app.py

from fastapi import FastAPI, Request, Response
import httpx

app = FastAPI()
BACKEND = "http://127.0.0.1:8000"

@app.get("/")
async def root():
    return {
        "ok": True,
        "message": "GGUF OpenAI-compatible API is running",
        "chat_completions": "/v1/chat/completions"
    }

@app.get("/health")
async def health():
    async with httpx.AsyncClient(timeout=10) as client:
        r = await client.get(f"{BACKEND}/health")
        return Response(
            content=r.content,
            status_code=r.status_code,
            media_type=r.headers.get("content-type", "application/json"),
        )

@app.api_route("/v1/{path:path}", methods=["GET", "POST", "PUT", "PATCH", "DELETE"])
async def proxy(path: str, request: Request):
    url = f"{BACKEND}/v1/{path}"
    body = await request.body()
    headers = dict(request.headers)
    headers.pop("host", None)

    async with httpx.AsyncClient(timeout=None) as client:
        r = await client.request(
            method=request.method,
            url=url,
            headers=headers,
            content=body,
            params=request.query_params,
        )

    return Response(
        content=r.content,
        status_code=r.status_code,
        media_type=r.headers.get("content-type", "application/json"),
    )