Yves Gugger
8f6e13ffcf
Some checks failed
CI / Frontend Lint & Type Check (push) Has been cancelled
CI / Frontend Build (push) Has been cancelled
CI / Backend Lint (push) Has been cancelled
CI / Backend Tests (push) Has been cancelled
CI / Docker Build (push) Has been cancelled
CI / Security Scan (push) Has been cancelled
Deploy / Build & Push Images (push) Has been cancelled
Deploy / Deploy to Server (push) Has been cancelled
Deploy / Notify (push) Has been cancelled
73 lines
2.3 KiB
Python
73 lines
2.3 KiB
Python
"""
|
|
LLM Agent endpoint (Tool Calling, Trader/Tycoon).
|
|
|
|
This endpoint runs a small tool loop:
|
|
- LLM requests tools via strict JSON
|
|
- Backend executes read-only tools against live DB/API helpers
|
|
- Final answer is streamed to the client (SSE) using the gateway
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
from typing import Any, Literal, Optional
|
|
|
|
from fastapi import APIRouter, Depends, HTTPException, Request, status
|
|
from fastapi.responses import JSONResponse, StreamingResponse
|
|
from pydantic import BaseModel, Field
|
|
from sqlalchemy.ext.asyncio import AsyncSession
|
|
|
|
from app.api.deps import get_current_user
|
|
from app.database import get_db
|
|
from app.models.user import User
|
|
from app.services.llm_agent import run_agent, stream_final_answer
|
|
|
|
|
|
router = APIRouter(prefix="/llm", tags=["LLM"])
|
|
|
|
|
|
class ChatMessage(BaseModel):
|
|
role: Literal["system", "user", "assistant"]
|
|
content: str
|
|
|
|
|
|
class AgentRequest(BaseModel):
|
|
messages: list[ChatMessage] = Field(default_factory=list, min_length=1)
|
|
path: str = Field(default="/terminal/hunt")
|
|
model: Optional[str] = None
|
|
temperature: float = Field(default=0.7, ge=0.0, le=2.0)
|
|
stream: bool = True
|
|
|
|
|
|
@router.post("/agent")
|
|
async def llm_agent(
|
|
payload: AgentRequest,
|
|
current_user: User = Depends(get_current_user),
|
|
db: AsyncSession = Depends(get_db),
|
|
):
|
|
try:
|
|
convo = await run_agent(
|
|
db,
|
|
current_user,
|
|
messages=[m.model_dump() for m in payload.messages],
|
|
path=payload.path,
|
|
model=payload.model,
|
|
temperature=payload.temperature,
|
|
)
|
|
except PermissionError as e:
|
|
raise HTTPException(status_code=status.HTTP_403_FORBIDDEN, detail=str(e))
|
|
except Exception as e:
|
|
raise HTTPException(status_code=500, detail=f"Agent failed: {type(e).__name__}: {e}")
|
|
|
|
if payload.stream:
|
|
return StreamingResponse(
|
|
stream_final_answer(convo, model=payload.model, temperature=payload.temperature),
|
|
media_type="text/event-stream",
|
|
headers={"Cache-Control": "no-cache", "Connection": "keep-alive"},
|
|
)
|
|
|
|
# Non-stream fallback: produce final answer in one shot by asking the model again.
|
|
# (kept simple; streaming path is preferred)
|
|
return JSONResponse({"ok": True, "messages": convo[-8:]})
|
|
|
|
|