Backend MVP: FastAPI + Claude integration

Core components:
- FastAPI server with health endpoints
- AI client (Anthropic Claude integration)
- Context manager (loads company knowledge)
- Chat API (non-streaming and streaming)
- Requirements and environment setup

Ready to run: python backend/main.py

backend/api/chat.py

@@ -0,0 +1,137 @@
+"""
+Chat API - Main endpoint for interacting with Grimlock
+"""
+
+from fastapi import APIRouter, Depends, HTTPException
+from fastapi.responses import StreamingResponse
+from pydantic import BaseModel
+from typing import List, Optional
+import logging
+
+from core.context_manager import ContextManager
+from core.ai_client import AIClient
+import main
+
+logger = logging.getLogger(__name__)
+
+router = APIRouter()
+
+class Message(BaseModel):
+    role: str
+    content: str
+
+class ChatRequest(BaseModel):
+    messages: List[Message]
+    role: Optional[str] = None
+    stream: bool = False
+
+class ChatResponse(BaseModel):
+    response: str
+    context_used: bool
+
+@router.post("/", response_model=ChatResponse)
+async def chat(
+    request: ChatRequest,
+    context_manager: ContextManager = Depends(main.get_context_manager),
+    ai_client: AIClient = Depends(main.get_ai_client)
+):
+    """
+    Chat with Grimlock
+    
+    Args:
+        request: Chat request with messages and optional role
+        
+    Returns:
+        ChatResponse with AI response
+    """
+    try:
+        # Get the last user message for context
+        user_message = None
+        for msg in reversed(request.messages):
+            if msg.role == "user":
+                user_message = msg.content
+                break
+        
+        # Get relevant context
+        context = ""
+        if user_message:
+            context = context_manager.get_context_for_query(
+                user_message,
+                role=request.role
+            )
+        
+        # Build system prompt
+        system_prompt = context_manager.get_system_prompt(role=request.role)
+        
+        # Add context to system prompt if available
+        if context:
+            system_prompt += f"\n\n# Company Context\n{context}"
+        
+        # Convert messages to API format
+        api_messages = [
+            {"role": msg.role, "content": msg.content}
+            for msg in request.messages
+        ]
+        
+        # Get response from AI
+        response = await ai_client.chat(
+            messages=api_messages,
+            system_prompt=system_prompt
+        )
+        
+        return ChatResponse(
+            response=response,
+            context_used=bool(context)
+        )
+        
+    except Exception as e:
+        logger.error(f"Error in chat endpoint: {e}")
+        raise HTTPException(status_code=500, detail=str(e))
+
+@router.post("/stream")
+async def chat_stream(
+    request: ChatRequest,
+    context_manager: ContextManager = Depends(main.get_context_manager),
+    ai_client: AIClient = Depends(main.get_ai_client)
+):
+    """Stream chat response from Grimlock"""
+    
+    try:
+        # Get context
+        user_message = None
+        for msg in reversed(request.messages):
+            if msg.role == "user":
+                user_message = msg.content
+                break
+        
+        context = ""
+        if user_message:
+            context = context_manager.get_context_for_query(
+                user_message,
+                role=request.role
+            )
+        
+        # Build system prompt
+        system_prompt = context_manager.get_system_prompt(role=request.role)
+        if context:
+            system_prompt += f"\n\n# Company Context\n{context}"
+        
+        # Convert messages
+        api_messages = [
+            {"role": msg.role, "content": msg.content}
+            for msg in request.messages
+        ]
+        
+        # Stream response
+        async def generate():
+            async for chunk in ai_client.chat_stream(
+                messages=api_messages,
+                system_prompt=system_prompt
+            ):
+                yield chunk
+        
+        return StreamingResponse(generate(), media_type="text/plain")
+        
+    except Exception as e:
+        logger.error(f"Error in chat stream: {e}")
+        raise HTTPException(status_code=500, detail=str(e))