style: Action and Observation use schema for unifying (#494)

* style: Action and Observation use schema for unifying

* merge from upstream/main

* merge from upstream/main

agenthub/monologue_agent/agent.py

@@ -2,6 +2,7 @@ from typing import List
 from opendevin.agent import Agent
 from opendevin.state import State
 from opendevin.llm.llm import LLM
+from opendevin.schema import ActionType, ObservationType
 
 from opendevin.action import (
     Action,
@@ -98,7 +99,7 @@ class MonologueAgent(Agent):
 
     def _add_event(self, event: dict):
         """
-        Adds a new event to the agent's monologue and memory. 
+        Adds a new event to the agent's monologue and memory.
         Monologue automatically condenses when it gets too large.
 
         Parameters:
@@ -107,8 +108,14 @@ class MonologueAgent(Agent):
 
         if "extras" in event and "screenshot" in event["extras"]:
             del event["extras"]["screenshot"]
-        if 'args' in event and 'output' in event['args'] and len(event['args']['output']) > MAX_OUTPUT_LENGTH:
-            event['args']['output'] = event['args']['output'][:MAX_OUTPUT_LENGTH] + "..."
+        if (
+            "args" in event
+            and "output" in event["args"]
+            and len(event["args"]["output"]) > MAX_OUTPUT_LENGTH
+        ):
+            event["args"]["output"] = (
+                event["args"]["output"][:MAX_OUTPUT_LENGTH] + "..."
+            )
 
         self.monologue.add_event(event)
         self.memory.add_event(event)
@@ -121,7 +128,7 @@ class MonologueAgent(Agent):
         Short circuted to return when already initialized.
 
         Parameters:
-        - task (str): The initial goal statement provided by the user 
+        - task (str): The initial goal statement provided by the user
 
         Raises:
         - ValueError: If task is not provided
@@ -140,14 +147,18 @@ class MonologueAgent(Agent):
             thought = thought.replace("$TASK", task)
             if output_type != "":
                 observation: Observation = NullObservation(content="")
-                if output_type == "run":
-                    observation = CmdOutputObservation(content=thought, command_id=0, command="")
-                elif output_type == "read":
+                if output_type == ObservationType.RUN:
+                    observation = CmdOutputObservation(
+                        content=thought, command_id=0, command=""
+                    )
+                elif output_type == ObservationType.READ:
                     observation = FileReadObservation(content=thought, path="")
-                elif output_type == "recall":
+                elif output_type == ObservationType.RECALL:
                     observation = AgentRecallObservation(content=thought, memories=[])
-                elif output_type == "browse":
-                    observation = BrowserOutputObservation(content=thought, url="", screenshot="")
+                elif output_type == ObservationType.BROWSE:
+                    observation = BrowserOutputObservation(
+                        content=thought, url="", screenshot=""
+                    )
                 self._add_event(observation.to_dict())
                 output_type = ""
             else:
@@ -155,7 +166,7 @@ class MonologueAgent(Agent):
                 if thought.startswith("RUN"):
                     command = thought.split("RUN ")[1]
                     action = CmdRunAction(command)
-                    output_type = "run"
+                    output_type = ActionType.RUN
                 elif thought.startswith("WRITE"):
                     parts = thought.split("WRITE ")[1].split(" > ")
                     path = parts[1]
@@ -164,15 +175,15 @@ class MonologueAgent(Agent):
                 elif thought.startswith("READ"):
                     path = thought.split("READ ")[1]
                     action = FileReadAction(path=path)
-                    output_type = "read"
+                    output_type = ActionType.READ
                 elif thought.startswith("RECALL"):
                     query = thought.split("RECALL ")[1]
                     action = AgentRecallAction(query=query)
-                    output_type = "recall"
+                    output_type = ActionType.RECALL
                 elif thought.startswith("BROWSE"):
                     url = thought.split("BROWSE ")[1]
                     action = BrowseURLAction(url=url)
-                    output_type = "browse"
+                    output_type = ActionType.BROWSE
                 else:
                     action = AgentThinkAction(thought=thought)
                 self._add_event(action.to_dict())
@@ -200,9 +211,9 @@ class MonologueAgent(Agent):
             self.monologue.get_thoughts(),
             state.background_commands_obs,
         )
-        messages = [{"content": prompt,"role": "user"}]
+        messages = [{"content": prompt, "role": "user"}]
         resp = self.llm.completion(messages=messages)
-        action_resp = resp['choices'][0]['message']['content']
+        action_resp = resp["choices"][0]["message"]["content"]
         action = prompts.parse_action_response(action_resp)
         self.latest_action = action
         return action
@@ -211,7 +222,7 @@ class MonologueAgent(Agent):
         """
         Uses VectorIndexRetriever to find related memories within the long term memory.
         Uses search to produce top 10 results.
-        
+
         Parameters:
         - query (str): The query that we want to find related memories for
 
@@ -219,4 +230,3 @@ class MonologueAgent(Agent):
         - List[str]: A list of top 10 text results that matched the query
         """
         return self.memory.search(query)
-

agenthub/planner_agent/prompt.py

@@ -5,6 +5,7 @@ from opendevin.controller.agent_controller import print_with_color
 from opendevin.plan import Plan
 from opendevin.action import Action, action_from_dict
 from opendevin.observation import Observation
+from opendevin.schema import ActionType
 
 from opendevin.action import (
     NullAction,
@@ -26,17 +27,17 @@ from opendevin.observation import (
 )
 
 ACTION_TYPE_TO_CLASS: Dict[str, Type[Action]] = {
-    "run": CmdRunAction,
-    "kill": CmdKillAction,
-    "browse": BrowseURLAction,
-    "read": FileReadAction,
-    "write": FileWriteAction,
-    "recall": AgentRecallAction,
-    "think": AgentThinkAction,
-    "summarize": AgentSummarizeAction,
-    "finish": AgentFinishAction,
-    "add_task": AddTaskAction,
-    "modify_task": ModifyTaskAction,
+    ActionType.RUN: CmdRunAction,
+    ActionType.KILL: CmdKillAction,
+    ActionType.BROWSE: BrowseURLAction,
+    ActionType.READ: FileReadAction,
+    ActionType.WRITE: FileWriteAction,
+    ActionType.RECALL: AgentRecallAction,
+    ActionType.THINK: AgentThinkAction,
+    ActionType.SUMMARIZE: AgentSummarizeAction,
+    ActionType.FINISH: AgentFinishAction,
+    ActionType.ADD_TASK: AddTaskAction,
+    ActionType.MODIFY_TASK: ModifyTaskAction,
 }
 
 HISTORY_SIZE = 10
@@ -129,9 +130,10 @@ What is your next thought or action? Again, you must reply with JSON, and only w
 %(hint)s
 """
 
+
 def get_prompt(plan: Plan, history: List[Tuple[Action, Observation]]) -> str:
     """
-    Gets the prompt for the planner agent. 
+    Gets the prompt for the planner agent.
     Formatted with the most recent action-observation pairs, current task, and hint based on last action
 
     Parameters:
@@ -152,7 +154,10 @@ def get_prompt(plan: Plan, history: List[Tuple[Action, Observation]]) -> str:
             latest_action = action
         if not isinstance(observation, NullObservation):
             observation_dict = observation.to_dict()
-            if "extras" in observation_dict and "screenshot" in observation_dict["extras"]:
+            if (
+                "extras" in observation_dict
+                and "screenshot" in observation_dict["extras"]
+            ):
                 del observation_dict["extras"]["screenshot"]
             history_dicts.append(observation_dict)
     history_str = json.dumps(history_dicts, indent=2)
@@ -167,41 +172,42 @@ def get_prompt(plan: Plan, history: List[Tuple[Action, Observation]]) -> str:
         plan_status = "You're not currently working on any tasks. Your next action MUST be to mark a task as in_progress."
         hint = plan_status
 
-    latest_action_id = latest_action.to_dict()['action']
+    latest_action_id = latest_action.to_dict()["action"]
 
     if current_task is not None:
         if latest_action_id == "":
             hint = "You haven't taken any actions yet. Start by using `ls` to check out what files you're working with."
-        elif latest_action_id == "run":
+        elif latest_action_id == ActionType.RUN:
             hint = "You should think about the command you just ran, what output it gave, and how that affects your plan."
-        elif latest_action_id == "read":
+        elif latest_action_id == ActionType.READ:
             hint = "You should think about the file you just read, what you learned from it, and how that affects your plan."
-        elif latest_action_id == "write":
+        elif latest_action_id == ActionType.WRITE:
             hint = "You just changed a file. You should think about how it affects your plan."
-        elif latest_action_id == "browse":
+        elif latest_action_id == ActionType.BROWSE:
             hint = "You should think about the page you just visited, and what you learned from it."
-        elif latest_action_id == "think":
+        elif latest_action_id == ActionType.THINK:
             hint = "Look at your last thought in the history above. What does it suggest? Don't think anymore--take action."
-        elif latest_action_id == "recall":
+        elif latest_action_id == ActionType.RECALL:
             hint = "You should think about the information you just recalled, and how it should affect your plan."
-        elif latest_action_id == "add_task":
+        elif latest_action_id == ActionType.ADD_TASK:
             hint = "You should think about the next action to take."
-        elif latest_action_id == "modify_task":
+        elif latest_action_id == ActionType.MODIFY_TASK:
             hint = "You should think about the next action to take."
-        elif latest_action_id == "summarize":
+        elif latest_action_id == ActionType.SUMMARIZE:
             hint = ""
-        elif latest_action_id == "finish":
+        elif latest_action_id == ActionType.FINISH:
             hint = ""
 
     print_with_color("HINT:\n" + hint, "INFO")
     return prompt % {
-        'task': plan.main_goal,
-        'plan': plan_str,
-        'history': history_str,
-        'hint': hint,
-        'plan_status': plan_status,
+        "task": plan.main_goal,
+        "plan": plan_str,
+        "history": history_str,
+        "hint": hint,
+        "plan_status": plan_status,
     }
 
+
 def parse_response(response: str) -> Action:
     """
     Parses the model output to find a valid action to take
@@ -216,9 +222,8 @@ def parse_response(response: str) -> Action:
     json_end = response.rfind("}") + 1
     response = response[json_start:json_end]
     action_dict = json.loads(response)
-    if 'contents' in action_dict:
+    if "contents" in action_dict:
         # The LLM gets confused here. Might as well be robust
-        action_dict['content'] = action_dict.pop('contents')
+        action_dict["content"] = action_dict.pop("contents")
     action = action_from_dict(action_dict)
     return action
-

frontend/src/components/Terminal.tsx

@@ -3,6 +3,8 @@ import { IDisposable, Terminal as XtermTerminal } from "@xterm/xterm";
 import "@xterm/xterm/css/xterm.css";
 import { useSelector } from "react-redux";
 import { FitAddon } from "xterm-addon-fit";
+import ActionType from "../types/ActionType";
+import ObservationType from "../types/ObservationType";
 import Socket from "../services/socket";
 import { RootState } from "../store";
 
@@ -22,10 +24,10 @@ class JsonWebsocketAddon {
     );
     Socket.addEventListener("message", (event) => {
       const { action, args, observation, content } = JSON.parse(event.data);
-      if (action === "run") {
+      if (action === ActionType.RUN) {
         terminal.writeln(args.command);
       }
-      if (observation === "run") {
+      if (observation === ObservationType.RUN) {
         content.split("\n").forEach((line: string) => {
           terminal.writeln(line);
         });

frontend/src/services/actions.ts

@@ -7,11 +7,12 @@ import store from "../store";
 import { ActionMessage } from "../types/Message";
 import { SocketMessage } from "../types/ResponseType";
 import { handleObservationMessage } from "./observations";
+import ActionType from "../types/ActionType";
 
 let isInitialized = false;
 
 const messageActions = {
-  initialize: () => {
+  [ActionType.INIT]: () => {
     store.dispatch(setInitialized(true));
     if (isInitialized) {
       return;
@@ -23,23 +24,23 @@ const messageActions = {
     );
     isInitialized = true;
   },
-  browse: (message: ActionMessage) => {
+  [ActionType.BROWSE]: (message: ActionMessage) => {
     const { url, screenshotSrc } = message.args;
     store.dispatch(setUrl(url));
     store.dispatch(setScreenshotSrc(screenshotSrc));
   },
-  write: (message: ActionMessage) => {
+  [ActionType.WRITE]: (message: ActionMessage) => {
     const { path, content } = message.args;
     store.dispatch(updatePath(path));
     store.dispatch(setCode(content));
   },
-  think: (message: ActionMessage) => {
+  [ActionType.THINK]: (message: ActionMessage) => {
     store.dispatch(appendAssistantMessage(message.args.thought));
   },
-  finish: (message: ActionMessage) => {
+  [ActionType.FINISH]: (message: ActionMessage) => {
     store.dispatch(appendAssistantMessage(message.message));
   },
-  run: (message: ActionMessage) => {
+  [ActionType.RUN]: (message: ActionMessage) => {
     store.dispatch(appendInput(message.args.command));
   },
 };

frontend/src/services/chatService.ts

@@ -9,12 +9,13 @@ import {
 } from "../state/chatSlice";
 import Socket from "./socket";
 import store from "../store";
+import ActionType from "../types/ActionType";
 import { SocketMessage } from "../types/ResponseType";
 import { ActionMessage } from "../types/Message";
 
 export function sendChatMessage(message: string): void {
   store.dispatch(appendUserMessage(message));
-  const event = { action: "start", args: { task: message } };
+  const event = { action: ActionType.START, args: { task: message } };
   const eventString = JSON.stringify(event);
   Socket.send(eventString);
 }

frontend/src/services/settingsService.ts

@@ -1,6 +1,7 @@
 import { appendAssistantMessage } from "../state/chatSlice";
 import { setInitialized } from "../state/taskSlice";
 import store from "../store";
+import ActionType from "../types/ActionType";
 import Socket from "./socket";
 import {
   setAgent,
@@ -61,7 +62,7 @@ export function saveSettings(
         value,
       ]),
     );
-    const event = { action: "initialize", args: socketSettings };
+    const event = { action: ActionType.INIT, args: socketSettings };
     const eventString = JSON.stringify(event);
     store.dispatch(setInitialized(false));
     Socket.send(eventString);

opendevin/action/__init__.py

@@ -2,7 +2,13 @@ from .base import Action, NullAction
 from .bash import CmdRunAction, CmdKillAction
 from .browse import BrowseURLAction
 from .fileop import FileReadAction, FileWriteAction
-from .agent import AgentRecallAction, AgentThinkAction, AgentFinishAction, AgentEchoAction, AgentSummarizeAction
+from .agent import (
+    AgentRecallAction,
+    AgentThinkAction,
+    AgentFinishAction,
+    AgentEchoAction,
+    AgentSummarizeAction,
+)
 from .tasks import AddTaskAction, ModifyTaskAction
 
 actions = (
@@ -18,7 +24,8 @@ actions = (
     ModifyTaskAction,
 )
 
-ACTION_TYPE_TO_CLASS = {action_class.action:action_class for action_class in actions} # type: ignore[attr-defined]
+ACTION_TYPE_TO_CLASS = {action_class.action: action_class for action_class in actions}  # type: ignore[attr-defined]
+
 
 def action_from_dict(action: dict) -> Action:
     action = action.copy()
@@ -26,10 +33,13 @@ def action_from_dict(action: dict) -> Action:
         raise KeyError(f"'action' key is not found in {action=}")
     action_class = ACTION_TYPE_TO_CLASS.get(action["action"])
     if action_class is None:
-        raise KeyError(f"'{action['action']=}' is not defined. Available actions: {ACTION_TYPE_TO_CLASS.keys()}")
+        raise KeyError(
+            f"'{action['action']=}' is not defined. Available actions: {ACTION_TYPE_TO_CLASS.keys()}"
+        )
     args = action.get("args", {})
     return action_class(**args)
 
+
 __all__ = [
     "Action",
     "NullAction",
@@ -44,5 +54,5 @@ __all__ = [
     "AgentEchoAction",
     "AgentSummarizeAction",
     "AddTaskAction",
-    "ModifyTaskAction"
+    "ModifyTaskAction",
 ]

opendevin/action/agent.py

@@ -1,8 +1,14 @@
 from dataclasses import dataclass
 from typing import TYPE_CHECKING
 
-from opendevin.observation import AgentRecallObservation, AgentMessageObservation, Observation
+from opendevin.observation import (
+    AgentRecallObservation,
+    AgentMessageObservation,
+    Observation,
+)
+from opendevin.schema import ActionType
 from .base import ExecutableAction, NotExecutableAction
+
 if TYPE_CHECKING:
     from opendevin.controller import AgentController
 
@@ -10,22 +16,23 @@ if TYPE_CHECKING:
 @dataclass
 class AgentRecallAction(ExecutableAction):
     query: str
-    action: str = "recall"
+    action: str = ActionType.RECALL
 
     def run(self, controller: "AgentController") -> AgentRecallObservation:
         return AgentRecallObservation(
             content="Recalling memories...",
-            memories=controller.agent.search_memory(self.query)
+            memories=controller.agent.search_memory(self.query),
         )
 
     @property
     def message(self) -> str:
         return f"Let me dive into my memories to find what you're looking for! Searching for: '{self.query}'. This might take a moment."
 
+
 @dataclass
 class AgentThinkAction(NotExecutableAction):
     thought: str
-    action: str = "think"
+    action: str = ActionType.THINK
 
     def run(self, controller: "AgentController") -> "Observation":
         raise NotImplementedError
@@ -34,6 +41,7 @@ class AgentThinkAction(NotExecutableAction):
     def message(self) -> str:
         return self.thought
 
+
 @dataclass
 class AgentEchoAction(ExecutableAction):
     content: str
@@ -46,19 +54,21 @@ class AgentEchoAction(ExecutableAction):
     def message(self) -> str:
         return self.content
 
+
 @dataclass
 class AgentSummarizeAction(NotExecutableAction):
     summary: str
 
-    action: str = "summarize"
+    action: str = ActionType.SUMMARIZE
 
     @property
     def message(self) -> str:
         return self.summary
 
+
 @dataclass
 class AgentFinishAction(NotExecutableAction):
-    action: str = "finish"
+    action: str = ActionType.FINISH
 
     def run(self, controller: "AgentController") -> "Observation":
         raise NotImplementedError

opendevin/action/base.py

@@ -1,10 +1,12 @@
 from dataclasses import dataclass, asdict
 from typing import TYPE_CHECKING
+from opendevin.schema import ActionType
 
 if TYPE_CHECKING:
     from opendevin.controller import AgentController
     from opendevin.observation import Observation
 
+
 @dataclass
 class Action:
     def run(self, controller: "AgentController") -> "Observation":
@@ -13,10 +15,10 @@ class Action:
     def to_dict(self):
         d = asdict(self)
         try:
-            v = d.pop('action')
+            v = d.pop("action")
         except KeyError:
-            raise NotImplementedError(f'{self=} does not have action attribute set')
-        return {'action': v, "args": d, "message": self.message}
+            raise NotImplementedError(f"{self=} does not have action attribute set")
+        return {"action": v, "args": d, "message": self.message}
 
     @property
     def executable(self) -> bool:
@@ -26,6 +28,7 @@ class Action:
     def message(self) -> str:
         raise NotImplementedError
 
+
 @dataclass
 class ExecutableAction(Action):
     @property
@@ -39,12 +42,14 @@ class NotExecutableAction(Action):
     def executable(self) -> bool:
         return False
 
+
 @dataclass
 class NullAction(NotExecutableAction):
     """An action that does nothing.
     This is used when the agent need to receive user follow-up messages from the frontend.
     """
-    action: str = "null"
+
+    action: str = ActionType.NULL
 
     @property
     def message(self) -> str:

opendevin/action/bash.py

@@ -2,6 +2,7 @@ from dataclasses import dataclass
 from typing import TYPE_CHECKING
 
 from .base import ExecutableAction
+from opendevin.schema import ActionType
 
 if TYPE_CHECKING:
     from opendevin.controller import AgentController
@@ -12,7 +13,7 @@ if TYPE_CHECKING:
 class CmdRunAction(ExecutableAction):
     command: str
     background: bool = False
-    action: str = "run"
+    action: str = ActionType.RUN
 
     def run(self, controller: "AgentController") -> "CmdOutputObservation":
         return controller.command_manager.run_command(self.command, self.background)
@@ -21,14 +22,15 @@ class CmdRunAction(ExecutableAction):
     def message(self) -> str:
         return f"Running command: {self.command}"
 
+
 @dataclass
 class CmdKillAction(ExecutableAction):
     id: int
-    action: str = "kill"
+    action: str = ActionType.KILL
 
     def run(self, controller: "AgentController") -> "CmdOutputObservation":
         return controller.command_manager.kill_command(self.id)
 
     @property
     def message(self) -> str:
-        return f"Killing command: {self.id}"
+        return f"Killing command: {self.id}"

opendevin/action/browse.py

@@ -1,6 +1,7 @@
 import base64
 from dataclasses import dataclass
 from opendevin.observation import BrowserOutputObservation
+from opendevin.schema import ActionType
 from typing import TYPE_CHECKING
 from playwright.async_api import async_playwright
 
@@ -9,12 +10,13 @@ from .base import ExecutableAction
 if TYPE_CHECKING:
     from opendevin.controller import AgentController
 
+
 @dataclass
 class BrowseURLAction(ExecutableAction):
     url: str
-    action: str = "browse"
+    action: str = ActionType.BROWSE
 
-    async def run(self, controller: "AgentController") -> BrowserOutputObservation: # type: ignore
+    async def run(self, controller: "AgentController") -> BrowserOutputObservation:  # type: ignore
         try:
             async with async_playwright() as p:
                 browser = await p.chromium.launch()
@@ -34,12 +36,9 @@ class BrowseURLAction(ExecutableAction):
                 )
         except Exception as e:
             return BrowserOutputObservation(
-                content=str(e),
-                screenshot="", 
-                error=True,
-                url=self.url
+                content=str(e), screenshot="", error=True, url=self.url
             )
-        
+
     @property
     def message(self) -> str:
-        return f"Browsing URL: {self.url}"
+        return f"Browsing URL: {self.url}"

opendevin/action/fileop.py

@@ -2,6 +2,7 @@ import os
 from dataclasses import dataclass
 
 from opendevin.observation import FileReadObservation, FileWriteObservation
+from opendevin.schema import ActionType
 
 from .base import ExecutableAction
 
@@ -19,7 +20,7 @@ def resolve_path(base_path, file_path):
 @dataclass
 class FileReadAction(ExecutableAction):
     path: str
-    action: str = "read"
+    action: str = ActionType.READ
 
     def run(self, controller) -> FileReadObservation:
         path = resolve_path(controller.workdir, self.path)
@@ -35,7 +36,7 @@ class FileReadAction(ExecutableAction):
 class FileWriteAction(ExecutableAction):
     path: str
     content: str
-    action: str = "write"
+    action: str = ActionType.WRITE
 
     def run(self, controller) -> FileWriteObservation:
         whole_path = resolve_path(controller.workdir, self.path)

opendevin/action/tasks.py

@@ -1,25 +1,27 @@
 from dataclasses import dataclass, field
 
 from .base import NotExecutableAction
+from opendevin.schema import ActionType
+
 
 @dataclass
 class AddTaskAction(NotExecutableAction):
     parent: str
     goal: str
     subtasks: list = field(default_factory=list)
-    action: str = "add_task"
+    action: str = ActionType.ADD_TASK
 
     @property
     def message(self) -> str:
         return f"Added task: {self.goal}"
 
+
 @dataclass
 class ModifyTaskAction(NotExecutableAction):
     id: str
     state: str
-    action: str = "modify_task"
+    action: str = ActionType.MODIFY_TASK
 
     @property
     def message(self) -> str:
         return f"Set task {self.id} to {self.state}"
-

opendevin/mock/listen.py

@@ -1,13 +1,18 @@
 import uvicorn
 from fastapi import FastAPI, WebSocket
+from opendevin.schema import ActionType
 
 app = FastAPI()
+
+
 @app.websocket("/ws")
 async def websocket_endpoint(websocket: WebSocket):
     await websocket.accept()
     # send message to mock connection
-    await websocket.send_json({"action": "initialize", "message": "Control loop started."})
-    
+    await websocket.send_json(
+        {"action": ActionType.INIT, "message": "Control loop started."}
+    )
+
     try:
         while True:
             # receive message
@@ -21,10 +26,12 @@ async def websocket_endpoint(websocket: WebSocket):
     except Exception as e:
         print(f"WebSocket Error: {e}")
 
+
 @app.get("/")
 def read_root():
     return {"message": "This is a mock server"}
 
+
 @app.get("/litellm-models")
 def read_llm_models():
     return [
@@ -34,6 +41,7 @@ def read_llm_models():
         "gpt-4-0613",
     ]
 
+
 @app.get("/litellm-agents")
 def read_llm_agents():
     return [
@@ -42,9 +50,11 @@ def read_llm_agents():
         "PlannerAgent",
     ]
 
+
 @app.get("/default-model")
 def read_default_model():
     return "gpt-4"
 
+
 if __name__ == "__main__":
     uvicorn.run(app, host="127.0.0.1", port=3000)

opendevin/observation/base.py

@@ -1,5 +1,7 @@
 import copy
 from dataclasses import dataclass
+from opendevin.schema import ObservationType
+
 
 @dataclass
 class Observation:
@@ -36,7 +38,8 @@ class NullObservation(Observation):
     This data class represents a null observation.
     This is used when the produced action is NOT executable.
     """
-    observation : str = "null"
+
+    observation: str = ObservationType.NULL
 
     @property
     def message(self) -> str:

opendevin/observation/browse.py

@@ -1,6 +1,8 @@
 from dataclasses import dataclass
 
 from .base import Observation
+from opendevin.schema import ObservationType
+
 
 @dataclass
 class BrowserOutputObservation(Observation):
@@ -12,11 +14,8 @@ class BrowserOutputObservation(Observation):
     screenshot: str
     status_code: int = 200
     error: bool = False
-    observation : str = "browse"
+    observation: str = ObservationType.BROWSE
 
     @property
     def message(self) -> str:
         return "Visited " + self.url
-
-
-

opendevin/observation/error.py

@@ -1,16 +1,17 @@
 from dataclasses import dataclass
 
 from .base import Observation
+from opendevin.schema import ObservationType
+
 
 @dataclass
 class AgentErrorObservation(Observation):
     """
     This data class represents an error encountered by the agent.
     """
-    observation : str = "error"
+
+    observation: str = ObservationType.ERROR
 
     @property
     def message(self) -> str:
         return "Oops. Something went wrong: " + self.content
-
-

opendevin/observation/files.py

@@ -1,6 +1,8 @@
 from dataclasses import dataclass
 
 from .base import Observation
+from opendevin.schema import ObservationType
+
 
 @dataclass
 class FileReadObservation(Observation):
@@ -9,12 +11,13 @@ class FileReadObservation(Observation):
     """
 
     path: str
-    observation : str = "read"
+    observation: str = ObservationType.READ
 
     @property
     def message(self) -> str:
         return f"I read the file {self.path}."
 
+
 @dataclass
 class FileWriteObservation(Observation):
     """
@@ -22,10 +25,8 @@ class FileWriteObservation(Observation):
     """
 
     path: str
-    observation : str = "write"
+    observation: str = ObservationType.WRITE
 
     @property
     def message(self) -> str:
         return f"I wrote to the file {self.path}."
-
-

opendevin/observation/message.py

@@ -1,6 +1,8 @@
 from dataclasses import dataclass
 
 from .base import Observation
+from opendevin.schema import ObservationType
+
 
 @dataclass
 class UserMessageObservation(Observation):
@@ -9,7 +11,7 @@ class UserMessageObservation(Observation):
     """
 
     role: str = "user"
-    observation : str = "message"
+    observation: str = ObservationType.MESSAGE
 
     @property
     def message(self) -> str:
@@ -23,11 +25,8 @@ class AgentMessageObservation(Observation):
     """
 
     role: str = "assistant"
-    observation : str = "message"
+    observation: str = ObservationType.MESSAGE
 
     @property
     def message(self) -> str:
         return ""
-
-
-

opendevin/observation/recall.py

@@ -2,6 +2,8 @@ from dataclasses import dataclass
 from typing import List
 
 from .base import Observation
+from opendevin.schema import ObservationType
+
 
 @dataclass
 class AgentRecallObservation(Observation):
@@ -11,11 +13,8 @@ class AgentRecallObservation(Observation):
 
     memories: List[str]
     role: str = "assistant"
-    observation : str = "recall"
+    observation: str = ObservationType.RECALL
 
     @property
     def message(self) -> str:
         return "The agent recalled memories."
-
-
-

opendevin/observation/run.py

@@ -1,6 +1,8 @@
 from dataclasses import dataclass
 
 from .base import Observation
+from opendevin.schema import ObservationType
+
 
 @dataclass
 class CmdOutputObservation(Observation):
@@ -11,7 +13,7 @@ class CmdOutputObservation(Observation):
     command_id: int
     command: str
     exit_code: int = 0
-    observation : str = "run"
+    observation: str = ObservationType.RUN
 
     @property
     def error(self) -> bool:
@@ -19,6 +21,4 @@ class CmdOutputObservation(Observation):
 
     @property
     def message(self) -> str:
-        return f'Command `{self.command}` executed with exit code {self.exit_code}.'
-
-
+        return f"Command `{self.command}` executed with exit code {self.exit_code}."

opendevin/schema/action.py

@@ -42,3 +42,13 @@ class ActionType(str, Enum):
     """If you're absolutely certain that you've completed your task and have tested your work, 
     use the finish action to stop working.
     """
+
+    CHAT = "chat"
+
+    SUMMARIZE = "summarize"
+
+    ADD_TASK = "add_task"
+
+    MODIFY_TASK = "modify_task"
+
+    NULL = "null"

opendevin/schema/observation.py

@@ -6,6 +6,8 @@ class ObservationType(str, Enum):
     """The content of a file
     """
 
+    WRITE = "write"
+
     BROWSE = "browse"
     """The HTML content of a URL
     """
@@ -21,3 +23,9 @@ class ObservationType(str, Enum):
     CHAT = "chat"
     """A message from the user
     """
+
+    MESSAGE = "message"
+
+    ERROR = "error"
+
+    NULL = "null"

opendevin/server/agent/manager.py

@@ -12,6 +12,7 @@ from opendevin.controller import AgentController
 from opendevin.llm.llm import LLM
 from opendevin.observation import NullObservation, Observation, UserMessageObservation
 from opendevin.server.session import session_manager
+from opendevin.schema import ActionType
 
 DEFAULT_API_KEY = config.get("LLM_API_KEY")
 DEFAULT_BASE_URL = config.get("LLM_BASE_URL")
@@ -69,14 +70,14 @@ class AgentManager:
             await self.send_error("Invalid action")
             return
 
-        if action == "initialize":
+        if action == ActionType.INIT:
             await self.create_controller(data)
-        elif action == "start":
+        elif action == ActionType.START:
             await self.start_task(data)
         else:
             if self.controller is None:
                 await self.send_error("No agent started. Please wait a second...")
-            elif action == "chat":
+            elif action == ActionType.CHAT:
                 self.controller.add_history(
                     NullAction(), UserMessageObservation(data["message"])
                 )
@@ -141,7 +142,7 @@ class AgentManager:
                 "Error creating controller. Please check Docker is running using `docker ps`."
             )
             return
-        await self.send({"action": "initialize", "message": "Control loop started."})
+        await self.send({"action": ActionType.INIT, "message": "Control loop started."})
 
     async def start_task(self, start_event):
         """Starts a task for the agent.

opendevin/server/session/msg_stack.py

@@ -5,7 +5,7 @@ import signal
 import uuid
 from typing import Dict, List
 
-from opendevin.server.schema.action import ActionType
+from opendevin.schema.action import ActionType
 
 
 CACHE_DIR = os.getenv("CACHE_DIR", "cache")