Unbound Integration

+

remove my defaults & dotenv version
squash with model list
minor fixes

.env.example

@@ -24,6 +24,9 @@ ALIBABA_API_KEY=
 MOONSHOT_ENDPOINT=https://api.moonshot.cn/v1
 MOONSHOT_API_KEY=
 
+UNBOUND_ENDPOINT=https://api.getunbound.ai
+UNBOUND_API_KEY=
+
 # Set to false to disable anonymized telemetry
 ANONYMIZED_TELEMETRY=false
 

requirements.txt

@@ -4,4 +4,4 @@ gradio==5.10.0
 json-repair
 langchain-mistralai==0.2.4
 langchain-google-genai==2.0.8
-MainContentExtractor==0.0.4
+MainContentExtractor==0.0.4

src/utils/llm.py

@@ -1,5 +1,7 @@
 from openai import OpenAI
 import pdb
+import os
+from dotenv import load_dotenv
 from langchain_openai import ChatOpenAI
 from langchain_core.globals import get_llm_cache
 from langchain_core.language_models.base import (
@@ -29,6 +31,9 @@ from langchain_ollama import ChatOllama
 from langchain_core.output_parsers.base import OutputParserLike
 from langchain_core.runnables import Runnable, RunnableConfig
 from langchain_core.tools import BaseTool
+from pydantic import Field, PrivateAttr
+import requests
+import urllib3
 
 from typing import (
     TYPE_CHECKING,
@@ -103,6 +108,72 @@ class DeepSeekR1ChatOpenAI(ChatOpenAI):
         return AIMessage(content=content, reasoning_content=reasoning_content)
 
 
+# Load environment variables
+load_dotenv()
+
+class UnboundChatOpenAI(ChatOpenAI):
+    """Chat model that uses Unbound's API."""
+    
+    _session: requests.Session = PrivateAttr()
+    
+    def __init__(self, *args: Any, **kwargs: Any) -> None:
+        kwargs["base_url"] = kwargs.get("base_url", os.getenv("UNBOUND_ENDPOINT", "https://api.getunbound.ai"))
+        kwargs["api_key"] = kwargs.get("api_key", os.getenv("UNBOUND_API_KEY"))
+        if not kwargs["api_key"]:
+            raise ValueError("UNBOUND_API_KEY environment variable is not set")
+        super().__init__(*args, **kwargs)
+        
+        self.client = OpenAI(
+            base_url=kwargs["base_url"],
+            api_key=kwargs["api_key"]
+        )
+        
+        urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning)
+        
+        self._session = requests.Session()
+        self._session.verify = False
+        
+    def invoke(
+        self,
+        input: LanguageModelInput,
+        config: Optional[RunnableConfig] = None,
+        *,
+        stop: Optional[list[str]] = None,
+        **kwargs: Any,
+    ) -> AIMessage:
+        message_history = []
+        for input_ in input:
+            if isinstance(input_, SystemMessage):
+                message_history.append({"role": "system", "content": input_.content})
+            elif isinstance(input_, AIMessage):
+                message_history.append({"role": "assistant", "content": input_.content})
+            else:
+                message_history.append({"role": "user", "content": input_.content})
+        
+        response = self._session.post(
+            f"{self.client.base_url}/v1/chat/completions",
+            headers={"Authorization": f"Bearer {self.client.api_key}", "Content-Type": "application/json"},
+            json={
+                "model": self.model_name or "gpt-4o-mini",
+                "messages": message_history
+            }
+        )
+        response.raise_for_status()
+        data = response.json()
+        content = data["choices"][0]["message"]["content"]
+        return AIMessage(content=content)
+    
+    async def ainvoke(
+        self,
+        input: LanguageModelInput,
+        config: Optional[RunnableConfig] = None,
+        *,
+        stop: Optional[list[str]] = None,
+        **kwargs: Any,
+    ) -> AIMessage:
+        return self.invoke(input, config, stop=stop, **kwargs)
+
+
 class DeepSeekR1ChatOllama(ChatOllama):
 
     async def ainvoke(

src/utils/utils.py

@@ -12,7 +12,7 @@ from langchain_ollama import ChatOllama
 from langchain_openai import AzureChatOpenAI, ChatOpenAI
 import gradio as gr
 
-from .llm import DeepSeekR1ChatOpenAI, DeepSeekR1ChatOllama
+from .llm import DeepSeekR1ChatOpenAI, DeepSeekR1ChatOllama, UnboundChatOpenAI
 
 PROVIDER_DISPLAY_NAMES = {
     "openai": "OpenAI",
@@ -21,7 +21,8 @@ PROVIDER_DISPLAY_NAMES = {
     "deepseek": "DeepSeek",
     "google": "Google",
     "alibaba": "Alibaba",
-    "moonshot": "MoonShot"
+    "moonshot": "MoonShot",
+    "unbound": "Unbound AI"
 }
 
 
@@ -151,7 +152,6 @@ def get_llm_model(provider: str, **kwargs):
             base_url=base_url,
             api_key=api_key,
         )
-
     elif provider == "moonshot":
         return ChatOpenAI(
             model=kwargs.get("model_name", "moonshot-v1-32k-vision-preview"),
@@ -159,6 +159,18 @@ def get_llm_model(provider: str, **kwargs):
             base_url=os.getenv("MOONSHOT_ENDPOINT"),
             api_key=os.getenv("MOONSHOT_API_KEY"),
         )
+    elif provider == "unbound":
+        if not kwargs.get("base_url", ""):
+            base_url = os.getenv("UNBOUND_ENDPOINT", "https://api.getunbound.ai")
+        else:
+            base_url = kwargs.get("base_url")
+
+        return UnboundChatOpenAI(
+            model=kwargs.get("model_name", "gpt-4o-mini"),
+            temperature=kwargs.get("temperature", 0.0),
+            base_url=base_url,
+            api_key=api_key,
+        )
     else:
         raise ValueError(f"Unsupported provider: {provider}")
 
@@ -176,6 +188,7 @@ model_names = {
     "mistral": ["mixtral-large-latest", "mistral-large-latest", "mistral-small-latest", "ministral-8b-latest"],
     "alibaba": ["qwen-plus", "qwen-max", "qwen-turbo", "qwen-long"],
     "moonshot": ["moonshot-v1-32k-vision-preview", "moonshot-v1-8k-vision-preview"],
+    "unbound": ["gemini-2.0-flash","gpt-4o-mini", "gpt-4o", "gpt-4.5-preview"]
 }