app.py

from fastapi import FastAPI, HTTPException
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel
from typing import List, Dict, Any
import requests
import os

app = FastAPI(title="Hyper-Fast Serverless Engine v3.5")

app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

# Serverless Inference API Endpoint
API_URL = "https://api-inference.huggingface.co/models/meta-llama/Meta-Llama-3-8B-Instruct"

# Fetching token from Space Secrets
HF_TOKEN = os.getenv("HF_TOKEN", "") 
headers = {"Authorization": f"Bearer {HF_TOKEN}"} if HF_TOKEN else {}

class ChatPayload(BaseModel):
    user_id: str
    user_message: str
    current_chat_history: List[Dict[str, Any]] = []
    user_files: Dict[str, Any] = {}

@app.get("/")
def status():
    return {"status": "online", "engine": "Serverless GPU Bridge Active"}

@app.post("/chat")
async def chat_endpoint(payload: ChatPayload):
    try:
        user_query = payload.user_message
        
        # 1. Base System Prompt Setup
        prompt = "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nYou are a fast, responsive AI assistant.<|eot_id|>"
        
        # 2. Append history with correct formatting (Fixed {} here)
        for turn in payload.current_chat_history[-4:]:
            role = "user" if turn.get("role") == "user" else "assistant"
            content = turn.get("content", "")
            prompt += f"<|start_header_id|>{role}<|end_header_id|>\n\mrow{content}<|eot_id|>"
            
        # 3. Append current user query
        prompt += f"<|start_header_id|>user<|end_header_id|>\n\n{user_query}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n"

        api_payload = {
            "inputs": prompt,
            "parameters": {
                "max_new_tokens": 256,
                "temperature": 0.7,
                "return_full_text": False
            }
        }

        # 4. Request to HuggingFace API Cluster
        response = requests.post(API_URL, headers=headers, json=api_payload)
        
        if response.status_code != 200:
            raise HTTPException(status_code=response.status_code, detail=f"HF API Error: {response.text}")

        result = response.json()
        
        # 5. Extract generated text safely
        if isinstance(result, list) and len(result) > 0:
            generated_text = result[0].get("generated_text", "").strip()
        elif isinstance(result, dict) and "generated_text" in result:
            generated_text = result.get("generated_text", "").strip()
        else:
            generated_text = str(result)

        # 6. Build response state
        updated_history = payload.current_chat_history + [
            {"role": "user", "content": user_query},
            {"role": "assistant", "content": generated_text}
        ]

        return {
            "updated_chat_history": updated_history,
            "updated_files": payload.user_files
        }

    except Exception as exec_err:
        # Pushes logs inside HuggingFace console for exact traceback if anything else breaks
        print(f"Exception Triggered: {str(exec_err)}")
        raise HTTPException(status_code=500, detail=f"Cloud Matrix Failure: {str(exec_err)}")