old-space heap** when needed (`analyze.ts`). If you still OOM on huge repos, close other processes, avoid `--embeddings` for a first pass, or analyze a smaller path if supported by your workflow.
+
+---
+
+## LadybugDB / lock errors
+
+Only one process should open a repo’s `.gitnexus/lbug` store at a time. If MCP and a second `analyze` run conflict, stop one process, then retry `analyze` or restart MCP.
+
+---
+
+## Where to dig deeper
+
+- Architecture overview: [ARCHITECTURE.md](ARCHITECTURE.md)  
+- Agent safety rules: [GUARDRAILS.md](GUARDRAILS.md)  
+- Tests: [TESTING.md](TESTING.md)

TESTING.md

@@ -0,0 +1,95 @@
+# Testing — GitNexus
+
+How we structure tests and which commands to run locally and in CI.
+
+## Packages
+
+| Package        | Path           | Runner   | Notes                          |
+| -------------- | -------------- | -------- | ------------------------------ |
+| CLI + MCP core | `gitnexus/`    | Vitest   | Primary test surface in CI     |
+| Web UI         | `gitnexus-web/`| Vitest   | Unit/component tests           |
+| Web UI E2E     | `gitnexus-web/`| Playwright | Run when changing UI flows   |
+
+## Commands (local)
+
+From repository root, unless noted:
+
+**`gitnexus` (CLI / library)**
+
+```bash
+cd gitnexus
+npm install
+npm run build
+npm test                    # unit: vitest run test/unit
+npm run test:integration    # integration suite
+npm run test:all
+npm run test:coverage
+npx tsc --noEmit            # typecheck (matches CI)
+```
+
+**`gitnexus-web`**
+
+```bash
+cd gitnexus-web
+npm install
+npm test                    # unit tests (vitest)
+npx tsc -b --noEmit         # typecheck (matches CI)
+npm run test:coverage
+npm run test:e2e            # Playwright (requires gitnexus serve + npm run dev)
+```
+
+## Pre-commit hook
+
+A husky pre-commit hook (`.husky/pre-commit`) runs automatically on every `git commit`:
+
+- **`gitnexus-web/` files staged** → `tsc -b --noEmit` + `vitest run`
+- **`gitnexus/` files staged** → `tsc --noEmit` + `vitest run --project default`
+
+Skip with `git commit --no-verify` (use sparingly).
+
+## Test categories
+
+- **Unit** — Pure logic, parsers, graph/query helpers; fast; no network.
+- **Integration** — Real combinations (filesystem, MCP wiring, larger pipelines) as already organized under `gitnexus/test/integration`.
+- **Eval-style / golden sets** — For agent- or classification-style behavior, keep labeled inputs and expected outputs (JSON or table-driven tests) and run them in CI when relevant.
+- **E2E (web)** — Critical user paths only; prefer `data-testid` attributes for stable selectors. Tests run against real backend (`gitnexus serve`) and Vite dev server.
+
+## Performance metrics (targets)
+
+Set targets to match team expectations, then tune to this repo’s CI reality:
+
+| Metric              | Target (initial) | Notes                                      |
+| ------------------- | ---------------- | ------------------------------------------ |
+| Unit coverage       | Align with CI    | CI runs Vitest with coverage in `gitnexus` |
+| Unit wall time      | Fast PR feedback | Use `vitest run test/unit` for tight loop  |
+| Integration duration| < few minutes | Guard heavy tests with env flags if needed |
+
+## Regression testing
+
+Re-run the full relevant suite when:
+
+- Prompt or agent-behavior documentation changes (if tests encode behavior)
+- Model or embedding-related code paths change
+- Graph schema, query contracts, or MCP tool shapes change
+- Dependencies with parsing or runtime impact upgrade
+
+## CI integration
+
+GitHub Actions (`.github/workflows/ci.yml`) orchestrate:
+
+- **`ci-quality.yml`** — `tsc --noEmit` for `gitnexus/` + `tsc -b --noEmit` for `gitnexus-web/`
+- **`ci-tests.yml`** — `vitest run` with coverage (ubuntu) + cross-platform (macOS, Windows)
+- **`ci-e2e.yml`** — Playwright E2E tests, gated on `gitnexus-web/**` changes
+
+Local checks before pushing:
+
+```bash
+cd gitnexus && npx tsc --noEmit && npm test
+cd ../gitnexus-web && npx tsc -b --noEmit && npm test
+```
+
+Or rely on the pre-commit hook which runs these automatically for staged files.
+
+## User acceptance / beta (optional)
+
+For staged releases or UI betas: deploy to a staging environment, collect structured feedback, watch errors and latency, then iterate before a wider release.

WORKER_POOL_IMPLEMENTATION_SUMMARY.md

@@ -1,375 +0,0 @@
-# Worker Pool Implementation Summary for Byterover
-
-## 🎯 Project Context
-**Project**: GitNexus - Client-side, edge-based code knowledge graph generator
-**Implementation Date**: December 2024
-**Primary Goal**: Massive performance improvement for large codebases through parallel processing
-
-## 🚀 Performance Benefits Achieved
-
-### **Expected Speedup by Codebase Size:**
-- **Small codebases (< 100 files)**: 1.5-2x speedup
-- **Medium codebases (100-1000 files)**: 2-4x speedup  
-- **Large codebases (1000+ files)**: 4-8x speedup
-
-### **Key Performance Improvements:**
-- **Parallel file parsing** - Multiple files processed simultaneously
-- **Concurrent Tree-sitter operations** - AST generation in parallel
-- **Better CPU utilization** - Leverages all available cores
-- **Improved UI responsiveness** - Main thread freed up
-
-## 📁 Files Created/Modified
-
-### **Core Implementation Files:**
-
-#### 1. `src/lib/web-worker-pool.ts` (NEW)
-**Purpose**: Browser-compatible Web Worker Pool implementation
-**Key Features**:
-- Replaces Node.js `worker_threads` with standard Web Workers
-- Manages worker lifecycle, task queuing, and error handling
-- Supports progress tracking and batch processing
-- Includes `FileProcessingPool` and `WebWorkerPoolUtils`
-
-**Critical Code Patterns**:
-```typescript
-export class WebWorkerPool {
-  private workers: Worker[] = [];
-  private availableWorkers: Worker[] = [];
-  private taskQueue: WorkerTask<unknown, unknown>[] = [];
-  private activeTasks: Map<string, WorkerTask<unknown, unknown>> = new Map();
-  
-  async execute<TInput, TOutput>(input: TInput): Promise<TOutput>
-  async executeWithProgress<TInput, TOutput>(inputs: TInput[], onProgress?: (completed: number, total: number) => void): Promise<TOutput[]>
-  async shutdown(): Promise<void>
-}
-```
-
-#### 2. `src/core/ingestion/parallel-parsing-processor.ts` (NEW)
-**Purpose**: Parallel file parsing using worker pool
-**Key Features**:
-- Replaces sequential `ParsingProcessor`
-- Uses `tree-sitter-worker.js` for parallel AST parsing
-- Integrates with `FunctionRegistryTrie` for optimized lookups
-- Handles worker pool initialization and cleanup
-
-**Critical Code Patterns**:
-```typescript
-export class ParallelParsingProcessor implements GraphProcessor<ParsingInput> {
-  private workerPool: WebWorkerPool;
-  
-  async process(graph: KnowledgeGraph, input: ParsingInput): Promise<void>
-  private async processFilesInParallel(filePaths: string[], fileContents: Map<string, string>): Promise<ParallelParsingResult[]>
-  private async processResults(results: ParallelParsingResult[], graph: KnowledgeGraph): Promise<void>
-}
-```
-
-#### 3. `src/core/ingestion/parallel-pipeline.ts` (NEW)
-**Purpose**: Parallel 4-pass ingestion pipeline
-**Key Features**:
-- Replaces original `GraphPipeline`
-- Integrates `ParallelParsingProcessor` for Pass 2
-- Provides progress callbacks and performance logging
-- Ensures proper worker resource cleanup
-
-**Critical Code Patterns**:
-```typescript
-export class ParallelGraphPipeline {
-  private parsingProcessor: ParallelParsingProcessor;
-  
-  public async run(input: PipelineInput): Promise<KnowledgeGraph>
-  public static isParallelProcessingSupported(): boolean
-  public static getOptimalWorkerCount(): number
-}
-```
-
-### **Worker Scripts:**
-
-#### 4. `public/workers/tree-sitter-worker.js` (NEW)
-**Purpose**: Dedicated Tree-sitter parsing worker
-**Key Features**:
-- Initializes Tree-sitter and language parsers in worker context
-- Supports TypeScript, JavaScript, Python parsing
-- Extracts definitions using Tree-sitter queries
-- Communicates results back to main thread
-
-#### 5. `public/workers/generic-worker.js` (NEW)
-**Purpose**: General-purpose processing worker
-**Key Features**:
-- Text analysis (word count, identifier extraction)
-- File analysis (basic stats, language detection)
-- Data processing (deduplication, filtering, transformation)
-- Pattern matching and statistical analysis
-
-#### 6. `public/workers/file-processing-worker.js` (NEW)
-**Purpose**: Specialized file processing worker
-**Key Features**:
-- Leverages tree-sitter worker for parsing
-- File structure analysis
-- Dependency extraction (ES6 imports, CommonJS requires)
-- Code complexity analysis
-
-### **Configuration & Testing:**
-
-#### 7. `src/config/feature-flags.ts` (MODIFIED)
-**Changes**: Added worker pool feature flags
-```typescript
-// New flags added:
-enableWorkerPool: boolean;
-enableParallelParsing: boolean;
-enableParallelProcessing: boolean;
-
-// New methods:
-enableWorkerPool(): void
-disableWorkerPool(): void
-```
-
-#### 8. `src/lib/worker-pool-test.ts` (NEW)
-**Purpose**: Comprehensive test suite
-**Key Features**:
-- Basic functionality tests
-- File processing tests
-- Performance benchmarking
-- Error handling tests
-- Browser console testing support
-
-#### 9. `WORKER_POOL_IMPLEMENTATION_GUIDE.md` (NEW)
-**Purpose**: Complete documentation
-**Contents**:
-- Performance benefits and benchmarks
-- File structure and architecture
-- Usage examples and configuration
-- Testing instructions
-- Migration guide from sequential to parallel
-
-## 🔧 Technical Architecture
-
-### **Worker Pool Design Pattern:**
-```typescript
-// Worker Pool Lifecycle
-1. Initialize pool with optimal worker count
-2. Queue tasks for processing
-3. Distribute tasks to available workers
-4. Collect results and handle errors
-5. Recycle workers for next tasks
-6. Shutdown and cleanup resources
-```
-
-### **Parallel Processing Flow:**
-```typescript
-// 4-Pass Pipeline with Parallel Pass 2
-Pass 1: Structure Analysis (Sequential - lightweight)
-Pass 2: Code Parsing (Parallel - CPU intensive) ← NEW
-Pass 3: Import Resolution (Sequential - depends on Pass 2)
-Pass 4: Call Resolution (Sequential - depends on Pass 3)
-```
-
-### **Worker Communication Pattern:**
-```typescript
-// Main Thread → Worker
-worker.postMessage({
-  taskId: string,
-  input: TaskInput
-});
-
-// Worker → Main Thread
-self.postMessage({
-  taskId: string,
-  result: TaskOutput | error: string
-});
-```
-
-## 🎯 Integration Points
-
-### **Feature Flag Integration:**
-```typescript
-// Check if worker pool is enabled
-if (isWorkerPoolEnabled()) {
-  // Use parallel processing
-  const pipeline = new ParallelGraphPipeline();
-} else {
-  // Fallback to sequential processing
-  const pipeline = new GraphPipeline();
-}
-```
-
-### **Performance Monitoring:**
-```typescript
-// Worker pool statistics
-const stats = workerPool.getStats();
-console.log('Worker Pool Stats:', {
-  totalWorkers: stats.totalWorkers,
-  availableWorkers: stats.availableWorkers,
-  activeTasks: stats.activeTasks,
-  queuedTasks: stats.queuedTasks
-});
-```
-
-## 🚨 Error Handling & Fallbacks
-
-### **Worker Pool Error Handling:**
-- Worker crashes are handled gracefully
-- Failed workers are replaced automatically
-- Task timeouts prevent hanging operations
-- Fallback to sequential processing if workers fail
-
-### **Browser Compatibility:**
-- Checks for Web Worker support
-- Graceful degradation for unsupported browsers
-- Hardware concurrency detection
-- Memory usage monitoring
-
-## 📊 Performance Metrics
-
-### **Benchmark Results:**
-- **File Processing**: 4-8x faster for large codebases
-- **Memory Usage**: Efficient worker recycling
-- **CPU Utilization**: Near 100% on multi-core systems
-- **UI Responsiveness**: Main thread remains responsive
-
-### **Scalability:**
-- **Worker Count**: Automatically optimized based on hardware
-- **Task Distribution**: Intelligent load balancing
-- **Memory Management**: Automatic cleanup and recycling
-- **Error Recovery**: Robust error handling and recovery
-
-## 🔄 Migration Strategy
-
-### **From Sequential to Parallel:**
-1. **Feature Flag**: Enable `enableWorkerPool` flag
-2. **Pipeline Switch**: Replace `GraphPipeline` with `ParallelGraphPipeline`
-3. **Processor Update**: Use `ParallelParsingProcessor` for Pass 2
-4. **Testing**: Run comprehensive test suite
-5. **Monitoring**: Track performance improvements
-
-### **Backward Compatibility:**
-- All existing APIs remain unchanged
-- Feature flags control behavior
-- Graceful fallback to sequential processing
-- No breaking changes to existing code
-
-## 🎯 Future Enhancements
-
-### **Planned Improvements:**
-1. **Dynamic Worker Scaling**: Adjust worker count based on load
-2. **Advanced Caching**: Cache parsed ASTs for repeated processing
-3. **Streaming Processing**: Process files as they're uploaded
-4. **Priority Queuing**: Prioritize critical files for processing
-5. **Distributed Processing**: Support for multiple browser tabs/workers
-
-### **Performance Optimizations:**
-1. **Worker Pool Pooling**: Reuse worker pools across sessions
-2. **Memory Optimization**: Better memory management for large files
-3. **Load Balancing**: Intelligent task distribution
-4. **Preemptive Processing**: Start processing before all files are loaded
-
-## 📝 Critical Implementation Details
-
-### **Worker Script Loading:**
-- Worker scripts are served from `/public/workers/`
-- ES6 modules are used for better code organization
-- Tree-sitter WASM files are loaded dynamically
-- Error handling for missing worker scripts
-
-### **Task Serialization:**
-- Tasks are serialized for worker communication
-- Complex objects are simplified for transfer
-- Function references are converted to strings
-- Results are deserialized on main thread
-
-### **Memory Management:**
-- Workers are recycled after task completion
-- Large objects are transferred, not copied
-- Memory usage is monitored and logged
-- Automatic cleanup on pipeline shutdown
-
-## 🔍 Testing Strategy
-
-### **Test Coverage:**
-- **Unit Tests**: Individual worker pool functions
-- **Integration Tests**: End-to-end pipeline testing
-- **Performance Tests**: Benchmarking with various file sizes
-- **Error Tests**: Worker failure and recovery scenarios
-- **Browser Tests**: Cross-browser compatibility
-
-### **Test Commands:**
-```typescript
-// Browser console testing
-window.testWorkerPoolBasic()
-window.testFileProcessingPool()
-window.testWorkerPoolPerformance()
-window.runWorkerPoolTests()
-```
-
-## 📚 Documentation & Resources
-
-### **Key Documentation Files:**
-- `WORKER_POOL_IMPLEMENTATION_GUIDE.md` - Complete implementation guide
-- `src/lib/worker-pool-test.ts` - Test suite with examples
-- `public/workers/*.js` - Worker script documentation
-
-### **Architecture Diagrams:**
-- Worker Pool Lifecycle
-- Parallel Processing Flow
-- Error Handling Flow
-- Performance Monitoring
-
-## 🎯 Success Metrics
-
-### **Performance Improvements:**
-- ✅ 4-8x speedup for large codebases
-- ✅ Improved UI responsiveness
-- ✅ Better CPU utilization
-- ✅ Reduced memory pressure
-
-### **Code Quality:**
-- ✅ Comprehensive error handling
-- ✅ Extensive test coverage
-- ✅ Clear documentation
-- ✅ Backward compatibility
-
-### **User Experience:**
-- ✅ Progress tracking and feedback
-- ✅ Graceful error recovery
-- ✅ Automatic optimization
-- ✅ Feature flag control
-
-## 🔧 Configuration Options
-
-### **Worker Pool Configuration:**
-```typescript
-const workerPool = new WebWorkerPool({
-  maxWorkers: navigator.hardwareConcurrency || 4,
-  workerScript: '/workers/tree-sitter-worker.js',
-  timeout: 60000, // 60 seconds
-  name: 'ParallelParsingPool'
-});
-```
-
-### **Feature Flags:**
-```typescript
-// Enable all worker pool features
-featureFlags.enableWorkerPool();
-
-// Disable worker pool features
-featureFlags.disableWorkerPool();
-
-// Check worker pool status
-const isEnabled = isWorkerPoolEnabled();
-```
-
-## 🚀 Deployment Notes
-
-### **Production Considerations:**
-- Worker scripts must be served from public directory
-- Tree-sitter WASM files must be available
-- Feature flags control rollout
-- Performance monitoring is essential
-- Error logging for debugging
-
-### **Browser Support:**
-- Modern browsers with Web Worker support
-- ES6 module support required
-- WASM support for Tree-sitter
-- Hardware concurrency detection
-
-This implementation represents a significant architectural improvement to GitNexus, providing massive performance benefits for large codebases while maintaining backward compatibility and robust error handling.

compound-engineering.local.md

@@ -0,0 +1,57 @@
+---
+review_agents: [kieran-typescript-reviewer, pattern-recognition-specialist, architecture-strategist, data-integrity-guardian, security-sentinel, performance-oracle, code-simplicity-reviewer]
+plan_review_agents: [kieran-typescript-reviewer, architecture-strategist, code-simplicity-reviewer]
+voltagent_agents: [voltagent-lang:typescript-pro, voltagent-qa-sec:security-auditor, voltagent-data-ai:database-optimizer]
+---
+
+# Review Context
+
+## Project Overview
+GitNexus is a code intelligence tool that builds a knowledge graph from source code using tree-sitter AST parsing across 12 languages and KuzuDB for graph storage. Two packages: `gitnexus/` (CLI/MCP, TypeScript) and `gitnexus-web/` (browser).
+
+## Cross-Language Pattern Consistency (pattern-recognition-specialist)
+- 12 language-specific type extractors in `gitnexus/src/core/ingestion/type-extractors/` must follow identical patterns for: async unwrapping, constructor binding, namespace handling, nullable type stripping, for-loop element typing.
+- Past bugs: C#/Rust missing `await_expression` unwrapping that TypeScript handled correctly; PHP backslash namespace splitting inconsistent with other languages' `::` / `.` splitting.
+- When reviewing type extractor changes, verify the same pattern exists in ALL applicable language files — asymmetry is the #1 source of bugs.
+
+## Data Integrity (data-integrity-guardian)
+- KuzuDB graph operations: schema in `gitnexus/src/core/kuzu/schema.ts`, adapter in `kuzu-adapter.ts`.
+- The ingestion pipeline writes symbols and relationships to the graph — changes to node/relation schemas or the ingestion pipeline can corrupt the index.
+- Known issue: KuzuDB `close()` hangs on Linux due to C++ destructor — use `detachKuzu()` pattern.
+- `lbug-adapter.ts` fallback path needs quote/newline escaping for Cypher injection prevention.
+
+## Security (security-sentinel)
+- Cypher query construction in `lbug-adapter.ts` and `kuzu-adapter.ts` — watch for injection via unescaped user-provided symbol names.
+- CLI accepts `--repo` parameter and file paths — validate against path traversal.
+- MCP server exposes tools to external AI agents — all tool inputs are untrusted.
+
+## Performance (performance-oracle)
+- Tree-sitter buffer size is adaptive (512KB–32MB) via `getTreeSitterBufferSize()` in `constants.ts`.
+- The ingestion pipeline processes entire repositories — O(n) per file with potential O(n²) in cross-file resolution.
+- KuzuDB batch inserts vs individual inserts matter for large repos.
+
+## Architecture (architecture-strategist)
+- Ingestion pipeline phases: structure → parsing → imports → calls → heritage → processes → type resolution.
+- Shared modules: `export-detection.ts`, `constants.ts`, `utils.ts` — changes here have wide blast radius.
+- `gitnexus-web` package drifts behind CLI — flag if a change should be mirrored.
+
+## Voltagent Supplementary Agents
+
+Invoke these via the Agent tool alongside `/ce:review` for deeper specialist analysis. These cover gaps that compound-engineering agents don't:
+
+### voltagent-lang:typescript-pro
+**When:** Changes touch type-resolution logic, generics, conditional types, or complex type-level programming in `type-env.ts`, `type-extractors/*.ts`, or `types.ts`.
+**Why:** The type resolution system uses advanced TypeScript patterns (discriminated unions, mapped types, recursive generics) that benefit from deep TS type-system review beyond what kieran-typescript-reviewer covers.
+
+### voltagent-qa-sec:security-auditor
+**When:** Changes touch MCP tool handlers, Cypher query construction, CLI argument parsing, or any code that processes external input.
+**Why:** GitNexus is an MCP server — all tool inputs come from untrusted AI agents. Systematic OWASP-level audit catches injection vectors that spot-checking misses. Past finding: `lbug-adapter.ts` fallback path had unescaped newlines in Cypher queries.
+
+### voltagent-data-ai:database-optimizer
+**When:** Changes touch `kuzu-adapter.ts`, `schema.ts`, `lbug-adapter.ts`, or any Cypher query construction/execution.
+**Why:** No CE agent specializes in graph database optimization. KuzuDB batch insert patterns, index usage, and query planning directly affect analysis speed on large repos.
+
+## Review Tooling
+- Use `gitnexus_impact()` before approving changes to any symbol — check d=1 (WILL BREAK) callers.
+- Use `gitnexus_detect_changes({scope: "compare", base_ref: "main"})` to map PR diffs to affected execution flows.
+- Use claude-mem to surface past architectural decisions relevant to the code under review.

debug-csv-issue.js

@@ -1 +0,0 @@
- 

debug-csv-line.js

@@ -1 +0,0 @@
- 

docs/code-indexing/cobol/README.md

@@ -0,0 +1,100 @@
+# COBOL Code Indexing
+
+GitNexus indexes COBOL codebases using a **regex-only extraction** strategy, bypassing tree-sitter entirely. This document explains why, how the pipeline works, and links to detailed sub-documents.
+
+## Why Regex-Only?
+
+The tree-sitter-cobol grammar (v0.0.1) has three critical limitations that make it unusable for production indexing:
+
+| Issue | Impact | Severity |
+|-------|--------|----------|
+| External scanner hangs on ~5% of files | No timeout mechanism exists for the C scanner; the process blocks indefinitely | **Blocking** |
+| Only ~15% of paragraph headers detected | Most procedure-division paragraphs are invisible to the grammar | High |
+| Patch markers in cols 1-6 cause parse errors | Enterprise COBOL uses non-standard sequence area content (e.g., `mzADD`, `estero`, `#FIX`) | High |
+
+Because the external scanner hang cannot be interrupted (there is no `setTimeoutMicros` equivalent for tree-sitter), using tree-sitter-cobol would hang the indexing pipeline on a non-trivial fraction of real-world files.
+
+The regex-only approach provides:
+
+- **Speed**: ~1ms per file average extraction time
+- **Reliability**: zero hangs, zero crashes across 13,000+ files
+- **Coverage**: captures all critical symbols -- program name, paragraphs, sections, CALL, PERFORM, COPY, data items (01-77, 88-level), file declarations, FD entries, EXEC SQL/CICS blocks, ENTRY points, and MOVE statements
+
+## Architecture
+
+```mermaid
+flowchart TD
+    A[Repository Scan] --> B{File Detection}
+    B -->|Extension match| C[COBOL file]
+    B -->|GITNEXUS_COBOL_DIRS match| C
+    B -->|No match| Z[Skip]
+
+    C --> D{Copybook?}
+    D -->|Yes| E[Add to Copybook Map]
+    D -->|No| F[Source Program]
+
+    E --> G[COPY Expansion Engine]
+    F --> G
+
+    G -->|Inline copybook content| H[Expanded Source]
+    H --> I[Patch Marker Cleanup]
+    I --> J[Regex State Machine]
+
+    J --> K[Extracted Symbols]
+    K --> L[Graph Model Builder]
+    L --> M[Knowledge Graph]
+
+    subgraph "Per-Chunk Processing"
+        G
+        H
+        I
+        J
+        K
+        L
+    end
+
+    subgraph "Post-Processing"
+        M --> N[Community Detection]
+        M --> O[Process Detection]
+        M --> P[Contract Detection]
+    end
+
+    style J fill:#e8f5e9,stroke:#2e7d32
+    style G fill:#e3f2fd,stroke:#1565c0
+```
+
+## COBOL vs Tree-Sitter Languages
+
+| Feature | COBOL (Regex) | Tree-Sitter Languages |
+|---------|--------------|----------------------|
+| Parser | Single-pass regex state machine | tree-sitter grammar + queries |
+| Speed | ~1ms/file | ~5ms/file |
+| AST available | No | Yes |
+| COPY expansion | Yes (pre-processing step) | N/A |
+| Deep indexing | Data items, SQL, CICS, FD, ENTRY | Type annotations, generics, etc. |
+| Call extraction | PERFORM (intra-file) + CALL (cross-program) | AST-based call site detection |
+| Import extraction | COPY statements | `import`/`require`/`use`/`#include` |
+| Coverage | All critical symbols | Language-dependent query coverage |
+| Failure mode | Never hangs | External scanner can hang (COBOL only) |
+
+## Sub-Documents
+
+| Document | Description |
+|----------|-------------|
+| [File Detection](./file-detection.md) | Extension mapping, `GITNEXUS_COBOL_DIRS`, copybook classification |
+| [COPY Expansion](./copy-expansion.md) | Copybook inlining, REPLACING transformations, cycle detection |
+| [Regex Extraction](./regex-extraction.md) | State machine, regex patterns, line processing |
+| [Deep Indexing](./deep-indexing.md) | Data items, EXEC SQL/CICS, file declarations, FD, ENTRY, MOVE |
+| [Graph Model](./graph-model.md) | COBOL-specific node types, edge types, full annotated example |
+| [Performance](./performance.md) | Benchmarks, worker pool tuning, caps, troubleshooting |
+
+## Key Source Files
+
+| File | Purpose |
+|------|---------|
+| `gitnexus/src/core/ingestion/cobol-preprocessor.ts` | Patch marker cleanup + regex extraction engine |
+| `gitnexus/src/core/ingestion/cobol-copy-expander.ts` | COPY statement expansion with REPLACING |
+| `gitnexus/src/core/ingestion/utils.ts` | `getLanguageFromPath`, `getLanguageFromFilename` |
+| `gitnexus/src/core/ingestion/pipeline.ts` | `isCobolCopybook`, `expandCobolCopies`, `detectCrossProgamContracts` |
+| `gitnexus/src/core/ingestion/workers/parse-worker.ts` | `processCobolRegexOnly` -- graph model builder |
+| `gitnexus/src/core/ingestion/workers/worker-pool.ts` | Configurable sub-batch size for COBOL |

docs/code-indexing/cobol/copy-expansion.md

@@ -0,0 +1,157 @@
+# COBOL COPY Expansion
+
+The COPY statement is COBOL's include mechanism -- analogous to `#include` in C or `import` in modern languages. GitNexus expands COPY statements **before** regex extraction so that symbols defined inside copybooks (data items, paragraphs, etc.) are visible in the program's extracted graph.
+
+## Supported Syntax
+
+### Basic COPY
+
+```cobol
+COPY CPSESP.
+COPY "WORKGRID.CPY".
+```
+
+Inlines the content of the named copybook, replacing the COPY line(s).
+
+### COPY with REPLACING
+
+```cobol
+COPY CPSESP REPLACING "ANAZI-KEY" BY "LK-KEY".
+COPY CPSESP REPLACING LEADING "ESP-" BY "LK-ESP-"
+                       LEADING "KPSESPL" BY "LK-KPSESPL".
+COPY LINKAGE REPLACING TRAILING "-IN" BY "-OUT".
+```
+
+Three REPLACING types are supported:
+
+| Type         | Syntax                               | Behavior                                | Example                          |
+| ------------ | ------------------------------------ | --------------------------------------- | -------------------------------- |
+| **EXACT**    | `REPLACING "OLD" BY "NEW"`           | Replace exact identifier matches        | `ANAZI-KEY` becomes `LK-KEY`     |
+| **LEADING**  | `REPLACING LEADING "PFX-" BY "NEW-"` | Replace prefix on all COBOL identifiers | `ESP-NAME` becomes `LK-ESP-NAME` |
+| **TRAILING** | `REPLACING TRAILING "-IN" BY "-OUT"` | Replace suffix on all COBOL identifiers | `DATA-IN` becomes `DATA-OUT`     |
+
+Multiple REPLACING clauses can appear in a single COPY statement. They are applied in order to each COBOL identifier in the copybook content.
+
+### Multi-Line COPY
+
+COPY statements can span multiple lines (standard COBOL continuation rules apply):
+
+```cobol
+       COPY CPSESP REPLACING
+      -    LEADING "ESP-" BY "LK-ESP-"
+      -    LEADING "KPSESPL" BY "LK-KPSESPL".
+```
+
+Continuation lines (indicator `-` in column 7) are merged before COPY statement scanning.
+
+## Expansion Flow
+
+```mermaid
+sequenceDiagram
+    participant Pipeline
+    participant Expander as COPY Expander
+    participant Resolver
+    participant Reader
+
+    Pipeline->>Pipeline: Identify all COBOL files
+    Pipeline->>Pipeline: Classify copybooks vs programs
+    Pipeline->>Reader: Read all copybook content upfront
+    Reader-->>Pipeline: Copybook content map (name -> content)
+
+    loop For each source file in chunk
+        Pipeline->>Expander: expandCopies(content, filePath, resolveFile, readFile)
+        Expander->>Expander: Merge continuation lines
+        Expander->>Expander: Detect COPY statements via regex
+
+        loop For each COPY statement (reverse order)
+            Expander->>Resolver: resolveFile(copyTarget)
+            Resolver-->>Expander: Copybook key or null
+
+            alt Resolved successfully
+                Expander->>Reader: readFile(resolvedKey)
+                Reader-->>Expander: Copybook content
+
+                Expander->>Expander: Apply REPLACING transformations
+                Expander->>Expander: Recurse for nested COPYs (depth + 1)
+                Expander->>Expander: Splice expanded content into output
+            else Not resolved
+                Expander->>Expander: Keep original COPY line
+            end
+        end
+
+        Expander-->>Pipeline: Expanded content + resolution metadata
+        Pipeline->>Pipeline: Replace file content with expanded content
+    end
+```
+
+The return type `CopyExpansionResult` contains `expandedContent` and `copyResolutions`. The `expansionDepth` field has been removed from the return type (it was unused by callers).
+
+COPY statement line numbers in `CopyResolution` are 1-based (consistent with the preprocessor's line numbering). The splice operation that replaces COPY lines with expanded content adjusts for 0-based array indexing internally.
+
+## Cycle Detection
+
+Circular COPY references (e.g., copybook A includes copybook B which includes copybook A) are detected and handled:
+
+1. Each expansion chain maintains a `visited` set of resolved copybook paths
+2. If a copybook path is already in the visited set, the expansion is skipped
+3. A `warnedCircular` set (internal to `expandCopies()`, not a parameter) deduplicates warning messages within a single file expansion
+
+Known circular copybooks in PROJECT-NAME: `ANAZI`, `ANDIP`, `QDIPE` (self-referential includes).
+
+## Max Depth
+
+Nested COPY expansion is limited to **10 levels** (`DEFAULT_MAX_DEPTH`). If a COPY chain exceeds this depth, a warning is logged and the remaining COPY statements are left unexpanded.
+
+## Max Total Expansions
+
+A breadth amplification guard caps the total number of COPY expansions across all branches within a single file to **500** (`MAX_TOTAL_EXPANSIONS`). This prevents exponential blowup from diamond-shaped COPY graphs where N copybooks each include N other copybooks. Once the limit is reached, further COPY statements in that file are left unexpanded and a single warning is logged.
+
+## REPLACING Application Detail
+
+The REPLACING engine works by scanning all COBOL identifiers (matching `\b[A-Z][A-Z0-9-]*\b`) in the copybook content and applying each replacement rule:
+
+```
+Original copybook content:
+       05  ESP-NAME          PIC X(30).
+       05  ESP-CODE          PIC X(10).
+       05  KPSESPL-FLAG      PIC X(01).
+
+After REPLACING LEADING "ESP-" BY "LK-ESP-" LEADING "KPSESPL" BY "LK-KPSESPL":
+       05  LK-ESP-NAME       PIC X(30).
+       05  LK-ESP-CODE       PIC X(10).
+       05  LK-KPSESPL-FLAG   PIC X(01).
+```
+
+For LEADING replacements, the engine checks if each identifier starts with the `from` prefix (case-insensitive) and replaces only the prefix portion, preserving the rest of the identifier.
+
+For TRAILING replacements, the same logic applies to suffixes.
+
+For EXACT replacements, only identifiers that match the `from` value exactly (case-insensitive) are replaced.
+
+## Copybook Resolution
+
+The resolver tries multiple strategies to match a COPY target name to a copybook file:
+
+1. **Exact match**: `COPY CPSESP` resolves to copybook named `CPSESP`
+2. **Strip extension**: `COPY WORKGRID.CPY` strips `.CPY` and resolves to `WORKGRID`
+3. **Add extension**: `COPY CPSESP` tries `CPSESP.CPY` and `CPSESP.COPY`
+
+If no match is found, the COPY statement is left in place (unexpanded) and a resolution record with `resolvedPath: null` is created.
+
+## Pipeline Integration
+
+The expansion runs **per chunk**, after file content is read but before dispatch to worker threads:
+
+1. All copybook files are read upfront (they are typically small, collectively under 100MB)
+2. Per chunk, the copybook map is merged with chunk content (in case a chunk contains copybooks)
+3. Only programs (not copybooks themselves) undergo expansion
+4. The expanded content replaces the original content in-place before worker dispatch
+
+## Inline Comment Handling
+
+The copy expander's `stripInlineComment()` helper is quote-aware: pipe characters (`|`) inside single- or double-quoted strings are preserved. This matches the same quote-aware logic used by the preprocessor.
+
+## Source Files
+
+- `gitnexus/src/core/ingestion/cobol-copy-expander.ts` -- `expandCopies()`, `parseReplacingClause()`, `applyReplacing()`
+- `gitnexus/src/core/ingestion/pipeline.ts` -- `expandCobolCopies()`, copybook map construction, chunk integration

docs/code-indexing/cobol/deep-indexing.md

@@ -0,0 +1,312 @@
+# COBOL Deep Indexing
+
+Beyond basic symbol extraction (program name, paragraphs, CALL, PERFORM, COPY), GitNexus performs deep indexing of COBOL-specific constructs: data items, EXEC SQL/CICS blocks, file declarations, FD entries, ENTRY points, and MOVE statements.
+
+## Data Items
+
+### Level Numbers
+
+| Level Range | Meaning | Graph Node Type |
+|-------------|---------|-----------------|
+| 01 | Record (group item) | `Record` |
+| 02-49 | Elementary/group items | `Property` |
+| 66 | RENAMES | `Property` |
+| 77 | Independent item | `Property` |
+| 88 | Condition name | `Const` |
+
+FILLER items are skipped (no useful name for the graph).
+
+### Clauses Parsed
+
+The `parseDataItemClauses()` function extracts these clauses from the trailing text of a data item declaration:
+
+| Clause | Pattern | Example |
+|--------|---------|---------|
+| `PIC` / `PICTURE` | `\bPIC(?:TURE)?\s+(?:IS\s+)?(\S+)` | `PIC X(30)`, `PICTURE IS 9(5)V99` |
+| `USAGE` | `\bUSAGE\s+(?:IS\s+)?(COMP\|BINARY\|...)` | `USAGE IS COMP-3`, `BINARY` |
+| `REDEFINES` | `\bREDEFINES\s+([A-Z][A-Z0-9-]+)` | `REDEFINES WK-DATE-NUM` |
+| `OCCURS` | `\bOCCURS\s+(\d+)` | `OCCURS 12 TIMES` |
+
+Standalone COMP variants (without the `USAGE` keyword) are also detected: `COMP`, `COMP-1` through `COMP-6`, `COMP-X`, `BINARY`, `PACKED-DECIMAL`.
+
+### Data Hierarchy
+
+Data items form a hierarchical structure based on level numbers. The extractor uses a **stack algorithm**:
+
+```
+Processing order:
+  01 WK-RECORD          -> push {01, WK-RECORD}   -> parent: Module
+  05 WK-NAME            -> push {05, WK-NAME}     -> parent: WK-RECORD (01 < 05)
+  10 WK-FIRST           -> push {10, WK-FIRST}    -> parent: WK-NAME (05 < 10)
+  10 WK-LAST            -> pop WK-FIRST, push      -> parent: WK-NAME (05 < 10)
+  05 WK-CODE            -> pop WK-LAST, WK-NAME    -> parent: WK-RECORD (01 < 05)
+  88 WK-ACTIVE          -> (88 handled separately)  -> parent: WK-CODE
+```
+
+The stack maintains items where each entry's level is strictly less than the next. When a new item arrives with a level <= the top of stack, items are popped until the stack top has a smaller level. A `CONTAINS` edge is created from the stack top to the new item.
+
+For 88-level condition names, the parent is the immediately preceding non-88 data item (found by scanning backwards).
+
+### Annotated Example
+
+```cobol
+       01  WK-EMPLOYEE.
+           05  WK-EMP-ID          PIC 9(6).
+           05  WK-EMP-NAME        PIC X(30).
+           05  WK-EMP-STATUS      PIC X(01).
+               88  WK-ACTIVE      VALUE "A".
+               88  WK-INACTIVE    VALUE "I".
+           05  WK-SALARY          PIC 9(7)V99 COMP-3.
+           05  WK-DEPT            PIC X(04) OCCURS 3 TIMES.
+```
+
+Produces:
+- `Record` node: `WK-EMPLOYEE` (level 01, section: working-storage)
+- `Property` nodes: `WK-EMP-ID`, `WK-EMP-NAME`, `WK-EMP-STATUS`, `WK-SALARY`, `WK-DEPT`
+- `Const` nodes: `WK-ACTIVE` (values: `A`), `WK-INACTIVE` (values: `I`)
+- `CONTAINS` edges: `WK-EMPLOYEE -> WK-EMP-ID`, `WK-EMPLOYEE -> WK-EMP-NAME`, etc.
+- `CONTAINS` edges: `WK-EMP-STATUS -> WK-ACTIVE`, `WK-EMP-STATUS -> WK-INACTIVE`
+
+### Data Item Cap
+
+A maximum of **500 data items per file** (`MAX_DATA_ITEMS_PER_FILE`) are processed. Some COBOL programs (especially after COPY expansion) can have 10,000+ data items, which would cause graph bloat and push the V8 relationship Map past its 16.7M entry limit across thousands of files.
+
+The cap applies after extraction: the first 500 items in source order are kept. Since 01-level records appear first, critical top-level structure is preserved.
+
+## EXEC SQL
+
+EXEC SQL blocks are accumulated across lines between `EXEC SQL` and `END-EXEC`, then parsed as a unit.
+
+### Operation Classification
+
+The first SQL keyword determines the operation:
+
+| First Keyword | Operation |
+|---------------|-----------|
+| `SELECT` | SELECT |
+| `INSERT` | INSERT |
+| `UPDATE` | UPDATE |
+| `DELETE` | DELETE |
+| `DECLARE` | DECLARE |
+| `OPEN` | OPEN |
+| `CLOSE` | CLOSE |
+| `FETCH` | FETCH |
+| *(anything else)* | OTHER |
+
+### Table Extraction
+
+Tables are extracted from SQL clauses:
+
+| Clause Pattern | Example |
+|----------------|---------|
+| `FROM <table>` | `SELECT * FROM EMPLOYEES` |
+| `INSERT INTO <table>` | `INSERT INTO EMPLOYEES` |
+| `UPDATE <table>` | `UPDATE EMPLOYEES SET ...` |
+| `JOIN <table>` | `LEFT JOIN DEPARTMENTS ON ...` |
+
+Note: The `INTO` pattern is restricted to `INSERT INTO` to avoid false positives from `FETCH ... INTO :host-var` and `SELECT ... INTO :host-var` statements, where `INTO` introduces host variables rather than table names.
+
+### Cursor Detection
+
+```cobol
+           EXEC SQL
+               DECLARE C-EMPLOYEES CURSOR FOR
+               SELECT EMP-ID, EMP-NAME FROM EMPLOYEES
+               WHERE DEPT = :WK-DEPT
+           END-EXEC
+```
+
+Extracts: cursor `C-EMPLOYEES`, table `EMPLOYEES`, host variable `WK-DEPT`.
+
+### Host Variables
+
+Host variables are COBOL variables referenced in SQL with a `:` prefix. The colon is stripped:
+
+```sql
+WHERE EMP-ID = :WK-EMP-ID AND DEPT = :WK-DEPT
+```
+
+Extracts: `WK-EMP-ID`, `WK-DEPT`.
+
+### Graph Output
+
+- `CodeElement` node per table, with description `sql-table op:{OP}`
+- `CodeElement` node per cursor, with description `sql-cursor`
+- `ACCESSES` edge from Module to each CodeElement
+- Deduplication: if the same table appears in multiple SQL blocks, only one node is created
+
+## EXEC CICS
+
+EXEC CICS blocks are accumulated and parsed similarly to SQL blocks.
+
+### Command Detection
+
+Two-word commands are detected first (matched against the block start):
+
+```
+SEND MAP, RECEIVE MAP, SEND TEXT, SEND CONTROL, READ NEXT, READ PREV
+```
+
+If no two-word command matches, the first word is used (e.g., `LINK`, `XCTL`, `RETURN`, `READ`, `WRITE`).
+
+### Extraction
+
+| Element | Pattern | Example |
+|---------|---------|---------|
+| MAP name | `MAP('name')` or `MAP("name")` | `EXEC CICS SEND MAP('EMPMENU')` |
+| PROGRAM name | `PROGRAM('name')` or `PROGRAM("name")` | `EXEC CICS LINK PROGRAM('BGTABUP')` |
+| TRANSID | `TRANSID('name')` or `TRANSID("name")` | `EXEC CICS START TRANSID('EMP1')` |
+
+### Graph Output
+
+- MAP: `CodeElement` node with description `cics-map cmd:{CMD}` + `ACCESSES` edge from Module
+- PROGRAM: `CALLS` edge (cross-program call via CICS LINK/XCTL)
+- TRANSID: `CodeElement` node with description `cics-transid cmd:{CMD}` + `ACCESSES` edge from Module
+
+### Annotated Example
+
+```cobol
+           EXEC CICS
+               SEND MAP('EMPMENU')
+               MAPSET('EMPSET')
+               FROM(WK-MAP-DATA)
+               ERASE
+           END-EXEC
+```
+
+Produces:
+- `CodeElement` node: `EMPMENU` (description: `cics-map cmd:SEND MAP`)
+- `ACCESSES` edge: Module -> `EMPMENU`
+
+## File Declarations
+
+SELECT statements in the INPUT-OUTPUT SECTION are accumulated across multiple lines (until a period terminator) and parsed for:
+
+| Clause | Pattern | Example |
+|--------|---------|---------|
+| SELECT | `SELECT <name>` | `SELECT MASTER-FILE` |
+| ASSIGN | `ASSIGN TO <file>` | `ASSIGN TO "MASTER.DAT"` |
+| ORGANIZATION | `ORGANIZATION IS <type>` | `ORGANIZATION IS INDEXED` |
+| ACCESS | `ACCESS MODE IS <mode>` | `ACCESS MODE IS DYNAMIC` |
+| RECORD KEY | `RECORD KEY IS <field>` | `RECORD KEY IS WK-EMP-ID` |
+| FILE STATUS | `FILE STATUS IS <field>` | `FILE STATUS IS WK-FILE-STATUS` |
+
+### Graph Output
+
+- `CodeElement` node with description containing all parsed clauses (e.g., `select org:INDEXED access:DYNAMIC key:WK-EMP-ID status:WK-FILE-STATUS assign:MASTER.DAT`)
+- `RECORD_KEY_OF` edge: from Property node to CodeElement (confidence 0.8)
+- `FILE_STATUS_OF` edge: from Property node to CodeElement (confidence 0.8)
+
+## FD Entries
+
+FD (File Description) entries associate a file name with its record layout:
+
+```cobol
+       FD  MASTER-FILE.
+       01  MASTER-RECORD.
+           05  MR-EMP-ID       PIC 9(6).
+           05  MR-EMP-NAME     PIC X(30).
+```
+
+The extractor tracks `pendingFdName` state: when an `FD` line is seen, the next 01-level data item becomes its record.
+
+### Graph Output
+
+- `CodeElement` node with description `fd record:{recordName}`
+- `CONTAINS` edge: FD CodeElement -> Record node
+- `CONTAINS` edge: SELECT CodeElement -> FD CodeElement (linking file declaration to file description)
+
+## ENTRY Points
+
+The `ENTRY` statement defines additional entry points into a COBOL program (in addition to the main program entry):
+
+```cobol
+       ENTRY "SUBPROG" USING WK-PARAM-1 WK-PARAM-2.
+```
+
+### Graph Output
+
+- `Constructor` node with description `entry params:{param1},{param2}` (or just `entry` if no parameters)
+- `CONTAINS` edge: Module -> Constructor
+- Symbol table entry (so the entry point is discoverable by name)
+
+## PROCEDURE DIVISION USING
+
+```cobol
+       PROCEDURE DIVISION USING WK-INPUT-REC WK-OUTPUT-REC.
+```
+
+The USING clause identifies parameters received by the program from its caller.
+
+### Graph Output
+
+- `RECEIVES` edge: Module -> Property (for each parameter name, confidence 0.8)
+
+## MOVE Statements
+
+MOVE statements produce `ACCESSES` edges in the graph:
+
+```cobol
+       MOVE WK-NAME TO OUT-NAME.
+       MOVE CORRESPONDING WK-INPUT TO WK-OUTPUT.
+       MOVE CORR WK-IN TO WK-OUT.
+```
+
+### Extraction Details
+
+- Source and target identifiers are captured
+- `CORRESPONDING` and its abbreviation `CORR` are both recognized (bulk field-by-field move)
+- Figurative constants (SPACES, ZEROS, LOW-VALUES, HIGH-VALUES, QUOTES, ALL) are skipped
+- The enclosing paragraph (`caller`) is tracked for context
+
+### MOVE CORRESPONDING / CORR Edge Reasons
+
+MOVE CORRESPONDING (and CORR) produces distinct edge reasons to differentiate from simple MOVE:
+
+| Edge | Reason (simple MOVE) | Reason (CORRESPONDING/CORR) |
+|------|---------------------|-----------------------------|
+| Read (source) | `cobol-move-read` | `cobol-move-corresponding-read` |
+| Write (target) | `cobol-move-write` | `cobol-move-corresponding-write` |
+
+This distinction allows queries to find bulk field-by-field moves separately from simple variable assignments.
+
+## GO TO DEPENDING ON
+
+The `GO TO` statement with multiple targets and a `DEPENDING ON` clause is a computed branch:
+
+```cobol
+       GO TO PARA-1 PARA-2 PARA-3
+           DEPENDING ON WK-SELECTOR.
+```
+
+All target paragraph names are extracted and emitted as separate `gotos` entries. Each target produces a `CALLS` edge in the graph (same semantics as PERFORM). The `DEPENDING ON` variable is not currently tracked as a data-flow dependency.
+
+## SORT INPUT/OUTPUT PROCEDURE
+
+SORT and MERGE statements can specify procedural entry points instead of file-based I/O:
+
+```cobol
+       SORT SORT-FILE ON ASCENDING KEY SORT-KEY
+           INPUT PROCEDURE IS PREPARE-INPUT
+           OUTPUT PROCEDURE IS FORMAT-OUTPUT.
+```
+
+`INPUT PROCEDURE IS` and `OUTPUT PROCEDURE IS` targets are extracted as control-flow targets (same as PERFORM). They produce `performs` entries and corresponding `CALLS` edges in the graph.
+
+## Fixed-Format Literal Continuation
+
+In fixed-format COBOL, string literals can span multiple lines using the continuation indicator (`-` in column 7). When a continuation line starts with a quote character, the extractor joins it with the predecessor by removing the trailing quote from the previous line and the opening quote from the continuation:
+
+```
+Line N:          MOVE "THIS IS A LONG STRI
+Line N+1 (cont): -    "NG VALUE" TO WK-FIELD.
+Merged:          MOVE "THIS IS A LONG STRING VALUE" TO WK-FIELD.
+```
+
+The trailing `"` on line N and the opening `"` on line N+1 are both removed, producing a seamless literal. If no matching quote is found on the predecessor line, the continuation is appended as-is.
+
+## Source Files
+
+- `gitnexus/src/core/ingestion/cobol-preprocessor.ts` -- All extraction logic, clause parsers, EXEC block parsers
+- `gitnexus/src/core/ingestion/workers/parse-worker.ts` -- `processCobolRegexOnly()`, graph node/edge emission
+- `gitnexus/src/core/ingestion/parsing-processor.ts` -- Sequential fallback with same `MAX_DATA_ITEMS_PER_FILE` cap

docs/code-indexing/cobol/file-detection.md

@@ -0,0 +1,126 @@
+# COBOL File Detection
+
+GitNexus detects COBOL files through two mechanisms: extension-based mapping and directory-based override for extensionless files. This document covers both, plus the copybook/program classification logic.
+
+## Extension Mapping
+
+### Program Extensions
+
+| Extension | Type |
+|-----------|------|
+| `.cbl` | COBOL program |
+| `.cob` | COBOL program |
+| `.cobol` | COBOL program |
+
+### Copybook Extensions
+
+| Extension | Type | Notes |
+|-----------|------|-------|
+| `.cpy` | Copybook | Standard |
+| `.copy` | Copybook | Standard |
+| `.gnm` / `.GNM` | Copybook | Enterprise (GnuCOBOL naming) |
+| `.fd` / `.FD` | Copybook | File Description fragment |
+| `.wrk` / `.WRK` | Copybook | Working-Storage fragment |
+| `.sel` / `.SEL` | Copybook | SELECT clause fragment |
+| `.open` / `.OPEN` | Copybook | File OPEN fragment |
+| `.close` / `.CLOSE` | Copybook | File CLOSE fragment |
+| `.ini` / `.INI` | Copybook | Initialization fragment |
+| `.def` / `.DEF` | Copybook | Definition fragment |
+
+All extension matching is case-sensitive in `getLanguageFromFilename` (the extensions above are matched as written, including uppercase variants like `.GNM`).
+
+## Extensionless File Detection: `GITNEXUS_COBOL_DIRS`
+
+Many enterprise COBOL repositories use extensionless files -- the filename alone identifies the program (e.g., `s/BGTABFL` is the source for program `BGTABFL`). GitNexus handles this via the `GITNEXUS_COBOL_DIRS` environment variable.
+
+### Configuration
+
+Set `GITNEXUS_COBOL_DIRS` to a comma-separated list of directory names:
+
+```bash
+# Files in s/, c/, and wfproc/ directories (at any depth) are treated as COBOL
+export GITNEXUS_COBOL_DIRS=s,c,wfproc
+```
+
+The matching is **case-insensitive** and checks all path segments:
+
+- `/repo/s/BGTABFL` -- matches segment `s` -- COBOL
+- `/repo/src/c/CPSESP` -- matches segment `c` -- COBOL
+- `/repo/wfproc/WF001` -- matches segment `wfproc` -- COBOL
+- `/repo/docs/README` -- no matching segment -- skipped
+
+### Decision Tree
+
+```mermaid
+flowchart TD
+    A[getLanguageFromPath] --> B[getLanguageFromFilename]
+    B --> C{Known extension?}
+    C -->|Yes .cbl/.cob/.cobol/.cpy/...| D[Return COBOL]
+    C -->|Yes .ts/.py/.java/...| E[Return other language]
+    C -->|No match| F{Has extension?}
+
+    F -->|"Has dot in basename"| G[Return null]
+    F -->|"No dot = extensionless"| H{GITNEXUS_COBOL_DIRS set?}
+
+    H -->|No| G
+    H -->|Yes| I{Any path segment<br/>matches a configured dir?}
+
+    I -->|Yes| D
+    I -->|No| G
+
+    style D fill:#e8f5e9,stroke:#2e7d32
+    style G fill:#ffebee,stroke:#c62828
+```
+
+### Implementation Detail
+
+The `GITNEXUS_COBOL_DIRS` value is parsed once (on first call) and cached in a `Set<string>`:
+
+```typescript
+// From gitnexus/src/core/ingestion/utils.ts
+const getCobolDirs = (): Set<string> => {
+  if (_cobolDirs) return _cobolDirs;
+  const raw = process.env.GITNEXUS_COBOL_DIRS;
+  _cobolDirs = raw
+    ? new Set(raw.split(',').map(d => d.trim().toLowerCase()))
+    : new Set();
+  return _cobolDirs;
+};
+```
+
+The path segment check splits the full path on `/` and tests each segment against the cached set.
+
+## Copybook vs Program Classification
+
+After a file is identified as COBOL, it must be classified as either a **program** (to be parsed for symbols) or a **copybook** (to be loaded into the copybook map for COPY expansion).
+
+### Classification Rules
+
+A COBOL file is classified as a **copybook** if ANY of these conditions is true:
+
+1. It has a recognized copybook extension (`.cpy`, `.copy`, `.gnm`, `.fd`, `.wrk`, `.sel`, `.open`, `.close`, `.ini`, `.def`)
+2. It is an extensionless file whose path contains a directory segment matching one of: `c`, `copy`, `copybooks`, `copylib`, `cpy`
+
+A file is classified as a **program** if:
+
+1. It has a program extension (`.cbl`, `.cob`, `.cobol`), OR
+2. It is extensionless and does NOT match any copybook directory pattern
+
+### Copybook Name Resolution
+
+Copybook names are derived from the filename:
+
+- Strip the extension (if any)
+- Convert to uppercase
+
+Examples:
+- `c/CPSESP` -- name: `CPSESP`
+- `copy/workgrid.cpy` -- name: `WORKGRID`
+- `c/ANAZI.GNM` -- name: `ANAZI`
+
+This name is used to resolve `COPY CPSESP.` statements during expansion.
+
+## Source Files
+
+- `gitnexus/src/core/ingestion/utils.ts` -- `getLanguageFromPath()`, `getLanguageFromFilename()`, `getCobolDirs()`
+- `gitnexus/src/core/ingestion/pipeline.ts` -- `isCobolCopybook()`, `getCopybookName()`, `COPYBOOK_EXTENSIONS`, `COBOL_PROGRAM_EXTENSIONS`

docs/code-indexing/cobol/graph-model.md

@@ -0,0 +1,193 @@
+# COBOL Graph Model
+
+This document describes the graph nodes and edges that GitNexus creates for COBOL codebases. The COBOL graph model is richer than most tree-sitter languages because it captures domain-specific constructs: file declarations, FD entries, data hierarchies, SQL tables, CICS maps, and cross-program contracts.
+
+## Entity-Relationship Diagram
+
+```mermaid
+erDiagram
+    File ||--o{ Module : DEFINES
+    File ||--o{ Function : DEFINES
+    File ||--o{ Namespace : DEFINES
+    File ||--o{ Record : DEFINES
+    File ||--o{ Property : DEFINES
+    File ||--o{ Const : DEFINES
+    File ||--o{ CodeElement : DEFINES
+    File ||--o{ Constructor : DEFINES
+    File }o--o{ File : IMPORTS
+
+    Module ||--o{ Record : CONTAINS
+    Module ||--o{ Constructor : CONTAINS
+    Module }o--o{ CodeElement : ACCESSES
+    Module }o--o{ Module : CALLS
+    Module }o--o{ Module : CONTRACTS
+    Module }o--o{ Property : RECEIVES
+
+    Record ||--o{ Property : CONTAINS
+    Record ||--o{ Const : CONTAINS
+    Record }o--o{ Record : REDEFINES
+
+    Property ||--o{ Property : CONTAINS
+    Property ||--o{ Const : CONTAINS
+    Property }o--o{ Property : REDEFINES
+    Property }o--o{ CodeElement : RECORD_KEY_OF
+    Property }o--o{ CodeElement : FILE_STATUS_OF
+
+    CodeElement ||--o{ CodeElement : CONTAINS
+    CodeElement ||--o{ Record : CONTAINS
+
+    Function }o--o{ Function : CALLS
+```
+
+## Node Types
+
+| Node Type | COBOL Concept | Created From | Example |
+|-----------|--------------|--------------|---------|
+| `Module` | PROGRAM-ID | `PROGRAM-ID. BGTABFL` | Name: `BGTABFL`, description may include author and date |
+| `Function` | Paragraph | `PROCESS-RECORD.` at column 8 | Name: `PROCESS-RECORD` |
+| `Namespace` | Procedure section | `MAIN-LOGIC SECTION.` at column 8 | Name: `MAIN-LOGIC` |
+| `Record` | 01-level data item | `01 WK-EMPLOYEE.` | Description: `level:01 section:working-storage` |
+| `Property` | 02-49/66/77 data item | `05 WK-NAME PIC X(30).` | Description: `level:05 pic:X(30) section:working-storage` |
+| `Const` | 88-level condition | `88 WK-ACTIVE VALUE "A".` | Description: `level:88 values:A` |
+| `CodeElement` | SELECT, FD, SQL table, CICS map, cursor, transid | Various | Description varies by subtype |
+| `Constructor` | ENTRY point | `ENTRY "SUBPROG" USING WK-DATA` | Description: `entry params:WK-DATA` |
+
+### CodeElement Subtypes
+
+CodeElement is used for multiple COBOL constructs, distinguished by their description prefix:
+
+| Subtype | ID Pattern | Description Format | Example |
+|---------|-----------|-------------------|---------|
+| File SELECT | `CodeElement:{path}:SELECT:{name}` | `select org:INDEXED access:DYNAMIC ...` | `SELECT MASTER-FILE` |
+| FD entry | `CodeElement:{path}:FD:{name}` | `fd record:{recordName}` | `FD MASTER-FILE` |
+| SQL table | `CodeElement:{path}:sql-table:{name}` | `sql-table op:SELECT` | Table `EMPLOYEES` |
+| SQL cursor | `CodeElement:{path}:sql-cursor:{name}` | `sql-cursor` | Cursor `C-EMPLOYEES` |
+| CICS map | `CodeElement:{path}:cics-map:{name}` | `cics-map cmd:SEND MAP` | Map `EMPMENU` |
+| CICS transid | `CodeElement:{path}:cics-transid:{name}` | `cics-transid cmd:START` | Transid `EMP1` |
+
+## Edge Types
+
+| Edge Type | Source | Target | Created By | Confidence | Example |
+|-----------|--------|--------|-----------|------------|---------|
+| `DEFINES` | File | any node | File defines its symbols | 1.0 | File -> Module `BGTABFL` |
+| `CALLS` | Function | Function | `PERFORM X [THRU Y]` | (via call-processor) | `PROCESS-RECORD` -> `CALC-TAX` |
+| `CALLS` | Module | Module | `CALL "BGTABUP"` | (via call-processor) | `BGTABFL` -> `BGTABUP` |
+| `CALLS` | Module | Module | `EXEC CICS LINK PROGRAM('X')` | (via call-processor) | `BGTABFL` -> `BGTABUP` |
+| `IMPORTS` | File | File | `COPY copybook` | (via import-processor) | Source file -> Copybook file |
+| `CONTAINS` | Module | Record | Data hierarchy root | 1.0 | `BGTABFL` -> `WK-EMPLOYEE` |
+| `CONTAINS` | Record | Property | Data hierarchy | 1.0 | `WK-EMPLOYEE` -> `WK-NAME` |
+| `CONTAINS` | Property | Property | Nested data items | 1.0 | `WK-ADDRESS` -> `WK-CITY` |
+| `CONTAINS` | Record/Property | Const | 88-level parent | 1.0 | `WK-STATUS` -> `WK-ACTIVE` |
+| `CONTAINS` | CodeElement (FD) | Record | FD record link | 1.0 | `FD:MASTER-FILE` -> `MASTER-RECORD` |
+| `CONTAINS` | CodeElement (SELECT) | CodeElement (FD) | SELECT-FD link | 0.9 | `SELECT:MASTER-FILE` -> `FD:MASTER-FILE` |
+| `CONTAINS` | Module | Constructor | ENTRY in module | 1.0 | `BGTABFL` -> `SUBPROG` |
+| `REDEFINES` | Record | Record | `01 X REDEFINES Y` | 1.0 | `WK-DATE-NUM` -> `WK-DATE-ALPHA` |
+| `REDEFINES` | Property | Property | `05 X REDEFINES Y` | 1.0 | `WK-CODE-NUM` -> `WK-CODE-ALPHA` |
+| `RECORD_KEY_OF` | Property | CodeElement (SELECT) | `RECORD KEY IS field` | 0.8 | `WK-EMP-ID` -> `SELECT:MASTER-FILE` |
+| `FILE_STATUS_OF` | Property | CodeElement (SELECT) | `FILE STATUS IS field` | 0.8 | `WK-FS` -> `SELECT:MASTER-FILE` |
+| `ACCESSES` | Module | CodeElement | EXEC SQL/CICS | 0.9 | `BGTABFL` -> `sql-table:EMPLOYEES` |
+| `RECEIVES` | Module | Property | `PROCEDURE USING` | 0.8 | `BGTABFL` -> `WK-INPUT-REC` |
+| `CONTRACTS` | Module | Module | Shared copybook detection | 0.9 | `BGTABFL` -> `BGTABUP` (via `CPSESP`) |
+
+## Full Annotated Example
+
+Given this COBOL program:
+
+```cobol
+       IDENTIFICATION DIVISION.
+       PROGRAM-ID. EMPMAINT.
+       AUTHOR. Development Team.
+
+       ENVIRONMENT DIVISION.
+       INPUT-OUTPUT SECTION.
+       FILE-CONTROL.
+           SELECT EMP-FILE
+               ASSIGN TO "EMPLOYEE.DAT"
+               ORGANIZATION IS INDEXED
+               ACCESS MODE IS DYNAMIC
+               RECORD KEY IS EMP-ID
+               FILE STATUS IS WS-FILE-STATUS.
+
+       DATA DIVISION.
+       FILE SECTION.
+       FD  EMP-FILE.
+       01  EMP-RECORD.
+           05  EMP-ID             PIC 9(6).
+           05  EMP-NAME           PIC X(30).
+
+       WORKING-STORAGE SECTION.
+       01  WS-FLAGS.
+           05  WS-FILE-STATUS     PIC X(02).
+           05  WS-EOF-FLAG        PIC X(01).
+               88  WS-EOF         VALUE "Y".
+
+       LINKAGE SECTION.
+       01  LK-SEARCH-KEY          PIC 9(6).
+
+       PROCEDURE DIVISION USING LK-SEARCH-KEY.
+       MAIN-LOGIC SECTION.
+       MAIN-START.
+           PERFORM OPEN-FILE
+           PERFORM PROCESS-RECORDS
+           PERFORM CLOSE-FILE
+           STOP RUN.
+
+       OPEN-FILE.
+           OPEN I-O EMP-FILE.
+
+       PROCESS-RECORDS.
+           MOVE LK-SEARCH-KEY TO EMP-ID
+           EXEC SQL
+               SELECT EMP_SALARY INTO :WS-SALARY
+               FROM EMPLOYEES
+               WHERE EMP_ID = :EMP-ID
+           END-EXEC
+           CALL "EMPREPORT".
+
+       CLOSE-FILE.
+           CLOSE EMP-FILE.
+```
+
+The graph produced contains:
+
+**Nodes:**
+- `Module`: EMPMAINT (description: `author:Development Team`)
+- `Namespace`: MAIN-LOGIC
+- `Function`: MAIN-START, OPEN-FILE, PROCESS-RECORDS, CLOSE-FILE
+- `Record`: EMP-RECORD, WS-FLAGS, LK-SEARCH-KEY
+- `Property`: EMP-ID, EMP-NAME, WS-FILE-STATUS, WS-EOF-FLAG
+- `Const`: WS-EOF (values: Y)
+- `CodeElement`: SELECT:EMP-FILE, FD:EMP-FILE, sql-table:EMPLOYEES
+- (COPY imports, if any, would produce File IMPORTS edges)
+
+**Edges:**
+- `DEFINES`: File -> all nodes
+- `CONTAINS`: EMPMAINT -> EMP-RECORD, EMPMAINT -> WS-FLAGS, EMPMAINT -> LK-SEARCH-KEY
+- `CONTAINS`: EMP-RECORD -> EMP-ID, EMP-RECORD -> EMP-NAME
+- `CONTAINS`: WS-FLAGS -> WS-FILE-STATUS, WS-FLAGS -> WS-EOF-FLAG
+- `CONTAINS`: WS-EOF-FLAG -> WS-EOF
+- `CONTAINS`: FD:EMP-FILE -> EMP-RECORD
+- `CONTAINS`: SELECT:EMP-FILE -> FD:EMP-FILE
+- `CALLS`: MAIN-START -> OPEN-FILE, MAIN-START -> PROCESS-RECORDS, MAIN-START -> CLOSE-FILE
+- `CALLS`: EMPMAINT -> EMPREPORT (external CALL)
+- `ACCESSES`: EMPMAINT -> sql-table:EMPLOYEES
+- `RECEIVES`: EMPMAINT -> LK-SEARCH-KEY (PROCEDURE USING)
+- `RECORD_KEY_OF`: EMP-ID -> SELECT:EMP-FILE
+- `FILE_STATUS_OF`: WS-FILE-STATUS -> SELECT:EMP-FILE
+
+## How COBOL Differs from Tree-Sitter Languages
+
+| Aspect | COBOL | Tree-Sitter Languages |
+|--------|-------|----------------------|
+| Node variety | 8 types (Module, Function, Namespace, Record, Property, Const, CodeElement, Constructor) | Typically 4-6 (Function, Class, Method, Interface, Module, Const) |
+| Domain edges | RECORD_KEY_OF, FILE_STATUS_OF, ACCESSES, RECEIVES, CONTRACTS, REDEFINES | Primarily CALLS, IMPORTS, EXTENDS, IMPLEMENTS |
+| Data hierarchy | Deep CONTAINS chains (01 -> 05 -> 10 -> 88) | Flat class members |
+| Cross-program calls | CALL "name" + CICS LINK PROGRAM | Import-based resolution |
+| Contract detection | Shared COPY copybook between caller/callee | Not applicable |
+| Metadata | AUTHOR, DATE-WRITTEN on Module | JSDoc/docstring (not indexed) |
+
+## Source Files
+
+- `gitnexus/src/core/ingestion/workers/parse-worker.ts` -- `processCobolRegexOnly()`, node/edge emission logic
+- `gitnexus/src/core/ingestion/pipeline.ts` -- `detectCrossProgamContracts()` for CONTRACTS edges
+- `gitnexus/src/core/ingestion/cobol-preprocessor.ts` -- `CobolRegexResults` interface (all extracted data)

docs/code-indexing/cobol/performance.md

@@ -0,0 +1,261 @@
+# COBOL Performance and Tuning
+
+This document covers real-world benchmarks, worker pool configuration, memory management, known limitations, and troubleshooting for COBOL indexing.
+
+## PROJECT-NAME Benchmark
+
+The PROJECT-NAME project is a large Italian payroll system written in COBOL. It serves as the primary benchmark for COBOL indexing performance.
+
+### Input
+
+| Metric                      | Value                                                                        |
+| --------------------------- | ---------------------------------------------------------------------------- |
+| Paths scanned               | 14,217                                                                       |
+| Parseable files             | 13,129                                                                       |
+| Total source size           | 224 MB                                                                       |
+| Chunks                      | 12 (at 20 MB budget)                                                         |
+| Copybooks loaded            | 2,976                                                                        |
+| Copybooks used in expansion | 2,955                                                                        |
+| Key directories             | `s/` (7773 programs), `c/` (3036 copybooks), `wfproc/` (1973 workflow files) |
+
+### Output
+
+| Metric                 | Value  |
+| ---------------------- | ------ |
+| Graph nodes            | 2.79M  |
+| Graph edges            | 5.67M  |
+| Clusters (communities) | 16,679 |
+| Execution flows        | 300    |
+
+### Timing
+
+| Phase                           | Duration          |
+| ------------------------------- | ----------------- |
+| Total                           | ~251s             |
+| KuzuDB write                    | 132s              |
+| Full-text search indexing       | 6.7s              |
+| Regex extraction (avg per file) | ~1ms              |
+| COPY expansion + deep indexing  | Remainder (~112s) |
+
+### Indexing Command
+
+```bash
+cd /path/to/PROJECT-NAME
+GITNEXUS_COBOL_DIRS=s,c,wfproc GITNEXUS_VERBOSE=1 node --max-old-space-size=8192 \
+  /path/to/gitnexus/dist/cli/index.js analyze --force
+```
+
+## Open-Source Benchmarks
+
+### CardDemo (AWS)
+
+| Metric | Value |
+| ------ | ----- |
+| Graph nodes | 12,323 |
+| Graph edges | 8,893 |
+| Total time | 7.4s |
+
+### ACAS
+
+| Metric | Value |
+| ------ | ----- |
+| Graph nodes | 14,016 |
+| Graph edges | 15,452 |
+| Total time | 9.3s |
+
+### Micro-Benchmark (Single-File Extraction)
+
+| Metric | Value |
+| ------ | ----- |
+| Per-iteration | 0.65ms |
+| Throughput | ~382K lines/sec |
+
+## Worker Pool Tuning
+
+### Sub-Batch Size
+
+The worker pool splits each worker's chunk into sub-batches to bound peak memory per `postMessage` serialization. COBOL repos use a smaller sub-batch size than the default:
+
+| Parameter             | Default     | COBOL Mode          |
+| --------------------- | ----------- | ------------------- |
+| Sub-batch size        | 1,500 files | 200 files           |
+| Per sub-batch timeout | 120s        | 120s (configurable) |
+
+**Why 200?** COBOL regex extraction + preprocessing takes ~1ms per file on average, but with COPY expansion and deep indexing the effective time is ~150ms per file. At sub-batch size 1500, that would be ~225s per sub-batch, exceeding the 120s timeout.
+
+COBOL mode is activated automatically when `GITNEXUS_COBOL_DIRS` is set:
+
+```typescript
+// From pipeline.ts
+const cobolSubBatch = process.env.GITNEXUS_COBOL_DIRS ? 200 : undefined;
+workerPool = createWorkerPool(workerUrl, undefined, cobolSubBatch);
+```
+
+### Worker Count
+
+Workers default to `min(8, cpus - 1)`. For COBOL repos, this is usually sufficient since regex extraction is CPU-bound but fast. The bottleneck is typically KuzuDB write, not extraction.
+
+### Timeout Configuration
+
+| Environment Variable                 | Default         | Purpose                                             |
+| ------------------------------------ | --------------- | --------------------------------------------------- |
+| `GITNEXUS_WORKER_TIMEOUT_MS`         | 120,000 (2 min) | Per sub-batch processing timeout                    |
+| `GITNEXUS_WORKER_STARTUP_TIMEOUT_MS` | 60,000 (1 min)  | Worker initialization timeout (tree-sitter loading) |
+
+For COBOL-only repos, worker startup is faster because tree-sitter native modules are loaded lazily (skipped entirely if only COBOL files are present).
+
+## Data Item Cap
+
+### Configuration
+
+```typescript
+const MAX_DATA_ITEMS_PER_FILE = 500;
+```
+
+This constant appears in both `parse-worker.ts` (worker path) and `parsing-processor.ts` (sequential fallback).
+
+### Rationale
+
+Some COBOL programs, especially after COPY expansion, can have 10,000+ data items. At that scale:
+
+- The in-memory relationship Map (for CONTAINS, REDEFINES, etc.) approaches the V8 16.7M entry limit across thousands of files
+- KuzuDB write time increases linearly with edge count
+- Most deep-nested items (level 20+) are rarely queried individually
+
+### Impact
+
+The cap truncates data items beyond the 500th in source order. Since 01-level Records appear first in COBOL source, the cap preserves:
+
+- All 01-level record definitions
+- The most important 02-49 level items (those closest to the record root)
+- 88-level conditions associated with early items
+
+To increase the cap for specific needs, modify the `MAX_DATA_ITEMS_PER_FILE` constant in both files.
+
+## Memory Management
+
+### COPY Expansion Breadth Guard
+
+A per-file `MAX_TOTAL_EXPANSIONS = 500` limit prevents exponential blowup from diamond-shaped COPY graphs (e.g., N copybooks each containing N COPY statements). Once the limit is reached, further COPY statements in that file are left unexpanded. See [copy-expansion.md](copy-expansion.md) for details.
+
+### COPY Expansion Memory
+
+All copybook content is loaded upfront into a Map before chunk processing begins. For PROJECT-NAME:
+
+- 2,976 copybooks, typically under 100MB total
+- The Map is shared (read-only) across chunk iterations
+- Per-chunk, the copybook map is merged with chunk file content (in case a chunk contains copybooks not in the pre-loaded set)
+- After all chunks are processed, the copybook map is freed (`cobolCopybookContents = undefined`)
+
+### Chunk Budget
+
+Source files are grouped into chunks of max 20MB (`CHUNK_BYTE_BUDGET`). Each chunk's lifecycle:
+
+1. Read file content into memory
+2. Expand COPY statements (mutates content in-place)
+3. Dispatch to workers for extraction
+4. Workers return serialized results
+5. Merge results into graph
+6. Chunk content goes out of scope (GC reclaims)
+
+This ensures only ~20MB of source + ~200-400MB of working memory (ASTs, extracted records, serialization) is active at any time.
+
+### Shared Warning Deduplication
+
+The `warnedCircular` set (used by the COPY expansion engine) is shared across all files in a chunk. This prevents the same circular copybook warning (e.g., `ANAZI includes itself`) from being logged thousands of times.
+
+## Known Limitations
+
+| Limitation                               | Impact                                                                                                          | Workaround                                                 |
+| ---------------------------------------- | --------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------- |
+| tree-sitter-cobol hangs on ~5% of files  | Cannot use tree-sitter for COBOL                                                                                | Regex-only extraction (current approach)                   |
+| Data item cap (500/file)                 | May miss deeply nested items in large programs                                                                  | Increase `MAX_DATA_ITEMS_PER_FILE` in source               |
+| Circular copybooks (ANAZI, ANDIP, QDIPE) | Self-referential includes cannot be expanded                                                                    | Detected and skipped with warning                          |
+| wfproc/ files may not be pure COBOL      | Workflow files may produce extraction noise                                                                     | Exclude `wfproc` from `GITNEXUS_COBOL_DIRS` if problematic |
+| No MOVE DATA_FLOW edges yet              | Data flow between variables not in graph                                                                        | Reserved for future release                                |
+| Continuation line handling               | Some complex multi-line continuations (especially in string literals spanning 3+ lines) may not merge correctly | Known edge case; affects <0.1% of lines                    |
+| Single-line EXEC blocks                  | `EXEC SQL SELECT ... END-EXEC` on one line is handled, but pathological nesting is not                          | Extremely rare in practice                                 |
+| Extension case sensitivity               | `.GNM` and `.gnm` are matched differently                                                                       | Use the exact case from the codebase                       |
+
+## Troubleshooting
+
+### "COPY expansion failed"
+
+```
+[pipeline] COPY expansion failed for s/BGTABFL: Cannot read properties of null
+```
+
+**Cause:** A copybook referenced by a COPY statement cannot be found.
+
+**Fix:**
+
+1. Verify `GITNEXUS_COBOL_DIRS` includes the directory containing copybooks (typically `c`)
+2. Check that copybook filenames match the COPY target (case-insensitive, after stripping extensions)
+3. Ensure copybook files are not in `.gitignore`
+
+### Worker sub-batch timeout
+
+```
+Worker 3 sub-batch timed out after 120s (chunk: 200 items)
+```
+
+**Cause:** A sub-batch took longer than the timeout. Typically happens when one file is extremely large (50,000+ lines after COPY expansion).
+
+**Fix:** Increase the timeout:
+
+```bash
+GITNEXUS_WORKER_TIMEOUT_MS=300000 gitnexus analyze
+```
+
+### Memory errors (heap out of memory)
+
+```
+FATAL ERROR: CALL_AND_RETRY_LAST Allocation failed - JavaScript heap out of memory
+```
+
+**Fix:** Increase Node.js heap size:
+
+```bash
+node --max-old-space-size=16384 /path/to/gitnexus/dist/cli/index.js analyze
+```
+
+For very large repos (>500MB source), consider `--max-old-space-size=32768`.
+
+### Concurrent analyze corruption
+
+**Rule:** Only ONE `gitnexus analyze` process should run at a time per repository. Concurrent writes to KuzuDB corrupt the database.
+
+If corruption occurs:
+
+```bash
+# Remove the KuzuDB directory and re-index
+rm -rf .gitnexus/kuzu
+gitnexus analyze --force
+```
+
+### Slow KuzuDB write phase
+
+The KuzuDB write phase (132s for PROJECT-NAME) is the bottleneck for large COBOL repos. This is proportional to the number of nodes and edges being written. Reducing `MAX_DATA_ITEMS_PER_FILE` or excluding non-essential directories from `GITNEXUS_COBOL_DIRS` can help.
+
+### Verbose output
+
+Enable verbose logging to see per-phase timing and statistics:
+
+```bash
+GITNEXUS_VERBOSE=1 gitnexus analyze
+```
+
+This outputs:
+
+- Scan statistics (paths, parseable files, chunk count)
+- Worker pool configuration (worker count, sub-batch size)
+- COPY expansion statistics (copybooks loaded, files expanded)
+- Community and process detection results
+- Contract detection results
+
+## Source Files
+
+- `gitnexus/src/core/ingestion/workers/worker-pool.ts` -- `DEFAULT_SUB_BATCH_SIZE`, `SUB_BATCH_TIMEOUT_MS`, `WORKER_STARTUP_TIMEOUT_MS`
+- `gitnexus/src/core/ingestion/pipeline.ts` -- `CHUNK_BYTE_BUDGET`, COBOL sub-batch configuration, chunk lifecycle
+- `gitnexus/src/core/ingestion/workers/parse-worker.ts` -- `MAX_DATA_ITEMS_PER_FILE`, `processCobolRegexOnly()`
+- `gitnexus/src/core/ingestion/parsing-processor.ts` -- Sequential fallback `MAX_DATA_ITEMS_PER_FILE`

docs/code-indexing/cobol/regex-extraction.md

@@ -0,0 +1,206 @@
+# COBOL Regex Extraction
+
+The `extractCobolSymbolsWithRegex()` function in `cobol-preprocessor.ts` performs single-pass, state-machine-driven extraction of all COBOL symbols. This document describes the state machine, line processing flow, and every regex pattern used.
+
+## State Machine: Division Tracking
+
+The extractor tracks which COBOL division is currently being processed. Division transitions are detected by the `RE_DIVISION` pattern.
+
+```mermaid
+stateDiagram-v2
+    [*] --> null : Start of file
+    null --> identification : IDENTIFICATION DIVISION
+    identification --> environment : ENVIRONMENT DIVISION
+    environment --> data : DATA DIVISION
+    data --> procedure : PROCEDURE DIVISION
+
+    note right of identification
+        Extracts: PROGRAM-ID, AUTHOR, DATE-WRITTEN
+    end note
+    note right of environment
+        Extracts: SELECT ... ASSIGN ... (file declarations)
+    end note
+    note right of data
+        Extracts: FD entries, data items (01-77, 88), COPY
+    end note
+    note right of procedure
+        Extracts: paragraphs, sections, PERFORM, CALL,
+        ENTRY, MOVE, EXEC SQL/CICS
+    end note
+```
+
+## State Machine: Data Section Tracking
+
+Within the DATA DIVISION, a secondary state machine tracks the current section to tag data items with their origin.
+
+```mermaid
+stateDiagram-v2
+    [*] --> unknown : DATA DIVISION entered
+    unknown --> working_storage : WORKING-STORAGE SECTION
+    unknown --> linkage : LINKAGE SECTION
+    unknown --> file : FILE SECTION
+    unknown --> local_storage : LOCAL-STORAGE SECTION
+    working_storage --> linkage : LINKAGE SECTION
+    working_storage --> file : FILE SECTION
+    linkage --> working_storage : WORKING-STORAGE SECTION
+    file --> working_storage : WORKING-STORAGE SECTION
+    file --> linkage : LINKAGE SECTION
+    local_storage --> working_storage : WORKING-STORAGE SECTION
+```
+
+Within the ENVIRONMENT DIVISION, the `currentEnvSection` tracks whether we are in `INPUT-OUTPUT` or `CONFIGURATION` section. SELECT statement accumulation only occurs in `INPUT-OUTPUT`.
+
+## Line Processing Flow
+
+Each raw source line goes through this pipeline:
+
+```
+Raw line
+  |
+  v
+Length < 7? ---------> Skip (flush pending if any)
+  |
+  v
+Indicator col 7
+  |
+  +-- '*' or '/' -----> Comment: skip entirely
+  |
+  +-- '-' ------------> Continuation: append to pending line
+  |
+  +-- other ----------> Normal: flush pending, strip inline comments (|),
+                        buffer as new pending logical line
+```
+
+After all lines are processed, the final pending line is flushed, along with any accumulated SELECT statement, SORT/MERGE accumulator, and any open EXEC block (truncated file without `END-EXEC`).
+
+### Inline Comment Stripping
+
+Enterprise COBOL (particularly Italian dialect) uses the pipe character `|` as an inline comment marker. The `stripInlineComment()` helper is **quote-aware**: it tracks whether the scan position is inside a single- or double-quoted string and only treats `|` as a comment marker when outside quotes. Pipe characters inside string literals are preserved.
+
+Free-format `*>` inline comment stripping uses the same quote-aware approach: the scanner walks character by character, toggling quote state, and only recognizes `*>` as a comment marker when not inside a quoted string.
+
+### Patch Marker Handling
+
+The `preprocessCobolSource()` function (run before extraction in the worker) replaces non-standard content in columns 1-6. Standard COBOL expects spaces or digit sequence numbers in this area. If any letter or `#` character is found, the entire sequence area is replaced with 6 spaces:
+
+```
+Before: mzADD MOVE WK-AMT TO WK-TOTAL
+After:        MOVE WK-AMT TO WK-TOTAL
+```
+
+This preserves exact line count for position mapping.
+
+## Regex Pattern Reference
+
+All patterns are compiled once as module-level constants and reused across calls.
+
+### Division and Section Detection
+
+| Constant | Pattern | Purpose | Example Match |
+|----------|---------|---------|---------------|
+| `RE_DIVISION` | `\b(IDENTIFICATION\|ENVIRONMENT\|DATA\|PROCEDURE)\s+DIVISION\b` | Division boundary | `PROCEDURE DIVISION` |
+| `RE_SECTION` | `\b(WORKING-STORAGE\|LINKAGE\|FILE\|LOCAL-STORAGE\|INPUT-OUTPUT\|CONFIGURATION)\s+SECTION\b` | Section boundary | `WORKING-STORAGE SECTION` |
+
+### IDENTIFICATION DIVISION
+
+| Constant | Pattern | Purpose | Example Match |
+|----------|---------|---------|---------------|
+| `RE_PROGRAM_ID` | `\bPROGRAM-ID\.\s*([A-Z][A-Z0-9-]*)` | Program name | `PROGRAM-ID. BGTABFL` |
+| `RE_AUTHOR` | `^\s+AUTHOR\.\s*(.+)` | Author metadata | `AUTHOR. D. Smith` |
+| `RE_DATE_WRITTEN` | `^\s+DATE-WRITTEN\.\s*(.+)` | Date metadata | `DATE-WRITTEN. 2024-01-15` |
+
+### ENVIRONMENT DIVISION
+
+| Constant | Pattern | Purpose | Example Match |
+|----------|---------|---------|---------------|
+| `RE_SELECT_START` | `\bSELECT\s+(?:OPTIONAL\s+)?([A-Z][A-Z0-9-]+)` | File SELECT start (with optional `SELECT OPTIONAL` support) | `SELECT MASTER-FILE`, `SELECT OPTIONAL TRANS-FILE` |
+
+SELECT statements are accumulated across multiple lines until a period terminator is found, then parsed for ASSIGN, ORGANIZATION, ACCESS, RECORD KEY, and FILE STATUS clauses.
+
+### DATA DIVISION
+
+| Constant | Pattern | Purpose | Example Match |
+|----------|---------|---------|---------------|
+| `RE_FD` | `^\s+FD\s+([A-Z][A-Z0-9-]+)` | File description | `FD MASTER-FILE` |
+| `RE_DATA_ITEM` | `^\s+(\d{1,2})\s+([A-Z][A-Z0-9-]+)\s*(.*)` | Data item (01-77) | `05 WK-NAME PIC X(30)` |
+| `RE_ANONYMOUS_REDEFINES` | `^\s+(\d{1,2})\s+REDEFINES\s+([A-Z][A-Z0-9-]+)` | Anonymous REDEFINES | `01 REDEFINES WK-REC` |
+| `RE_88_LEVEL` | `^\s+88\s+([A-Z][A-Z0-9-]+)\s+VALUES?\s+(?:ARE\s+)?(.+)` | Condition name | `88 WK-ACTIVE VALUE "Y"` |
+
+The trailing clauses of `RE_DATA_ITEM` are parsed by `parseDataItemClauses()` for PIC, USAGE, OCCURS, and REDEFINES.
+
+### PROCEDURE DIVISION
+
+| Constant | Pattern | Purpose | Example Match |
+|----------|---------|---------|---------------|
+| `RE_PROC_SECTION` | `^       ([A-Z][A-Z0-9-]+)\s+SECTION\.\s*$` | Procedure section header | `       MAIN-LOGIC SECTION.` |
+| `RE_PROC_PARAGRAPH` | `^       ([A-Z][A-Z0-9-]+)\.\s*$` | Paragraph header | `       PROCESS-RECORD.` |
+| `RE_PERFORM` | `\bPERFORM\s+([A-Z][A-Z0-9-]+)(?:\s+THRU\s+([A-Z][A-Z0-9-]+))?` | PERFORM call | `PERFORM CALC-TAX THRU CALC-TAX-EXIT` |
+| `RE_PROC_USING` | `\bPROCEDURE\s+DIVISION\s+USING\s+([\s\S]*?)(?:\.\|$)` | USING parameters | `PROCEDURE DIVISION USING WK-PARAM` |
+| `RE_ENTRY` | `\bENTRY\s+"([^"]+)"(?:\s+USING\s+([\s\S]*?))?(?:\.\|$)` | ENTRY point | `ENTRY "SUBPROG" USING WK-DATA` |
+| `RE_MOVE` | `\bMOVE\s+((?:CORRESPONDING\|CORR)\s+)?([A-Z][A-Z0-9-]+)\s+TO\s+(.+)` | MOVE statement (supports CORR abbreviation and multi-target) | `MOVE WK-NAME TO OUT-NAME`, `MOVE CORR WK-IN TO WK-OUT` |
+
+The USING parameter list (`RE_PROC_USING`) is split on `\bRETURNING\b` before tokenization -- any RETURNING clause and everything after it is excluded from the parameter list (`.split(/\bRETURNING\b/i)[0]`).
+
+Note: `RE_PROC_SECTION` and `RE_PROC_PARAGRAPH` require exactly 7 spaces of leading indentation (COBOL area A starting at column 8). This is the standard COBOL paragraph indentation.
+
+### All-Division Patterns
+
+These patterns are checked regardless of current division:
+
+| Constant | Pattern | Purpose | Example Match |
+|----------|---------|---------|---------------|
+| `RE_CALL` | `\bCALL\s+"([^"]+)"` | External program call | `CALL "BGTABUP"` |
+| `RE_COPY_UNQUOTED` | `\bCOPY\s+([A-Z][A-Z0-9-]+)(?:\s\|\.)` | COPY (unquoted) | `COPY CPSESP.` |
+| `RE_COPY_QUOTED` | `\bCOPY\s+"([^"]+)"(?:\s\|\.)` | COPY (quoted) | `COPY "WORKGRID.CPY".` |
+
+### SORT/MERGE Support
+
+| Constant | Purpose |
+|----------|---------|
+| `SORT_CLAUSE_NOISE` | Set of SORT/MERGE clause keywords filtered from USING/GIVING file lists: `ON`, `ASCENDING`, `DESCENDING`, `KEY`, `WITH`, `DUPLICATES`, `IN`, `ORDER`, `COLLATING`, `SEQUENCE`, `IS`, `THROUGH`, `THRU`, `INPUT`, `OUTPUT`, `PROCEDURE` |
+
+SORT and MERGE statements are accumulated across multiple lines (like SELECT) until a period terminator is found, then parsed for USING/GIVING file lists and INPUT/OUTPUT PROCEDURE targets. The `flushSort()` helper encapsulates the flush-and-parse logic, mirroring the existing `flushSelect()` pattern. Both helpers are called at EOF to handle truncated files.
+
+### GO TO Multi-Target
+
+`RE_GOTO` captures all paragraph names in a `GO TO` statement, including the multi-target form `GO TO p1 p2 p3 DEPENDING ON x`. The captured group contains all target names (space-separated), which are split into individual targets. Each target produces a separate `gotos` entry.
+
+### PROGRAM-ID Detection
+
+PROGRAM-ID is detected regardless of the current division state. This handles sibling programs that appear after `END PROGRAM` and omit the `IDENTIFICATION DIVISION` header -- the extractor will still capture the PROGRAM-ID and push a new program boundary.
+
+### EXEC Block Patterns
+
+| Constant | Pattern | Purpose | Example Match |
+|----------|---------|---------|---------------|
+| `RE_EXEC_SQL_START` | `\bEXEC\s+SQL\b` | Start of EXEC SQL block | `EXEC SQL` |
+| `RE_EXEC_CICS_START` | `\bEXEC\s+CICS\b` | Start of EXEC CICS block | `EXEC CICS` |
+| `RE_END_EXEC` | `\bEND-EXEC\b` | End of EXEC block | `END-EXEC` |
+
+EXEC blocks accumulate all lines between `EXEC SQL/CICS` and `END-EXEC`, then delegate to `parseExecSqlBlock()` or `parseExecCicsBlock()` for detailed extraction.
+
+## Excluded Paragraph Names
+
+The following names are excluded from paragraph detection to avoid false positives from division/section headers:
+
+```
+DECLARATIVES, END, PROCEDURE, IDENTIFICATION,
+ENVIRONMENT, DATA, WORKING-STORAGE, LINKAGE,
+FILE, LOCAL-STORAGE, COMMUNICATION, REPORT,
+SCREEN, INPUT-OUTPUT, CONFIGURATION
+```
+
+Additionally, paragraph candidates containing `DIVISION` or `SECTION` as substrings are excluded.
+
+## MOVE Skip List (Figurative Constants)
+
+MOVE statements where the source is a figurative constant are skipped:
+
+```
+SPACES, ZEROS, ZEROES, LOW-VALUES, LOW-VALUE,
+HIGH-VALUES, HIGH-VALUE, QUOTES, QUOTE, ALL
+```
+
+## Source Files
+
+- `gitnexus/src/core/ingestion/cobol-preprocessor.ts` -- `preprocessCobolSource()`, `extractCobolSymbolsWithRegex()`, all regex constants

docs/plans/2026-03-26-feat-cobol-full-language-coverage-plan.md

@@ -0,0 +1,326 @@
+---
+title: "feat: Complete COBOL language feature coverage for maximum knowledge graph value"
+type: feat
+status: active
+date: 2026-03-26
+origin: Feature audit from v3-integration-architect agent (session 8642401e)
+---
+
+## Enhancement Summary
+
+**Deepened on:** 2026-03-26
+**Research agents used:** COBOL expert (Phase 1+2), graph value analyst, codebase explorer
+**Sections enhanced:** Phase 1 (5 features), Phase 2 (4 features), graph value ranking
+
+### Key Improvements from Research
+1. **CALL USING** is the #1 highest-value edge type (9.2/10) — fixes ~40% of missing caller references
+2. **EXEC DLI** requires dual-interface support (EXEC DLI + CBLTDLI CALL) for full IMS coverage
+3. **DECLARATIVES** is lowest-risk Phase 2 item — existing section/paragraph detection already captures structure
+4. **SET TO TRUE** accounts for 80-90% of all SET statements — prioritize this form
+5. **INSPECT** needs multi-line accumulator (like SORT) — can span 5+ continuation lines
+6. **Graph value ranking**: cobol-call-using (9.2) > cobol-error-handler (9.0) > dli-gu (8.2) > cobol-string (6.2)
+
+### New Edge Cases Discovered
+- CALL USING supports mixed modes: `USING BY REFERENCE WS-A BY CONTENT WS-B BY VALUE WS-C`
+- CALL USING `ADDRESS OF` and `OMITTED` must be filtered from parameter lists
+- EXEC DLI can have multiple SEGMENT levels in hierarchical retrieval (use matchAll)
+- DECLARATIVES can have multiple USE sections (one per file + catch-all for INPUT/OUTPUT/I-O/EXTEND)
+- INSPECT TALLYING can have multiple counters in a single statement
+- STRING/UNSTRING can span multiple lines (need accumulator pattern)
+
+---
+
+# Complete COBOL Language Feature Coverage
+
+## Overview
+
+Implement the remaining 25 unhandled COBOL language features and fix 10 partial features to achieve ~95% coverage (up from 71.9%). The goal is to build the richest possible knowledge graph from COBOL codebases, enabling a future `modernize` MCP command (out of scope for this plan) that would use the graph to assist with COBOL-to-modern-language migration.
+
+## Problem Statement
+
+The COBOL processor currently handles 54 of 89 applicable language features (71.9%). The 25 unhandled features represent real data loss in the knowledge graph:
+- **Cross-program data flow** is invisible (CALL ... USING parameters not extracted)
+- **IMS/DB programs** produce empty graphs (EXEC DLI not recognized)
+- **String transformation logic** is invisible (STRING/UNSTRING/INSPECT not tracked)
+- **SQL copybook dependencies** are missing (EXEC SQL INCLUDE not mapped)
+- **Error handling flows** are lost (DECLARATIVES/USE AFTER not captured)
+
+## Proposed Solution
+
+Implement features in 4 phases, ordered by graph value density (edges created per LOC of implementation). Each phase is independently shippable and testable.
+
+## Technical Approach
+
+### Phase 1: High-Value Data Flow Edges (~150 LOC, ~8 new edge types)
+
+The highest-ROI features: they create new ACCESSES and IMPORTS edges that directly improve impact analysis.
+
+**Critical research finding**: Multi-line statement accumulation is the dominant challenge. CALL USING, STRING/UNSTRING, and multi-line data item clauses all span multiple lines in production COBOL. The free-format path processes each line independently — these features need statement accumulators (like SORT/SELECT) or the free-format path needs multi-line awareness. Estimated LOC increased from 110 to 150 to account for accumulator infrastructure.
+
+#### 1.1 EXEC SQL INCLUDE -> IMPORTS edges
+- **File:** `cobol-preprocessor.ts` (parseExecSqlBlock)
+- **What:** Detect `INCLUDE` as the operation, extract member name, emit as a `copies[]` entry
+- **Graph:** IMPORTS edge from File to included copybook/SQLCA with reason `sql-include`
+- **Tests:** Unit test for `EXEC SQL INCLUDE SQLCA END-EXEC` and `EXEC SQL INCLUDE CUSTCOPY END-EXEC`
+
+**Research insights (EXEC SQL INCLUDE):**
+- DB2 member names can contain underscores: `EXEC SQL INCLUDE CUST_TBL_DCL END-EXEC` — regex must use `[A-Z][A-Z0-9_-]+`
+- Quoted literal form: `EXEC SQL INCLUDE 'DBRMLIB.MEMBER' END-EXEC` (z/OS PDS qualified name)
+- SQLCA/SQLDA are DB2 builtins — won't resolve to repo files. Emit unresolved IMPORTS edge (still valuable)
+- No REPLACING support on EXEC SQL INCLUDE (unlike COPY)
+- Add `INCLUDE` to `OP_MAP` in `parseExecSqlBlock`; extract member via `RE_SQL_INCLUDE = /^INCLUDE\s+(?:'([^']+)'|"([^"]+)"|([A-Z][A-Z0-9_-]+))/i`
+
+#### 1.2 CALL ... USING parameter extraction -> ACCESSES edges (Graph value: 9.2/10)
+- **File:** `cobol-preprocessor.ts` (processLogicalLine CALL section)
+- **What:** After capturing CALL target, scan for USING clause. Extract parameter names (reuse USING_KEYWORDS filter). Store as `calls[].parameters: string[]`
+- **Interface:** Add `parameters?: string[]` to calls array type in CobolRegexResults
+- **File:** `cobol-processor.ts` (CALL edge block)
+- **Graph:** For each USING parameter, create ACCESSES edge from caller to data item Property node with reason `cobol-call-using`
+- **Tests:** `CALL 'AUDITLOG' USING CUST-ID WS-AMOUNT` -> 2 ACCESSES edges
+
+**Research insights (CALL USING forms):**
+- Mixed modes: `CALL 'PGM' USING BY REFERENCE WS-A BY CONTENT WS-B BY VALUE WS-C`
+- Pointer passing: `CALL 'PGM' USING ADDRESS OF WS-A`
+- Placeholder: `CALL 'PGM' USING OMITTED WS-B`
+- Filter keywords: add `ADDRESS`, `OMITTED`, `LENGTH` to USING_KEYWORDS (already has BY/VALUE/REFERENCE/CONTENT)
+- **Impact tool enhancement:** CALL-USING edges enable BFS traversal through parameter data flow — single most impactful edge type for COBOL impact analysis
+
+#### 1.3 STRING/UNSTRING data flow -> ACCESSES edges
+- **File:** `cobol-preprocessor.ts` (new section in extractProcedure)
+- **What:** Accumulate multi-line STRING/UNSTRING until period or END-STRING/END-UNSTRING. Extract sources and INTO targets.
+- **Interface:** Add `strings: Array<{ sources: string[]; target: string; type: 'string' | 'unstring'; line: number; caller: string | null }>` to CobolRegexResults
+- **Graph:** read-ACCESSES on sources, write-ACCESSES on INTO target with reason `cobol-string-read` / `cobol-string-write`
+- **Tests:** 2 unit tests + integration test assertions
+
+**Research insights (STRING/UNSTRING):**
+- **Needs statement accumulator** — STRING/UNSTRING always span multiple lines in production
+- Terminate accumulation at: period, END-STRING/END-UNSTRING, or start of next COBOL verb
+- STRING sources: identifiers before each `DELIMITED BY`. Filter: STRING, DELIMITED, BY, SIZE, ALL, INTO, WITH, POINTER, ON, OVERFLOW, NOT, END-STRING
+- UNSTRING: source is first identifier after UNSTRING; INTO targets are identifiers after INTO. Filter: DELIMITER, IN, COUNT, TALLYING, OR
+- WITH POINTER field is both read AND written (starting position updated)
+- TALLYING IN / COUNT IN fields are write targets
+- Literal sources (`'text'`) must be filtered — quote-aware tokenization needed
+- **Edge case**: STRING terminated by next verb, not period — existing fixture has `STRING ... DISPLAY` without period between them
+
+#### 1.4 OCCURS DEPENDING ON -> ACCESSES edge
+- **File:** `cobol-preprocessor.ts` (parseDataItemClauses)
+- **What:** Extend OCCURS regex to capture DEPENDING ON field, KEY fields, and INDEXED BY names
+- **Interface:** Add `dependingOn?: string`, `occursMax?: number`, `occursKeys?: Array<{direction: string; fields: string[]}>`, `indexedBy?: string[]` to data items
+- **Graph:** ACCESSES edge from table item to controlling field with reason `cobol-depends-on`
+- **Tests:** `05 WS-TABLE OCCURS 100 DEPENDING ON WS-COUNT` -> edge
+
+**Research insights (OCCURS):**
+- IBM allows `OCCURS 0 TO n DEPENDING ON` (zero minimum) and `OCCURS UNBOUNDED DEPENDING ON` (V6.4)
+- Subscripted controlling fields: `DEPENDING ON WS-COUNT(WS-IDX)` — strip subscripts before storing
+- **Pre-existing gap**: Multi-line data item clauses without continuation indicator are NOT captured. `05 WS-TABLE\n    OCCURS 100\n    DEPENDING ON WS-COUNT.` — the current RE_DATA_ITEM only gets the first line, `rest` is empty. Fixing properly requires a data item accumulator (like SELECT). **Defer full fix to Phase 3; implement same-line capture now.**
+- KEY IS fields: `ASCENDING KEY IS WS-KEY-1 WS-KEY-2` — capture for SEARCH ALL resolution
+- INDEXED BY: `INDEXED BY IDX-1 IDX-2` — capture for SET/SEARCH context
+
+#### 1.5 VALUE clause for standard data items
+- **File:** `cobol-preprocessor.ts` (parseDataItemClauses)
+- **What:** Extract VALUE using a pragmatic function that handles quoted strings, numerics, figurative constants, hex/national literals
+- **Interface:** Already exists as `values?: string[]` on data items (currently only populated for 88-level)
+- **Graph:** Stored in Property node description (no new edges)
+- **Tests:** `01 WS-STATUS PIC X VALUE 'A'` -> values: ['A']
+
+**Research insights (VALUE forms):**
+- Hex literals: `VALUE X'F1F2F3F4'`, National: `VALUE N'text'`, DBCS: `VALUE G'text'`
+- Figurative constants: SPACES, ZEROS, ZEROES, LOW-VALUES, HIGH-VALUES, QUOTES, NULL, NULLS
+- ALL literal: `VALUE ALL '*'`
+- Numeric with sign/decimal: `VALUE -123.45`, `VALUE +1`
+- `VALUE IS` optional — both `VALUE 'A'` and `VALUE IS 'A'` valid
+- **Decimal vs period ambiguity**: `VALUE 100.` — is `.` decimal or terminator? `parseDataItemClauses` already strips trailing period, so this is handled
+- IBM V6.4: floating-point `VALUE 1.0E5` — extend numeric regex if needed
+- Implementation: use a pragmatic `extractValue(rest)` function, not a single complex regex
+
+### Phase 2: EXEC DLI + DECLARATIVES (~90 LOC, ~4 new edge types)
+
+IMS/DB support and error handling flows.
+
+#### 2.1 EXEC DLI (IMS/DB) -> ACCESSES edges (Graph value: 8.2/10)
+- **File:** `cobol-preprocessor.ts` (processLogicalLine — add RE_EXEC_DLI_START check alongside SQL/CICS)
+- **What:** Accumulate EXEC DLI blocks like EXEC SQL. Parse DLI verbs (GU, GN, GNP, GHU, GHN, GHNP, ISRT, DLET, REPL, CHKP, SCHD, TERM). Extract segment name, PCB number, INTO/FROM areas, WHERE fields, PSB name.
+- **Interface:** Add `execDliBlocks: Array<{ line: number; verb: string; pcbNumber?: number; segmentName?: string; intoField?: string; fromField?: string; whereField?: string; psbName?: string }>` to CobolRegexResults
+- **Graph:** CodeElement node + ACCESSES edge to `<ims>:<segmentName>` Record node with reason `dli-{verb}`; ACCESSES edges to INTO/FROM data areas; PSB ACCESSES for SCHD
+- **Tests:** `EXEC DLI GU USING PCB(1) SEGMENT(CUSTOMER) INTO(WS-CUST) END-EXEC`
+
+**Research insights (dual IMS interface):**
+- **EXEC DLI**: Embedded command interface for CICS-DL/I programs only
+- **CBLTDLI CALL**: Batch interface via `CALL 'CBLTDLI' USING function-code PCB io-area SSA1..SSA15`
+- CBLTDLI is already captured as a CALL to 'CBLTDLI' — enrich with USING parameter semantics later
+- Multiple SEGMENT levels in hierarchical retrieval — use `matchAll` on segment regex
+- DLI verbs: GU (most common), GN, GNP, GHU, GHN, GHNP, ISRT, REPL, DLET, CHKP, SCHD, TERM, ROLL, ROLB
+- **Edge case**: DLET/REPL have no SEGMENT clause (operate on current position)
+- **Recommended order**: Implement AFTER DECLARATIVES and SET (lower risk, higher frequency)
+
+#### 2.2 DECLARATIVES / USE AFTER STANDARD EXCEPTION (Graph value: 9.0/10)
+- **File:** `cobol-preprocessor.ts` (processLogicalLine — detect DECLARATIVES keyword, track USE AFTER blocks)
+- **What:** When `DECLARATIVES.` is encountered, switch to declaratives mode. Extract USE statements binding sections to files/modes.
+- **Interface:** Add `declaratives: Array<{ sectionName: string; useType: 'error' | 'debug' | 'label' | 'reporting'; target: string; line: number }>` to CobolRegexResults
+- **Graph:** ACCESSES edge from declarative Namespace to file Record with reason `cobol-declarative-error-handler`
+- **Tests:** Unit test with DECLARATIVES section, integration test for error flow
+
+**Research insights (DECLARATIVES syntax):**
+- `USE AFTER STANDARD {EXCEPTION|ERROR} ON {file-name|INPUT|OUTPUT|I-O|EXTEND}`
+- EXCEPTION and ERROR are synonymous; STANDARD is optional in IBM dialects
+- Multiple USE sections allowed (one per file + catch-all for I/O modes)
+- `END DECLARATIVES.` must NOT reset PROCEDURE DIVISION state
+- `DECLARATIVES` is already in EXCLUDED_PARA_NAMES — no false paragraph risk
+- Existing section/paragraph detection already captures structural elements — just need USE binding
+- **Lowest risk Phase 2 item** — implement first
+
+#### 2.3 SET statement -> ACCESSES edges
+- **File:** `cobol-preprocessor.ts` (extractProcedure — new RE_SET regex)
+- **Interface:** Add `sets: Array<{ targets: string[]; form: 'to-true'|'to-value'|'up-by'|'down-by'|'address-of'|'to-null'|'to-entry'; value?: string; entryTarget?: string; entryIsLiteral?: boolean; line: number; caller: string | null }>` to CobolRegexResults
+- **Graph:** ACCESSES write edge with reason `cobol-set-condition` (TO TRUE), `cobol-set-index` (TO/UP/DOWN), `cobol-set-address` (ADDRESS OF). SET ENTRY with literal -> CALLS edge.
+- **Tests:** `SET WS-EOF TO TRUE`, `SET IDX-1 TO 5`, `SET IDX-1 UP BY 1`
+
+**Research insights (SET forms by frequency):**
+- `SET condition TO TRUE` — 80-90% of all SET usage. Multiple targets: `SET COND-A COND-B TO TRUE`
+- `SET index TO/UP BY/DOWN BY` — ~8%. Multiple indices: `SET IDX-1 IDX-2 UP BY 1`
+- `SET pointer TO ADDRESS OF data-item` / `SET ADDRESS OF data-item TO pointer` — ~2%
+- `SET proc-ptr TO ENTRY "PROGNAME"` — rare but creates CALLS edge (like dynamic CALL)
+- Filter OF/IN qualifiers: `SET COND-A OF WS-RECORD TO TRUE` (strip OF WS-RECORD)
+- **Prioritize**: SET TO TRUE alone covers 80-90% — implement this form first
+
+#### 2.4 INSPECT -> ACCESSES edges
+- **File:** `cobol-preprocessor.ts` (extractProcedure — new `inspectAccum` accumulator like SORT)
+- **What:** Accumulate multi-line INSPECT until period. Extract inspected field + tally counters.
+- **Interface:** Add `inspects: Array<{ inspectedField: string; counters: string[]; form: 'tallying'|'replacing'|'converting'|'tallying-replacing'; line: number; caller: string | null }>` to CobolRegexResults
+- **Graph:** ACCESSES read on inspected field always; write if REPLACING/CONVERTING. Write edges for tally counters. Reason: `cobol-inspect-read`/`cobol-inspect-write`/`cobol-inspect-tally`
+- **Tests:** `INSPECT WS-FIELD TALLYING WS-COUNT FOR ALL 'A'` -> read on WS-FIELD, write on WS-COUNT
+
+**Research insights (INSPECT forms by frequency):**
+- REPLACING (~60%): `INSPECT WS-STR REPLACING ALL 'A' BY 'B'`
+- TALLYING (~25%): `INSPECT WS-STR TALLYING WS-CNT FOR ALL 'A'` — multiple counters possible
+- CONVERTING (~10%): `INSPECT WS-STR CONVERTING 'abc' TO 'ABC'`
+- Combined (~5%): TALLYING + REPLACING in single statement
+- **Needs multi-line accumulator** — INSPECT frequently spans 3-5 lines in production
+- Extract tally counters with `([A-Z][A-Z0-9-]+)\s+FOR\b` matchAll pattern
+- Filter figurative constants (SPACES, ZEROS) using existing MOVE_SKIP set
+
+### Phase 3: Completeness Fixes (~60 LOC)
+
+Fix the 10 partial features and small gaps.
+
+#### 3.1 CALL ... RETURNING extraction
+- Extend RE_CALL processing to capture RETURNING target after the USING clause
+- Store as `calls[].returning?: string`
+- Graph: ACCESSES write edge with reason `cobol-call-returning`
+
+#### 3.2 SELECT OPTIONAL flag preservation
+- Store `isOptional: boolean` in FileDeclaration interface
+- Include in Record node description
+
+#### 3.3 ALTERNATE RECORD KEY extraction
+- Add regex in parseSelectStatement: `/\bALTERNATE\s+RECORD\s+KEY\s+(?:IS\s+)?([A-Z][A-Z0-9-]+)/i`
+- Store as `alternateKeys?: string[]`
+
+#### 3.4 COMMON attribute on nested programs
+- Extend RE_PROGRAM_ID: `/\bPROGRAM-ID\.\s*([A-Z][A-Z0-9-]+)(?:\s+IS\s+COMMON)?/i`
+- Store `isCommon: boolean` on Module node
+- Affects cross-program CALL resolution scope
+
+#### 3.5 IS EXTERNAL / IS GLOBAL as first-class properties
+- Change from usage string hack to proper boolean fields on data items
+- Add `isExternal?: boolean`, `isGlobal?: boolean` to data item interface
+
+#### 3.6 AUTHOR / DATE-WRITTEN mapped to Module node
+- Already extracted as programMetadata — map to Module node properties
+- `graph.addNode({ ..., properties: { ..., author, dateWritten } })`
+
+#### 3.7 REPLACE statement
+- Track REPLACE / REPLACE OFF state in preprocessor
+- Apply text substitutions during preprocessing (before regex extraction)
+- Complex: requires careful scoping rules
+
+### Phase 4: Niche Features (~30 LOC)
+
+Low-priority but nice for completeness.
+
+#### 4.1 INITIALIZE statement -> write ACCESSES
+- `/\bINITIALIZE\s+([A-Z][A-Z0-9-]+)/i`
+- ACCESSES write edge with reason `cobol-initialize`
+
+#### 4.2 Remaining IDENTIFICATION DIVISION paragraphs
+- DATE-COMPILED, INSTALLATION, SECURITY, REMARKS
+- Map to Module node description properties
+
+#### 4.3 EXEC SQL INCLUDE -> IMPORTS edge (expansion)
+- For EXEC SQL INCLUDE inside EXEC blocks that reference copybooks containing SQL
+- Create IMPORTS edge similar to COPY
+
+## Acceptance Criteria
+
+### Functional Requirements
+
+- [ ] Phase 1: All 5 features implemented with unit + integration tests
+- [ ] Phase 2: All 4 features implemented with unit + integration tests
+- [ ] Phase 3: All 7 partial features fixed
+- [ ] Phase 4: At least 2 of 3 niche features implemented
+- [ ] All existing 145 tests continue to pass
+- [ ] TypeScript compiles cleanly
+
+### Non-Functional Requirements
+
+- [ ] No performance regression: CardDemo benchmark stays under 8s
+- [ ] No file exceeds 1500 LOC (preprocessor currently 1326)
+- [ ] ACAS benchmark shows increased node/edge counts (more data extracted)
+- [ ] CardDemo benchmark shows increased edge counts (CALL USING, STRING, etc.)
+
+### Quality Gates
+
+- [ ] Each phase has its own commit
+- [ ] Integration test assertions updated with exact counts per phase
+- [ ] Benchmark run after each phase to track graph growth
+
+## Dependencies & Risks
+
+### Dependencies
+- None. All changes are additive to existing COBOL processor code.
+- No LanguageProvider changes needed.
+- No graph schema changes needed (all new constructs map to existing node labels + edge types).
+
+### Risks
+- **preprocessor.ts size**: Currently 1326 LOC. Phase 1+2 adds ~200 LOC -> 1526 LOC. May need to extract helpers into a separate `cobol-data-flow.ts` module if it exceeds 1500.
+- **REPLACE statement** (Phase 3.7) is the most complex feature — requires tracking text substitution state across logical lines. Consider deferring to a separate PR if it takes >100 LOC.
+- **EXEC DLI** (Phase 2.1) is only testable against IMS codebases. Need fixture data or synthetic test cases.
+
+## Graph Value Ranking by MCP Tool Impact
+
+Research agent analyzed all 5 MCP tools (query, context, impact, detect_changes, rename) against planned edge types:
+
+| Edge Type | QUERY | CONTEXT | IMPACT | DETECT | RENAME | **Overall** |
+|-----------|-------|---------|--------|--------|--------|-------------|
+| `cobol-call-using` | 4/5 | 5/5 | 5/5 | 4/5 | 4/5 | **9.2/10** |
+| `cobol-error-handler` | 5/5 | 4/5 | 5/5 | 5/5 | 2/5 | **9.0/10** |
+| `dli-*` (IMS verbs) | 4/5 | 4/5 | 5/5 | 4/5 | 2/5 | **8.2/10** |
+| `cobol-string-*` | 4/5 | 3/5 | 3/5 | 3/5 | 2/5 | **6.2/10** |
+
+**Key finding**: `cobol-call-using` alone would fix ~40% of missing caller references in COBOL graphs.
+
+## Future Considerations
+
+This plan provides the graph data foundation for a future `modernize` MCP command (out of scope) that would:
+- Use CALL USING edges to map data contracts between programs
+- Use STRING/UNSTRING edges to identify data transformation logic
+- Use EXEC SQL/DLI edges to map database access patterns
+- Use DECLARATIVES to understand error handling architecture
+- Use the complete knowledge graph to generate migration plans
+
+**MCP tool enhancements needed** (after this plan ships):
+- Add `cobol-call-using`, `cobol-error-handler`, `dli-*` to IMPACT tool's default `relationTypes` for COBOL repos
+- Add confidence floors for new edge types in `IMPACT_RELATION_CONFIDENCE`
+- Register new edge types in `VALID_RELATION_TYPES` set (`local-backend.ts:52`)
+
+## Sources & References
+
+### Internal References
+- Feature audit: session 8642401e (COBOL expert agent, 123 features audited)
+- Prior plans: `docs/plans/2026-03-25-feat-cobol-100-percent-feature-coverage-plan.md`
+- Architecture: `docs/code-indexing/cobol/` (7 documentation files)
+
+### External References
+- COBOL features reference: mainframestechhelp.com/tutorials/cobol/features.htm
+- COBOL-85 standard: ISO/IEC 1989:1985
+- IBM Enterprise COBOL reference

eslint.config.js

@@ -1,28 +0,0 @@
-import js from '@eslint/js'
-import globals from 'globals'
-import reactHooks from 'eslint-plugin-react-hooks'
-import reactRefresh from 'eslint-plugin-react-refresh'
-import tseslint from 'typescript-eslint'
-
-export default tseslint.config(
-  { ignores: ['dist'] },
-  {
-    extends: [js.configs.recommended, ...tseslint.configs.recommended],
-    files: ['**/*.{ts,tsx}'],
-    languageOptions: {
-      ecmaVersion: 2020,
-      globals: globals.browser,
-    },
-    plugins: {
-      'react-hooks': reactHooks,
-      'react-refresh': reactRefresh,
-    },
-    rules: {
-      ...reactHooks.configs.recommended.rules,
-      'react-refresh/only-export-components': [
-        'warn',
-        { allowConstantExport: true },
-      ],
-    },
-  },
-)

eslint.config.mjs

@@ -0,0 +1,83 @@
+import tsPlugin from '@typescript-eslint/eslint-plugin';
+import tsParser from '@typescript-eslint/parser';
+import unusedImports from 'eslint-plugin-unused-imports';
+import reactHooks from 'eslint-plugin-react-hooks';
+import prettierConfig from 'eslint-config-prettier';
+
+export default [
+  // Global ignores
+  {
+    ignores: [
+      '**/dist/**',
+      '**/node_modules/**',
+      '**/coverage/**',
+      'gitnexus/vendor/**',
+      'gitnexus-web/src/vendor/**',
+      'gitnexus/test/fixtures/**',
+      'gitnexus-web/playwright-report/**',
+      'gitnexus-web/test-results/**',
+      '**/*.d.ts',
+      '.claude/**',
+      '.history/**',
+    ],
+  },
+
+  // Base TypeScript config for all packages
+  {
+    files: ['**/*.{ts,tsx}'],
+    languageOptions: {
+      parser: tsParser,
+      parserOptions: {
+        ecmaVersion: 2022,
+        sourceType: 'module',
+      },
+    },
+    plugins: {
+      '@typescript-eslint': tsPlugin,
+      'unused-imports': unusedImports,
+    },
+    rules: {
+      // Unused imports — auto-fixable
+      'unused-imports/no-unused-imports': 'error',
+      'unused-imports/no-unused-vars': [
+        'warn',
+        { vars: 'all', varsIgnorePattern: '^_', args: 'after-used', argsIgnorePattern: '^_' },
+      ],
+
+      // TypeScript quality
+      '@typescript-eslint/no-unused-vars': 'off', // handled by unused-imports plugin
+      'no-unused-vars': 'off', // handled by unused-imports plugin
+      '@typescript-eslint/no-explicit-any': 'warn',
+      '@typescript-eslint/no-non-null-assertion': 'warn',
+
+      // General quality
+      'no-debugger': 'error',
+      'prefer-const': 'error',
+      'no-var': 'error',
+      eqeqeq: ['error', 'always', { null: 'ignore' }],
+    },
+  },
+
+  // CLI package — allow console.log (it's a CLI tool)
+  {
+    files: ['gitnexus/src/cli/**/*.ts', 'gitnexus/src/server/**/*.ts'],
+    rules: {
+      'no-console': 'off',
+    },
+  },
+
+  // React-specific rules for gitnexus-web
+  {
+    files: ['gitnexus-web/src/**/*.{ts,tsx}'],
+    plugins: {
+      'react-hooks': reactHooks,
+    },
+    rules: {
+      'react-hooks/rules-of-hooks': 'error',
+      'react-hooks/exhaustive-deps': 'warn',
+    },
+  },
+
+  // Disable formatting rules (prettier handles those)
+  prettierConfig,
+];

eval/.env.example

@@ -0,0 +1,23 @@
+# ─── GitNexus SWE-bench Eval — API Keys ───
+# Copy this file to .env and fill in the keys you have.
+# You only need keys for the models you plan to test.
+
+# OpenRouter (covers Claude, MiniMax, GLM, and 200+ other models)
+# Get yours at: https://openrouter.ai/keys
+OPENROUTER_API_KEY=
+
+# Anthropic (direct — optional if using OpenRouter)
+# Get yours at: https://console.anthropic.com/
+ANTHROPIC_API_KEY=
+
+# ZhipuAI / GLM (direct — optional if using OpenRouter)
+# Get yours at: https://open.bigmodel.cn/
+ZHIPUAI_API_KEY=
+
+# MiniMax (direct — optional if using OpenRouter)
+MINIMAX_API_KEY=
+
+# ─── Optional ───
+
+# Cost tracking: set to "ignore_errors" if litellm can't find pricing for a model
+# MSWEA_COST_TRACKING=ignore_errors

eval/.gitignore

@@ -0,0 +1,16 @@
+# Evaluation results (large, should not be committed)
+results/
+*.traj.json
+preds.json
+
+# Python
+__pycache__/
+*.pyc
+*.egg-info/
+.eggs/
+dist/
+build/
+
+# Environment
+.env
+.venv/

eval/README.md

@@ -0,0 +1,214 @@
+# GitNexus SWE-bench Evaluation Harness
+
+Evaluate whether GitNexus code intelligence improves AI agent performance on real software engineering tasks. Runs SWE-bench instances across multiple models and compares baseline (no graph) vs GitNexus-enhanced configurations.
+
+## What This Tests
+
+**Hypothesis**: Giving AI agents structural code intelligence (call graphs, execution flows, blast radius analysis) improves their ability to resolve real GitHub issues — measured by resolve rate, cost, and efficiency.
+
+**Evaluation modes:**
+
+| Mode | What the agent gets |
+|------|-------------------|
+| `baseline` | Standard bash tools (grep, find, cat, sed) — control group |
+| `native` | Baseline + explicit GitNexus tools via eval-server (~100ms) |
+| `native_augment` | Native tools + grep results automatically enriched with graph context (**recommended**) |
+
+> **Recommended**: Use `native_augment` mode. It mirrors the Claude Code model — the agent gets both explicit GitNexus tools (fast bash commands) AND automatic enrichment of grep results with callers, callees, and execution flows. The agent decides when to use explicit tools vs rely on enriched search output.
+
+**Models supported:**
+
+- Claude 3.5 Haiku, Claude Sonnet 4, Claude Opus 4
+- MiniMax M1 2.5
+- GLM 4.7, GLM 5
+- Any model supported by litellm (add a YAML config)
+
+## Prerequisites
+
+- Python 3.11+
+- Docker (for SWE-bench containers)
+- Node.js 18+ (for GitNexus)
+- API keys for your chosen models
+
+## Setup
+
+```bash
+cd eval
+
+# Install dependencies
+pip install -e .
+
+# Set up API keys — copy the template and fill in your keys
+cp .env.example .env
+# Then edit .env and paste your key(s)
+```
+
+All models are routed through **OpenRouter** by default, so a single `OPENROUTER_API_KEY` is all you need. To use provider APIs directly (Anthropic, ZhipuAI, etc.), edit the model YAML in `configs/models/` and set the corresponding key in `.env`.
+
+```bash
+# Pull SWE-bench Docker images (pulled on-demand, but you can pre-pull)
+docker pull swebench/sweb.eval.x86_64.django_1776_django-16527:latest
+```
+
+### Debug logging
+
+Set `GITNEXUS_EVAL_DEBUG=1` to include full Python tracebacks in run summaries and logs. By default, errors are sanitized to avoid leaking host paths or stack traces.
+
+## Quick Start
+
+### Debug a single instance
+
+```bash
+# Fastest way to verify everything works
+python run_eval.py debug -m claude-haiku -i django__django-16527 --subset lite
+```
+
+### Run a single configuration
+
+```bash
+# 5 instances, Claude Sonnet, native_augment mode (default)
+python run_eval.py single -m claude-sonnet --subset lite --slice 0:5
+
+# Baseline comparison (no GitNexus)
+python run_eval.py single -m claude-sonnet --mode baseline --subset lite --slice 0:5
+
+# Full Lite benchmark, 4 parallel workers
+python run_eval.py single -m claude-sonnet --subset lite -w 4
+```
+
+### Run the full matrix
+
+```bash
+# All models x all modes
+python run_eval.py matrix --subset lite -w 4
+
+# Key comparison: baseline vs native_augment
+python run_eval.py matrix -m claude-sonnet -m claude-haiku --modes baseline --modes native_augment --subset lite --slice 0:50
+```
+
+### Analyze results
+
+```bash
+# Summary table
+python -m analysis.analyze_results results/
+
+# Compare modes for a specific model
+python -m analysis.analyze_results compare-modes results/ -m claude-sonnet
+
+# GitNexus tool usage analysis
+python -m analysis.analyze_results gitnexus-usage results/
+
+# Export as CSV for further analysis
+python -m analysis.analyze_results summary results/ --format csv > results.csv
+
+# Run official SWE-bench test evaluation
+python -m analysis.analyze_results summary results/ --swebench-eval
+```
+
+### List available configurations
+
+```bash
+python run_eval.py list-configs
+```
+
+## Architecture
+
+```
+eval/
+  run_eval.py              # Main entry point (single, matrix, debug commands)
+  agents/
+    gitnexus_agent.py      # GitNexusAgent: extends DefaultAgent with augmentation + metrics
+  environments/
+    gitnexus_docker.py     # Docker env with GitNexus + eval-server + standalone tool scripts
+  bridge/
+    gitnexus_tools.sh      # Bash wrappers (legacy — now standalone scripts are installed directly)
+    mcp_bridge.py          # Legacy MCP bridge (kept for reference)
+  prompts/
+    system_baseline.jinja          # System: persona + format rules
+    instance_baseline.jinja        # Instance: task + workflow
+    system_native.jinja            # System: + GitNexus tool reference
+    instance_native.jinja          # Instance: + GitNexus debugging workflow
+    system_native_augment.jinja    # System: + GitNexus tools + grep enrichment docs
+    instance_native_augment.jinja  # Instance: + GitNexus workflow + risk assessment
+  configs/
+    models/                # Per-model YAML configs
+    modes/                 # Per-mode YAML configs (baseline, native, native_augment)
+  analysis/
+    analyze_results.py     # Post-run comparative analysis
+  results/                 # Output directory (gitignored)
+```
+
+## How It Works
+
+### Template structure
+
+mini-swe-agent requires two Jinja templates:
+- **system_template** → system message: persona, format rules, tool reference (static)
+- **instance_template** → first user message: task, workflow, rules, examples (contains `{{task}}`)
+
+Each mode has a `system_{mode}.jinja` + `instance_{mode}.jinja` pair. The agent loads both automatically based on the configured mode.
+
+### Per-instance flow
+
+1. Docker container starts with SWE-bench instance (repo at specific commit)
+2. **GitNexus setup**: Node.js + gitnexus installed, `gitnexus analyze` runs (or restores from cache)
+3. **Eval-server starts**: `gitnexus eval-server` daemon (persistent HTTP server, keeps LadybugDB warm)
+4. **Standalone tool scripts installed** in `/usr/local/bin/` — works with `subprocess.run` (no `.bashrc` needed)
+5. Agent runs with the configured model + system prompt + GitNexus tools
+6. Agent's patch is extracted as a git diff
+7. Metrics collected: cost, tokens, tool calls, GitNexus usage, augmentation stats
+
+### Tool architecture
+
+```
+Agent → bash command → /usr/local/bin/gitnexus-query
+  → curl localhost:4848/tool/query     (fast path: eval-server, ~100ms)
+  → npx gitnexus query                 (fallback: cold CLI, ~5-10s)
+```
+
+Each tool script in `/usr/local/bin/` is standalone — no sourcing, no env inheritance needed. This is critical because mini-swe-agent runs every command via `subprocess.run` in a fresh subshell.
+
+### Eval-server
+
+The eval-server is a lightweight HTTP daemon that:
+- Keeps LadybugDB warm in memory (no cold start per tool call)
+- Returns LLM-friendly text (not raw JSON — saves tokens)
+- Includes next-step hints to guide tool chaining (query → context → impact → fix)
+- Auto-shuts down after idle timeout
+
+### Index caching
+
+SWE-bench repos repeat (Django has 200+ instances at different commits). The harness caches GitNexus indexes per `(repo, commit)` hash in `~/.gitnexus-eval-cache/` to avoid redundant re-indexing.
+
+### Grep augmentation (native_augment mode)
+
+When the agent runs `grep` or `rg`, the observation is post-processed: the agent class calls `gitnexus-augment` on the search pattern and appends `[GitNexus]` annotations showing callers, callees, and execution flows for matched symbols. This mirrors the Claude Code / Cursor hook integration.
+
+## Adding Models
+
+Create a YAML file in `configs/models/`:
+
+```yaml
+# configs/models/my-model.yaml
+model:
+  model_name: "openrouter/provider/model-name"
+  cost_tracking: "ignore_errors"  # if not in litellm's cost DB
+  model_kwargs:
+    max_tokens: 8192
+    temperature: 0
+```
+
+The model name follows [litellm conventions](https://docs.litellm.ai/docs/providers).
+
+## Metrics Collected
+
+| Metric | Description |
+|--------|-------------|
+| Patch Rate | % of instances where agent produced a patch |
+| Resolve Rate | % of instances where patch passes tests (requires --swebench-eval) |
+| Total Cost | API cost across all instances |
+| Avg Cost/Instance | Cost efficiency |
+| API Calls | Number of LLM calls |
+| GN Tool Calls | How many GitNexus tools the agent used |
+| Augment Hits | How many grep/find results got enriched |
+| Augment Hit Rate | % of search commands that got useful enrichment |

eval/__init__.py

@@ -0,0 +1 @@
+# GitNexus SWE-bench Evaluation Harness

eval/agents/gitnexus_agent.py

@@ -0,0 +1,199 @@
+"""
+GitNexus-Enhanced Agent for SWE-bench Evaluation
+
+Extends mini-swe-agent's DefaultAgent with:
+1. Native augment mode: GitNexus tools via eval-server + grep enrichment (recommended)
+2. Native mode: GitNexus tools via eval-server only
+3. Baseline mode: Pure mini-swe-agent (no GitNexus — control group)
+
+The agent class itself is minimal — the heavy lifting is in:
+- Prompt selection (system + instance templates per mode)
+- Observation post-processing (grep result augmentation)
+- Metrics tracking (which tools the agent actually uses)
+
+Template structure (matches mini-swe-agent's expectations):
+  system_template  → system message: persona + format rules + tool reference
+  instance_template → first user message: task + workflow + rules + examples
+"""
+
+import logging
+import re
+import time
+from enum import Enum
+from pathlib import Path
+
+from constants import AUGMENT_TIMEOUT_SECONDS
+from minisweagent import Environment, Model
+from minisweagent.agents.default import AgentConfig, DefaultAgent
+from tool_registry import BINARIES_BY_KEY, TOOL_METRIC_KEYS
+
+logger = logging.getLogger("gitnexus_agent")
+
+PROMPTS_DIR = Path(__file__).parent.parent / "prompts"
+
+
+class GitNexusMode(str, Enum):
+    """Evaluation modes for GitNexus integration."""
+    BASELINE = "baseline"               # No GitNexus — pure mini-swe-agent
+    NATIVE = "native"                   # GitNexus tools via eval-server
+    NATIVE_AUGMENT = "native_augment"   # Native tools + grep enrichment (recommended)
+
+
+class GitNexusAgentConfig(AgentConfig):
+    """Extended config for GitNexus evaluation agent."""
+    gitnexus_mode: GitNexusMode = GitNexusMode.BASELINE
+    augment_timeout: float = AUGMENT_TIMEOUT_SECONDS
+    augment_min_pattern_length: int = 3
+    track_gitnexus_usage: bool = True
+
+
+class GitNexusAgent(DefaultAgent):
+    """
+    Agent that optionally enriches its capabilities with GitNexus code intelligence.
+
+    In BASELINE mode, behaves identically to DefaultAgent.
+    In NATIVE mode, GitNexus tools are available as bash commands via eval-server.
+    In NATIVE_AUGMENT mode, GitNexus tools + automatic grep result enrichment.
+    """
+
+    def __init__(self, model: Model, env: Environment, *, config_class: type = GitNexusAgentConfig, **kwargs):
+        mode = kwargs.get("gitnexus_mode", GitNexusMode.BASELINE)
+        if isinstance(mode, str):
+            mode = GitNexusMode(mode)
+
+        # Load system template
+        system_file = PROMPTS_DIR / f"system_{mode.value}.jinja"
+        if system_file.exists() and "system_template" not in kwargs:
+            kwargs["system_template"] = system_file.read_text()
+
+        # Load instance template
+        instance_file = PROMPTS_DIR / f"instance_{mode.value}.jinja"
+        if instance_file.exists() and "instance_template" not in kwargs:
+            kwargs["instance_template"] = instance_file.read_text()
+
+        super().__init__(model, env, config_class=config_class, **kwargs)
+        self.gitnexus_mode = mode
+        self.gitnexus_metrics = GitNexusMetrics()
+
+    def execute_actions(self, message: dict) -> list[dict]:
+        """Execute actions with optional GitNexus augmentation and tracking."""
+        if self.config.track_gitnexus_usage:
+            self._track_tool_usage(message)
+
+        outputs = [self.env.execute(action) for action in message.get("extra", {}).get("actions", [])]
+
+        # Augment grep/find observations in NATIVE_AUGMENT mode
+        if self.gitnexus_mode == GitNexusMode.NATIVE_AUGMENT:
+            actions = message.get("extra", {}).get("actions", [])
+            for i, (action, output) in enumerate(zip(actions, outputs)):
+                augmented = self._maybe_augment(action, output)
+                if augmented:
+                    outputs[i] = augmented
+
+        return self.add_messages(
+            *self.model.format_observation_messages(message, outputs, self.get_template_vars())
+        )
+
+    def _maybe_augment(self, action: dict, output: dict) -> dict | None:
+        """
+        If the action is a search command (grep, find, rg, ag), augment the output
+        with GitNexus knowledge graph context.
+        """
+        command = action.get("command", "")
+        if not command:
+            return None
+
+        pattern = self._extract_search_pattern(command)
+        if not pattern or len(pattern) < self.config.augment_min_pattern_length:
+            return None
+
+        start = time.time()
+        try:
+            augment_result = self.env.execute({
+                "command": f'gitnexus-augment "{pattern}" 2>&1 || true',
+                "timeout": self.config.augment_timeout,
+            })
+            elapsed = time.time() - start
+            self.gitnexus_metrics.augmentation_calls += 1
+            self.gitnexus_metrics.augmentation_time += elapsed
+
+            augment_text = augment_result.get("output", "").strip()
+            if augment_text and "[GitNexus]" in augment_text:
+                original_output = output.get("output", "")
+                output = dict(output)
+                output["output"] = f"{original_output}\n\n{augment_text}"
+                self.gitnexus_metrics.augmentation_hits += 1
+                return output
+        except Exception as e:
+            logger.debug(f"Augmentation failed for pattern '{pattern}': {e}")
+            self.gitnexus_metrics.augmentation_errors += 1
+
+        return None
+
+    @staticmethod
+    def _extract_search_pattern(command: str) -> str | None:
+        """Extract the search pattern from a grep/find/rg command."""
+        patterns = [
+            r'(?:grep|rg|ag)\s+(?:-[a-zA-Z]*\s+)*["\']([^"\']+)["\']',
+            r'(?:grep|rg|ag)\s+(?:-[a-zA-Z]*\s+)*(\S+)',
+        ]
+
+        for pat in patterns:
+            match = re.search(pat, command)
+            if match:
+                result = match.group(1)
+                if result.startswith("/") or result.startswith("."):
+                    continue
+                if result.startswith("-"):
+                    continue
+                return result
+
+        return None
+
+    def _track_tool_usage(self, message: dict):
+        """Track which GitNexus tools the agent uses."""
+        for action in message.get("extra", {}).get("actions", []):
+            command = action.get("command", "")
+            for key, binary in BINARIES_BY_KEY.items():
+                if binary in command and key in self.gitnexus_metrics.tool_calls:
+                    self.gitnexus_metrics.tool_calls[key] += 1
+                    break
+
+    def serialize(self, *extra_dicts) -> dict:
+        """Serialize with GitNexus-specific metrics."""
+        gitnexus_data = {
+            "info": {
+                "gitnexus": {
+                    "mode": self.gitnexus_mode.value,
+                    "metrics": self.gitnexus_metrics.to_dict(),
+                },
+            },
+        }
+        return super().serialize(gitnexus_data, *extra_dicts)
+
+
+class GitNexusMetrics:
+    """Tracks GitNexus-specific metrics during evaluation."""
+
+    def __init__(self):
+        self.tool_calls: dict[str, int] = {key: 0 for key in TOOL_METRIC_KEYS}
+        self.augmentation_calls: int = 0
+        self.augmentation_hits: int = 0
+        self.augmentation_errors: int = 0
+        self.augmentation_time: float = 0.0
+        self.index_time: float = 0.0
+
+    @property
+    def total_tool_calls(self) -> int:
+        return sum(self.tool_calls.values())
+
+    def to_dict(self) -> dict:
+        return {
+            "tool_calls": dict(self.tool_calls),
+            "total_tool_calls": self.total_tool_calls,
+            "augmentation_calls": self.augmentation_calls,
+            "augmentation_hits": self.augmentation_hits,
+            "augmentation_errors": self.augmentation_errors,
+            "augmentation_time_seconds": round(self.augmentation_time, 2),
+            "index_time_seconds": round(self.index_time, 2),
+        }

eval/analysis/analyze_results.py

@@ -0,0 +1,455 @@
+#!/usr/bin/env python3
+"""
+Results Analyzer for GitNexus SWE-bench Evaluation
+
+Reads evaluation results and generates comparative analysis:
+- Resolve rate by model x mode
+- Cost comparison (total, per-instance)
+- Token/API call efficiency
+- GitNexus tool usage patterns
+- Augmentation hit rates
+
+Usage:
+    python -m analysis.analyze_results /path/to/results
+    python -m analysis.analyze_results /path/to/results --format markdown
+    python -m analysis.analyze_results /path/to/results --swebench-eval  # run actual test verification
+"""
+
+import json
+import logging
+import os
+import subprocess
+import sys
+from pathlib import Path
+from typing import Any
+
+import typer
+from rich.console import Console
+from rich.table import Table
+
+from tool_registry import TOOL_METRIC_KEYS
+
+logger = logging.getLogger("analyze_results")
+console = Console()
+app = typer.Typer(rich_markup_mode="rich", add_completion=False)
+
+
+def load_run_results(results_dir: Path) -> dict[str, dict]:
+    """
+    Load all run results from the results directory.
+    
+    Returns: {run_id: {summary, preds, instances}}
+    """
+    runs = {}
+
+    for run_dir in sorted(results_dir.iterdir()):
+        if not run_dir.is_dir():
+            continue
+
+        run_id = run_dir.name
+        run_data: dict[str, Any] = {"run_id": run_id, "dir": run_dir}
+
+        # Load summary
+        summary_path = run_dir / "summary.json"
+        if summary_path.exists():
+            run_data["summary"] = json.loads(summary_path.read_text())
+
+        # Load predictions
+        preds_path = run_dir / "preds.json"
+        if preds_path.exists():
+            run_data["preds"] = json.loads(preds_path.read_text())
+
+        # Load individual trajectories for detailed metrics
+        run_data["trajectories"] = {}
+        for traj_dir in run_dir.iterdir():
+            if not traj_dir.is_dir():
+                continue
+            for traj_file in traj_dir.glob("*.traj.json"):
+                try:
+                    traj = json.loads(traj_file.read_text())
+                    instance_id = traj.get("instance_id", traj_dir.name)
+                    run_data["trajectories"][instance_id] = traj
+                except Exception:
+                    pass
+
+        if run_data.get("preds") or run_data.get("summary"):
+            runs[run_id] = run_data
+
+    return runs
+
+
+def parse_run_id(run_id: str) -> tuple[str, str]:
+    """Parse 'model_mode' into (model, mode) using known suffixes."""
+    # Match the longest known suffix first to avoid hyphen collisions in model names.
+    known_modes = [
+        "native_augment",
+        "native",
+        "baseline",
+        "mcp",
+        "augment",
+        "full",
+    ]
+    for mode in known_modes:
+        suffix = f"_{mode}"
+        if run_id.endswith(suffix):
+            return run_id[: -len(suffix)], mode
+    return run_id, "unknown"
+
+
+def compute_metrics(run_data: dict) -> dict:
+    """Compute evaluation metrics for a single run."""
+    preds = run_data.get("preds", {})
+    summary = run_data.get("summary", {})
+    trajectories = run_data.get("trajectories", {})
+
+    n_instances = len(preds)
+    n_with_patch = sum(1 for p in preds.values() if p.get("model_patch", "").strip())
+
+    # Cost and API call metrics from trajectories
+    costs = []
+    api_calls = []
+    gn_tool_calls = []
+    gn_augment_hits = []
+    gn_augment_calls = []
+
+    for instance_id, traj in trajectories.items():
+        info = traj.get("info", {})
+        model_stats = info.get("model_stats", {})
+        costs.append(model_stats.get("instance_cost", 0))
+        api_calls.append(model_stats.get("api_calls", 0))
+
+        gn = info.get("gitnexus", {}).get("metrics", {})
+        if gn:
+            gn_tool_calls.append(gn.get("total_tool_calls", 0))
+            gn_augment_hits.append(gn.get("augmentation_hits", 0))
+            gn_augment_calls.append(gn.get("augmentation_calls", 0))
+
+    # Also try summary-level metrics
+    if not costs and summary:
+        results = summary.get("results", [])
+        for r in results:
+            costs.append(r.get("cost", 0))
+            api_calls.append(r.get("n_calls", 0))
+            gn = r.get("gitnexus_metrics", {})
+            if gn:
+                gn_tool_calls.append(gn.get("total_tool_calls", 0))
+                gn_augment_hits.append(gn.get("augmentation_hits", 0))
+                gn_augment_calls.append(gn.get("augmentation_calls", 0))
+
+    total_cost = sum(costs)
+    total_calls = sum(api_calls)
+
+    return {
+        "n_instances": n_instances,
+        "n_with_patch": n_with_patch,
+        "patch_rate": n_with_patch / max(n_instances, 1),
+        "total_cost": total_cost,
+        "avg_cost": total_cost / max(n_instances, 1),
+        "total_api_calls": total_calls,
+        "avg_api_calls": total_calls / max(n_instances, 1),
+        "total_gn_tool_calls": sum(gn_tool_calls),
+        "avg_gn_tool_calls": sum(gn_tool_calls) / max(len(gn_tool_calls), 1) if gn_tool_calls else 0,
+        "total_augment_hits": sum(gn_augment_hits),
+        "total_augment_calls": sum(gn_augment_calls),
+        "augment_hit_rate": sum(gn_augment_hits) / max(sum(gn_augment_calls), 1) if gn_augment_calls else 0,
+    }
+
+
+def run_swebench_evaluation(results_dir: Path, run_id: str, subset: str = "lite") -> dict | None:
+    """
+    Run the official SWE-bench evaluation on predictions.
+    
+    Requires: pip install swebench
+    """
+    preds_path = results_dir / run_id / "preds.json"
+    if not preds_path.exists():
+        return None
+
+    dataset_mapping = {
+        "lite": "princeton-nlp/SWE-Bench_Lite",
+        "verified": "princeton-nlp/SWE-Bench_Verified",
+        "full": "princeton-nlp/SWE-Bench",
+    }
+
+    try:
+        eval_output = results_dir / run_id / "swebench_eval"
+        cmd = [
+            sys.executable, "-m", "swebench.harness.run_evaluation",
+            "--dataset_name", dataset_mapping.get(subset, subset),
+            "--predictions_path", str(preds_path),
+            "--max_workers", "4",
+            "--run_id", run_id,
+            "--output_dir", str(eval_output),
+        ]
+
+        logger.info(f"Running SWE-bench evaluation for {run_id}...")
+        result = subprocess.run(cmd, capture_output=True, text=True, timeout=600)
+
+        if result.returncode == 0:
+            # Parse evaluation results
+            report_path = eval_output / run_id / "results.json"
+            if report_path.exists():
+                return json.loads(report_path.read_text())
+
+        logger.error(f"SWE-bench eval failed: {result.stderr[:500]}")
+        return None
+
+    except Exception as e:
+        logger.error(f"SWE-bench eval error: {e}")
+        return None
+
+
+# ─── CLI Commands ───────────────────────────────────────────────────────────
+
+
+@app.command()
+def summary(
+    results_dir: str = typer.Argument(..., help="Path to results directory"),
+    format: str = typer.Option("table", "--format", help="Output format: table, markdown, json, csv"),
+    swebench_eval: bool = typer.Option(False, "--swebench-eval", help="Run official SWE-bench test evaluation"),
+    subset: str = typer.Option("lite", "--subset", help="SWE-bench subset (for --swebench-eval)"),
+):
+    """Generate comparative analysis of evaluation results."""
+    results_path = Path(results_dir)
+    if not results_path.exists():
+        console.print(f"[red]Results directory not found: {results_path}[/red]")
+        raise typer.Exit(1)
+
+    runs = load_run_results(results_path)
+    if not runs:
+        console.print("[yellow]No evaluation results found[/yellow]")
+        raise typer.Exit(0)
+
+    console.print(f"\n[bold]Found {len(runs)} evaluation runs[/bold]\n")
+
+    # Compute metrics per run
+    all_metrics = {}
+    for run_id, run_data in runs.items():
+        model, mode = parse_run_id(run_id)
+        metrics = compute_metrics(run_data)
+        metrics["model"] = model
+        metrics["mode"] = mode
+
+        # Optionally run SWE-bench evaluation
+        if swebench_eval:
+            eval_result = run_swebench_evaluation(results_path, run_id, subset)
+            if eval_result:
+                metrics["resolved"] = eval_result.get("resolved", 0)
+                metrics["resolve_rate"] = eval_result.get("resolved", 0) / max(metrics["n_instances"], 1)
+
+        all_metrics[run_id] = metrics
+
+    if format == "table":
+        _print_table(all_metrics)
+    elif format == "markdown":
+        _print_markdown(all_metrics)
+    elif format == "json":
+        console.print(json.dumps(all_metrics, indent=2))
+    elif format == "csv":
+        _print_csv(all_metrics)
+
+
+@app.command()
+def compare_modes(
+    results_dir: str = typer.Argument(..., help="Path to results directory"),
+    model: str = typer.Option(..., "-m", "--model", help="Model to compare across modes"),
+):
+    """Compare modes for a specific model (baseline vs mcp vs augment vs full)."""
+    results_path = Path(results_dir)
+    runs = load_run_results(results_path)
+
+    # Filter to the specified model
+    model_runs = {
+        run_id: data for run_id, data in runs.items()
+        if parse_run_id(run_id)[0] == model
+    }
+
+    if not model_runs:
+        console.print(f"[yellow]No results found for model: {model}[/yellow]")
+        raise typer.Exit(1)
+
+    console.print(f"\n[bold]Mode comparison for {model}[/bold]\n")
+
+    metrics = {}
+    for run_id, run_data in model_runs.items():
+        _, mode = parse_run_id(run_id)
+        metrics[mode] = compute_metrics(run_data)
+
+    mode_order = [
+        mode
+        for mode in ["baseline", "native", "native_augment", "mcp", "augment", "full"]
+        if mode in metrics
+    ] or sorted(metrics.keys())
+
+    # Print comparison table
+    table = Table(title=f"Mode Comparison: {model}")
+    table.add_column("Metric", style="bold")
+    for mode in mode_order:
+        table.add_column(mode, justify="right")
+
+    rows = [
+        ("Instances", "n_instances", "d"),
+        ("With Patch", "n_with_patch", "d"),
+        ("Patch Rate", "patch_rate", ".1%"),
+        ("Total Cost", "total_cost", "$.4f"),
+        ("Avg Cost", "avg_cost", "$.4f"),
+        ("Total API Calls", "total_api_calls", "d"),
+        ("Avg API Calls", "avg_api_calls", ".1f"),
+        ("GN Tool Calls", "total_gn_tool_calls", "d"),
+        ("Augment Hits", "total_augment_hits", "d"),
+        ("Augment Hit Rate", "augment_hit_rate", ".1%"),
+    ]
+
+    for label, key, fmt in rows:
+        values = []
+        for mode in mode_order:
+            if mode in metrics:
+                v = metrics[mode].get(key, 0)
+                if fmt == ".1%":
+                    values.append(f"{v:.1%}")
+                elif fmt == "$.4f":
+                    values.append(f"${v:.4f}")
+                elif fmt == ".1f":
+                    values.append(f"{v:.1f}")
+                else:
+                    values.append(str(v))
+        table.add_row(label, *values)
+
+    # Add delta rows (improvement over baseline)
+    if "baseline" in metrics:
+        baseline_cost = metrics["baseline"]["avg_cost"]
+        baseline_calls = metrics["baseline"]["avg_api_calls"]
+
+        table.add_section()
+        for mode in mode_order:
+            if mode == "baseline":
+                continue
+            mode_cost = metrics[mode]["avg_cost"]
+            mode_calls = metrics[mode]["avg_api_calls"]
+
+            cost_delta = ((mode_cost - baseline_cost) / max(baseline_cost, 0.001)) * 100
+            calls_delta = ((mode_calls - baseline_calls) / max(baseline_calls, 1)) * 100
+
+            cost_str = f"{cost_delta:+.1f}%"
+            calls_str = f"{calls_delta:+.1f}%"
+
+            # Color-code: negative is good (cheaper/fewer calls)
+            cost_color = "green" if cost_delta < 0 else "red"
+            calls_color = "green" if calls_delta < 0 else "red"
+
+            console.print(f"  {mode} vs baseline: cost [{cost_color}]{cost_str}[/{cost_color}], calls [{calls_color}]{calls_str}[/{calls_color}]")
+
+    console.print(table)
+
+
+@app.command()
+def gitnexus_usage(
+    results_dir: str = typer.Argument(..., help="Path to results directory"),
+):
+    """Analyze GitNexus tool usage patterns across all runs."""
+    results_path = Path(results_dir)
+    runs = load_run_results(results_path)
+
+    console.print("\n[bold]GitNexus Tool Usage Analysis[/bold]\n")
+
+    table = Table(title="Tool Usage by Run")
+    table.add_column("Run", style="bold")
+    for key in TOOL_METRIC_KEYS:
+        table.add_column(key, justify="right")
+    table.add_column("Total", justify="right")
+    table.add_column("Augment Hits", justify="right")
+
+    for run_id, run_data in sorted(runs.items()):
+        _, mode = parse_run_id(run_id)
+        if mode == "baseline":
+            continue
+
+        # Aggregate tool calls across trajectories
+        tool_totals: dict[str, int] = {key: 0 for key in TOOL_METRIC_KEYS}
+        augment_hits = 0
+
+        for traj in run_data.get("trajectories", {}).values():
+            gn = traj.get("info", {}).get("gitnexus", {}).get("metrics", {})
+            for tool, count in gn.get("tool_calls", {}).items():
+                tool_totals[tool] = tool_totals.get(tool, 0) + count
+            augment_hits += gn.get("augmentation_hits", 0)
+
+        # Also check summary
+        for r in run_data.get("summary", {}).get("results", []):
+            gn = r.get("gitnexus_metrics", {})
+            for tool, count in gn.get("tool_calls", {}).items():
+                tool_totals[tool] = tool_totals.get(tool, 0) + count
+            augment_hits += gn.get("augmentation_hits", 0)
+
+        total = sum(tool_totals.values())
+        if total > 0 or augment_hits > 0:
+            table.add_row(
+                run_id,
+                *[str(tool_totals.get(key, 0)) for key in TOOL_METRIC_KEYS],
+                str(total),
+                str(augment_hits),
+            )
+
+    console.print(table)
+
+
+# ─── Output Formatters ─────────────────────────────────────────────────────
+
+
+def _print_table(all_metrics: dict):
+    """Print rich table summary."""
+    table = Table(title="Evaluation Results")
+    table.add_column("Run", style="bold")
+    table.add_column("Model")
+    table.add_column("Mode")
+    table.add_column("N", justify="right")
+    table.add_column("Patched", justify="right")
+    table.add_column("Rate", justify="right")
+    table.add_column("Cost", justify="right")
+    table.add_column("Calls", justify="right")
+    table.add_column("GN Tools", justify="right")
+
+    for run_id, m in sorted(all_metrics.items()):
+        resolved_str = ""
+        if "resolve_rate" in m:
+            resolved_str = f" ({m['resolve_rate']:.0%})"
+
+        table.add_row(
+            run_id,
+            m["model"],
+            m["mode"],
+            str(m["n_instances"]),
+            str(m["n_with_patch"]),
+            f"{m['patch_rate']:.0%}{resolved_str}",
+            f"${m['total_cost']:.2f}",
+            str(m["total_api_calls"]),
+            str(m["total_gn_tool_calls"]) if m["total_gn_tool_calls"] > 0 else "-",
+        )
+
+    console.print(table)
+
+
+def _print_markdown(all_metrics: dict):
+    """Print markdown table."""
+    print("| Run | Model | Mode | N | Patched | Rate | Cost | Calls | GN Tools |")
+    print("|-----|-------|------|---|---------|------|------|-------|----------|")
+    for run_id, m in sorted(all_metrics.items()):
+        gn = str(m["total_gn_tool_calls"]) if m["total_gn_tool_calls"] > 0 else "-"
+        print(f"| {run_id} | {m['model']} | {m['mode']} | {m['n_instances']} | {m['n_with_patch']} | {m['patch_rate']:.0%} | ${m['total_cost']:.2f} | {m['total_api_calls']} | {gn} |")
+
+
+def _print_csv(all_metrics: dict):
+    """Print CSV output."""
+    print("run_id,model,mode,n_instances,n_with_patch,patch_rate,total_cost,avg_cost,total_api_calls,avg_api_calls,total_gn_tool_calls,total_augment_hits,augment_hit_rate")
+    for run_id, m in sorted(all_metrics.items()):
+        print(
+            f"{run_id},{m['model']},{m['mode']},{m['n_instances']},{m['n_with_patch']},"
+            f"{m['patch_rate']:.4f},{m['total_cost']:.4f},{m['avg_cost']:.4f},"
+            f"{m['total_api_calls']},{m['avg_api_calls']:.1f},{m['total_gn_tool_calls']},"
+            f"{m['total_augment_hits']},{m['augment_hit_rate']:.4f}"
+        )
+
+
+if __name__ == "__main__":
+    logging.basicConfig(level=logging.INFO)
+    app()

eval/bridge/gitnexus_tools.sh

@@ -0,0 +1,155 @@
+#!/bin/bash
+# GitNexus CLI tool wrappers for SWE-bench evaluation
+#
+# These functions call the GitNexus eval-server (HTTP daemon) for near-instant
+# tool responses. The eval-server keeps KuzuDB warm in memory.
+#
+# If the eval-server is not running, falls back to direct CLI commands.
+#
+# Usage:
+#   gitnexus-query "how does authentication work"
+#   gitnexus-context "validateUser"
+#   gitnexus-impact "AuthService" upstream
+#   gitnexus-cypher "MATCH (n:Function) RETURN n.name LIMIT 10"
+#   gitnexus-overview
+
+GITNEXUS_EVAL_PORT="${GITNEXUS_EVAL_PORT:-4848}"
+GITNEXUS_EVAL_URL="http://127.0.0.1:${GITNEXUS_EVAL_PORT}"
+
+_gitnexus_call() {
+    local tool="$1"
+    shift
+    local json_body="$1"
+
+    # Try eval-server first (fastest path — KuzuDB stays warm)
+    local result
+    result=$(curl -sf -X POST "${GITNEXUS_EVAL_URL}/tool/${tool}" \
+        -H "Content-Type: application/json" \
+        -d "${json_body}" 2>/dev/null)
+
+    if [ $? -eq 0 ] && [ -n "$result" ]; then
+        echo "$result"
+        return 0
+    fi
+
+    # Fallback: direct CLI (cold start, slower but always works)
+    case "$tool" in
+        query)
+            local q=$(echo "$json_body" | python3 -c "import sys,json; print(json.load(sys.stdin).get('query',''))" 2>/dev/null)
+            npx gitnexus query "$q" 2>&1
+            ;;
+        context)
+            local n=$(echo "$json_body" | python3 -c "import sys,json; print(json.load(sys.stdin).get('name',''))" 2>/dev/null)
+            npx gitnexus context "$n" 2>&1
+            ;;
+        impact)
+            local t=$(echo "$json_body" | python3 -c "import sys,json; d=json.load(sys.stdin); print(d.get('target',''))" 2>/dev/null)
+            local d=$(echo "$json_body" | python3 -c "import sys,json; d=json.load(sys.stdin); print(d.get('direction','upstream'))" 2>/dev/null)
+            npx gitnexus impact "$t" --direction "$d" 2>&1
+            ;;
+        cypher)
+            local cq=$(echo "$json_body" | python3 -c "import sys,json; print(json.load(sys.stdin).get('query',''))" 2>/dev/null)
+            npx gitnexus cypher "$cq" 2>&1
+            ;;
+        *)
+            echo "Unknown tool: $tool" >&2
+            return 1
+            ;;
+    esac
+}
+
+gitnexus-query() {
+    local query="$1"
+    local task_context="${2:-}"
+    local goal="${3:-}"
+
+    if [ -z "$query" ]; then
+        echo "Usage: gitnexus-query <query> [task_context] [goal]"
+        echo "Search the code knowledge graph for execution flows related to a concept."
+        echo ""
+        echo "Examples:"
+        echo '  gitnexus-query "authentication flow"'
+        echo '  gitnexus-query "database connection" "fixing connection pool leak"'
+        return 1
+    fi
+
+    local args="{\"query\": \"$query\""
+    [ -n "$task_context" ] && args="$args, \"task_context\": \"$task_context\""
+    [ -n "$goal" ] && args="$args, \"goal\": \"$goal\""
+    args="$args}"
+
+    _gitnexus_call query "$args"
+}
+
+gitnexus-context() {
+    local name="$1"
+    local file_path="${2:-}"
+
+    if [ -z "$name" ]; then
+        echo "Usage: gitnexus-context <symbol_name> [file_path]"
+        echo "Get a 360-degree view of a code symbol: callers, callees, processes, file location."
+        echo ""
+        echo "Examples:"
+        echo '  gitnexus-context "validateUser"'
+        echo '  gitnexus-context "AuthService" "src/auth/service.py"'
+        return 1
+    fi
+
+    local args="{\"name\": \"$name\""
+    [ -n "$file_path" ] && args="$args, \"file_path\": \"$file_path\""
+    args="$args}"
+
+    _gitnexus_call context "$args"
+}
+
+gitnexus-impact() {
+    local target="$1"
+    local direction="${2:-upstream}"
+
+    if [ -z "$target" ]; then
+        echo "Usage: gitnexus-impact <symbol_name> [upstream|downstream]"
+        echo "Analyze the blast radius of changing a code symbol."
+        echo ""
+        echo "  upstream  = what depends on this (what breaks if you change it)"
+        echo "  downstream = what this depends on (what it uses)"
+        echo ""
+        echo "Examples:"
+        echo '  gitnexus-impact "AuthService" upstream'
+        echo '  gitnexus-impact "validateUser" downstream'
+        return 1
+    fi
+
+    _gitnexus_call impact "{\"target\": \"$target\", \"direction\": \"$direction\"}"
+}
+
+gitnexus-cypher() {
+    local query="$1"
+
+    if [ -z "$query" ]; then
+        echo "Usage: gitnexus-cypher <cypher_query>"
+        echo "Execute a raw Cypher query against the code knowledge graph."
+        echo ""
+        echo "Schema: Nodes: File, Function, Class, Method, Interface, Community, Process"
+        echo "Edges via CodeRelation.type: CALLS, IMPORTS, EXTENDS, IMPLEMENTS, DEFINES, MEMBER_OF, STEP_IN_PROCESS"
+        echo ""
+        echo "Examples:"
+        echo "  gitnexus-cypher 'MATCH (a)-[:CodeRelation {type: \"CALLS\"}]->(b:Function {name: \"save\"}) RETURN a.name, a.filePath'"
+        echo "  gitnexus-cypher 'MATCH (n:Class) RETURN n.name, n.filePath LIMIT 20'"
+        return 1
+    fi
+
+    _gitnexus_call cypher "{\"query\": \"$query\"}"
+}
+
+gitnexus-overview() {
+    echo "=== Code Knowledge Graph Overview ==="
+    _gitnexus_call list_repos '{}'
+}
+
+# Export functions so they're available in subshells
+export -f _gitnexus_call 2>/dev/null
+export -f gitnexus-query 2>/dev/null
+export -f gitnexus-context 2>/dev/null
+export -f gitnexus-impact 2>/dev/null
+export -f gitnexus-cypher 2>/dev/null
+export -f gitnexus-overview 2>/dev/null

eval/bridge/mcp_bridge.py

@@ -0,0 +1,378 @@
+"""
+MCP Bridge for GitNexus
+
+Starts the GitNexus MCP server as a subprocess and provides a Python interface
+to call MCP tools. Used by the bash wrapper scripts and the augmentation layer..
+
+The bridge communicates with the MCP server via stdio using the JSON-RPC protocol.
+"""
+
+import json
+import logging
+import os
+import subprocess
+import sys
+import threading
+import time
+from pathlib import Path
+from typing import Any
+
+from constants import (
+    MCP_FIND_GITNEXUS_FALLBACK_TIMEOUT_SECONDS,
+    MCP_FIND_GITNEXUS_TIMEOUT_SECONDS,
+    MCP_READ_TIMEOUT_SECONDS,
+    MCP_STOP_WAIT_SECONDS,
+)
+from utils.errors import is_debug_enabled, log_safe_exception
+
+logger = logging.getLogger("mcp_bridge")
+
+
+class MCPBridge:
+    """
+    Manages a GitNexus MCP server subprocess and proxies tool calls to it.
+    
+    Usage:
+        bridge = MCPBridge(repo_path="/path/to/repo")
+        bridge.start()
+        result = bridge.call_tool("query", {"query": "authentication"})
+        bridge.stop()
+    """
+
+    def __init__(self, repo_path: str | None = None):
+        self.repo_path = repo_path or os.getcwd()
+        self.process: subprocess.Popen | None = None
+        self._request_id = 0
+        self._lock = threading.Lock()
+        self._started = False
+
+    def start(self) -> bool:
+        """Start the GitNexus MCP server subprocess."""
+        if self._started:
+            return True
+
+        try:
+            # Find gitnexus binary
+            gitnexus_bin = self._find_gitnexus()
+            if not gitnexus_bin:
+                logger.error("GitNexus not found. Install with: npm install -g gitnexus")
+                return False
+
+            self.process = subprocess.Popen(
+                [gitnexus_bin, "mcp"],
+                stdin=subprocess.PIPE,
+                stdout=subprocess.PIPE,
+                stderr=subprocess.PIPE,
+                cwd=self.repo_path,
+                text=False,
+            )
+
+            # Send initialize request
+            init_result = self._send_request("initialize", {
+                "protocolVersion": "2024-11-05",
+                "capabilities": {},
+                "clientInfo": {"name": "gitnexus-eval", "version": "0.1.0"},
+            })
+
+            if init_result is None:
+                logger.error("MCP server failed to initialize")
+                self.stop()
+                return False
+
+            # Send initialized notification
+            self._send_notification("notifications/initialized", {})
+            self._started = True
+            logger.info("MCP bridge started successfully")
+            return True
+
+        except Exception as e:
+            log_safe_exception(logger, "Failed to start MCP bridge", e, include_debug=is_debug_enabled())
+            self.stop()
+            return False
+
+    def stop(self):
+        """Stop the MCP server subprocess."""
+        if self.process:
+            try:
+                if self.process.stdin:
+                    self.process.stdin.close()
+                if self.process.stdout:
+                    self.process.stdout.close()
+                if self.process.stderr:
+                    self.process.stderr.close()
+                self.process.terminate()
+                self.process.wait(timeout=MCP_STOP_WAIT_SECONDS)
+            except Exception:
+                try:
+                    self.process.kill()
+                except Exception:
+                    pass
+            self.process = None
+        self._started = False
+
+    def call_tool(self, tool_name: str, arguments: dict[str, Any] | None = None) -> dict[str, Any] | None:
+        """
+        Call a GitNexus MCP tool and return the result.
+        
+        Returns the tool result content or None on error.
+        """
+        if not self._started:
+            logger.error("MCP bridge not started")
+            return None
+
+        result = self._send_request("tools/call", {
+            "name": tool_name,
+            "arguments": arguments or {},
+        })
+
+        if result is None:
+            return None
+
+        # Extract text content from MCP response
+        content = result.get("content", [])
+        if content and isinstance(content, list):
+            texts = [item.get("text", "") for item in content if item.get("type") == "text"]
+            return {"text": "\n".join(texts), "raw": content}
+
+        return {"text": "", "raw": content}
+
+    def list_tools(self) -> list[dict]:
+        """List available MCP tools."""
+        result = self._send_request("tools/list", {})
+        if result:
+            return result.get("tools", [])
+        return []
+
+    def read_resource(self, uri: str) -> str | None:
+        """Read an MCP resource by URI."""
+        result = self._send_request("resources/read", {"uri": uri})
+        if result:
+            contents = result.get("contents", [])
+            if contents:
+                return contents[0].get("text", "")
+        return None
+
+    def _find_gitnexus(self) -> str | None:
+        """Find the gitnexus CLI binary."""
+        # Check if npx is available (preferred - uses local install)
+        for cmd in ["npx"]:
+            try:
+                result = subprocess.run(
+                    [cmd, "gitnexus", "--version"],
+                    capture_output=True,
+                    text=True,
+                    timeout=MCP_FIND_GITNEXUS_TIMEOUT_SECONDS,
+                    cwd=self.repo_path,
+                )
+                if result.returncode == 0:
+                    return cmd  # Will use "npx gitnexus mcp"
+            except Exception:
+                continue
+
+        # Check for global install
+        try:
+            result = subprocess.run(
+                ["gitnexus", "--version"],
+                capture_output=True,
+                text=True,
+                timeout=MCP_FIND_GITNEXUS_FALLBACK_TIMEOUT_SECONDS,
+            )
+            if result.returncode == 0:
+                return "gitnexus"
+        except Exception:
+            pass
+
+        return None
+
+    def _next_id(self) -> int:
+        with self._lock:
+            self._request_id += 1
+            return self._request_id
+
+    def _send_request(self, method: str, params: dict) -> dict | None:
+        """Send a JSON-RPC request and wait for response."""
+        if not self.process or not self.process.stdin or not self.process.stdout:
+            return None
+
+        request_id = self._next_id()
+        request = {
+            "jsonrpc": "2.0",
+            "id": request_id,
+            "method": method,
+            "params": params,
+        }
+
+        try:
+            message = json.dumps(request)
+            # MCP uses Content-Length header framing
+            header = f"Content-Length: {len(message.encode('utf-8'))}\r\n\r\n"
+            self.process.stdin.write(header.encode("utf-8"))
+            self.process.stdin.write(message.encode("utf-8"))
+            self.process.stdin.flush()
+
+            # Read response
+            response = self._read_response(timeout=MCP_READ_TIMEOUT_SECONDS)
+            if response and response.get("id") == request_id:
+                if "error" in response:
+                    logger.error(f"MCP error: {response['error']}")
+                    return None
+                return response.get("result")
+            return None
+
+        except Exception as e:
+            log_safe_exception(logger, "MCP request failed", e, include_debug=is_debug_enabled())
+            return None
+
+    def _send_notification(self, method: str, params: dict):
+        """Send a JSON-RPC notification (no response expected)."""
+        if not self.process or not self.process.stdin:
+            return
+
+        notification = {
+            "jsonrpc": "2.0",
+            "method": method,
+            "params": params,
+        }
+
+        try:
+            message = json.dumps(notification)
+            header = f"Content-Length: {len(message.encode('utf-8'))}\r\n\r\n"
+            self.process.stdin.write(header.encode("utf-8"))
+            self.process.stdin.write(message.encode("utf-8"))
+            self.process.stdin.flush()
+        except Exception as e:
+            log_safe_exception(logger, "MCP notification failed", e, include_debug=is_debug_enabled())
+
+    def _read_content_length(self, deadline: float) -> int | None:
+        """Read Content-Length header, returning the byte length or None."""
+        if not self.process or not self.process.stdout:
+            return None
+
+        header_line = b""
+        while time.time() < deadline:
+            byte = self.process.stdout.read(1)
+            if not byte:
+                return None
+            header_line += byte
+            if header_line.endswith(b"\r\n\r\n") or header_line.endswith(b"\n\n"):
+                break
+
+        if not header_line:
+            return None
+
+        header_str = header_line.decode("utf-8").strip()
+        for line in header_str.split("\r\n"):
+            if line.lower().startswith("content-length:"):
+                try:
+                    return int(line.split(":", 1)[1].strip())
+                except (ValueError, IndexError):
+                    return None
+        return None
+
+    def _read_body(self, content_length: int, deadline: float) -> bytes | None:
+        """Read a response body of the expected length before deadline."""
+        if not self.process or not self.process.stdout:
+            return None
+
+        remaining = content_length
+        chunks: list[bytes] = []
+
+        while remaining > 0 and time.time() < deadline:
+            chunk = self.process.stdout.read(remaining)
+            if not chunk:
+                return None
+            chunks.append(chunk)
+            remaining -= len(chunk)
+
+        if remaining > 0:
+            return None
+
+        return b"".join(chunks)
+
+    def _read_response(self, timeout: float = MCP_READ_TIMEOUT_SECONDS) -> dict | None:
+        """Read a JSON-RPC response from the MCP server."""
+        if not self.process or not self.process.stdout:
+            return None
+
+        try:
+            deadline = time.time() + timeout
+            while time.time() < deadline:
+                content_length = self._read_content_length(deadline)
+                if content_length is None:
+                    continue
+
+                body = self._read_body(content_length, deadline)
+                if not body:
+                    return None
+
+                message = json.loads(body.decode("utf-8"))
+
+                # Skip notifications (no id), return responses
+                if "id" in message:
+                    return message
+
+            return None
+
+        except Exception as e:
+            log_safe_exception(logger, "Error reading MCP response", e, include_debug=is_debug_enabled())
+            return None
+
+
+class MCPToolCLI:
+    """
+    CLI wrapper that exposes MCP tools as simple command-line calls.
+    Used by the bash wrapper scripts inside Docker containers.
+    
+    Usage from bash:
+        python -m bridge.mcp_bridge query '{"query": "authentication"}'
+        python -m bridge.mcp_bridge context '{"name": "validateUser"}'
+    """
+
+    def __init__(self):
+        self.bridge = MCPBridge()
+
+    def run(self, tool_name: str, args_json: str = "{}") -> int:
+        """Run a single tool call and print the result."""
+        try:
+            args = json.loads(args_json)
+        except json.JSONDecodeError:
+            # Try to parse as simple key=value pairs
+            args = self._parse_simple_args(args_json)
+
+        if not self.bridge.start():
+            print("ERROR: Failed to start GitNexus MCP bridge", file=sys.stderr)
+            return 1
+
+        try:
+            result = self.bridge.call_tool(tool_name, args)
+            if result:
+                print(result.get("text", ""))
+                return 0
+            else:
+                print("No results", file=sys.stderr)
+                return 1
+        finally:
+            self.bridge.stop()
+
+    @staticmethod
+    def _parse_simple_args(args_str: str) -> dict:
+        """Parse 'key=value key2=value2' style arguments."""
+        args = {}
+        for part in args_str.split():
+            if "=" in part:
+                key, value = part.split("=", 1)
+                args[key] = value
+        return args
+
+
+if __name__ == "__main__":
+    if len(sys.argv) < 2:
+        print("Usage: python -m bridge.mcp_bridge <tool_name> [args_json]", file=sys.stderr)
+        print("Tools: query, context, impact, cypher, list_repos, detect_changes, rename", file=sys.stderr)
+        sys.exit(1)
+
+    tool = sys.argv[1]
+    args_json = sys.argv[2] if len(sys.argv) > 2 else "{}"
+
+    cli = MCPToolCLI()
+    sys.exit(cli.run(tool, args_json))

eval/configs/models/claude-haiku.yaml

@@ -0,0 +1,8 @@
+# Claude Haiku 4.5 — fast, cheap, good baseline
+# Via OpenRouter (set OPENROUTER_API_KEY in .env)
+model:
+  model_name: 'openrouter/anthropic/claude-haiku-4.5'
+  cost_tracking: 'ignore_errors'
+  model_kwargs:
+    max_tokens: 8192
+    temperature: 0

eval/configs/models/claude-opus.yaml

@@ -0,0 +1,9 @@
+# Claude Opus 4 — most capable, highest cost
+# Via OpenRouter (set OPENROUTER_API_KEY in .env)
+# To use Anthropic directly, change to: anthropic/claude-opus-4-20250514
+model:
+  model_name: 'openrouter/anthropic/claude-opus-4'
+  cost_tracking: 'ignore_errors'
+  model_kwargs:
+    max_tokens: 16384
+    temperature: 0

eval/configs/models/claude-sonnet.yaml

@@ -0,0 +1,9 @@
+# Claude Sonnet 4 — strong all-around model
+# Via OpenRouter (set OPENROUTER_API_KEY in .env)
+# To use Anthropic directly, change to: anthropic/claude-sonnet-4-20250514
+model:
+  model_name: 'openrouter/anthropic/claude-sonnet-4'
+  cost_tracking: 'ignore_errors'
+  model_kwargs:
+    max_tokens: 16384
+    temperature: 0

eval/configs/models/deepseek-chat.yaml

@@ -0,0 +1,13 @@
+model: deepseek-ai/deepseek-chat
+provider: openrouter
+cost:
+  input: 0.14 # per 1M tokens
+  output: 0.28 # per 1M tokens
+
+# Native DeepSeek API (direct)
+api_key: null
+base_url: null
+
+# For OpenRouter, uncomment below and comment out direct config above
+# api_key: \${OPENROUTER_API_KEY}
+# base_url: https://openrouter.ai/api/v1

eval/configs/models/deepseek-v3.yaml

@@ -0,0 +1,15 @@
+model: deepseek-ai/DeepSeek-V3
+provider: openrouter
+cost:
+  input: 0.27 # per 1M tokens
+  output: 1.10 # per 1M tokens
+
+# Native DeepSeek API (direct)
+# Get your API key at: https://platform.deepseek.com/
+# Or use OpenRouter with: OPENROUTER_API_KEY
+api_key: null
+base_url: null
+
+# For OpenRouter, uncomment below and comment out direct config above
+# api_key: \${OPENROUTER_API_KEY}
+# base_url: https://openrouter.ai/api/v1

eval/configs/models/glm-4.7.yaml

@@ -0,0 +1,7 @@
+# GLM 4.7 — via OpenRouter (set OPENROUTER_API_KEY in .env)
+model:
+  model_name: 'openrouter/zhipuai/glm-4.7'
+  cost_tracking: 'ignore_errors'
+  model_kwargs:
+    max_tokens: 8192
+    temperature: 0

eval/configs/models/glm-5.yaml

@@ -0,0 +1,7 @@
+# GLM 5 — via OpenRouter (set OPENROUTER_API_KEY in .env)
+model:
+  model_name: 'openrouter/zhipuai/glm-5'
+  cost_tracking: 'ignore_errors'
+  model_kwargs:
+    max_tokens: 8192
+    temperature: 0

eval/configs/models/minimax-2.5.yaml

@@ -0,0 +1,7 @@
+# MiniMax M1 2.5 — via OpenRouter (set OPENROUTER_API_KEY in .env)
+model:
+  model_name: 'openrouter/minimax/minimax-m1-2.5'
+  cost_tracking: 'ignore_errors'
+  model_kwargs:
+    max_tokens: 8192
+    temperature: 0