app/main.py

import os
import torch
from fastapi import FastAPI, Request
from fastapi.responses import JSONResponse
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
from starlette.middleware.cors import CORSMiddleware

# === Setup FastAPI ===
app = FastAPI()

# === CORS for frontend testing (optional) ===
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

# === Load Secret API Key from Hugging Face Secrets ===
API_KEY = os.getenv("API_KEY", "undefined")

# === Load Model and Adapter (CPU only) ===
BASE_MODEL = "Qwen/Qwen2-0.5B-Instruct"
ADAPTER_PATH = "adapter"

print("🔧 Loading tokenizer...")
tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL, trust_remote_code=True)

print("🧠 Loading base model on CPU...")
base_model = AutoModelForCausalLM.from_pretrained(
    BASE_MODEL,
    trust_remote_code=True,
    torch_dtype=torch.float32  # CPU only
).cpu()

print("🔗 Applying LoRA adapter...")
model = PeftModel.from_pretrained(base_model, ADAPTER_PATH).cpu()
model.eval()

print("✅ Model and adapter loaded.")


# === Root route for test ===
@app.get("/")
def read_root():
    return {"message": "🧠 Qwen2.5-0.5B-Instruct API is running on CPU!"}


# === POST /v1/chat/completions (OpenAI-style) ===
@app.post("/v1/chat/completions")
async def chat(request: Request):
    # ✅ Check API key from headers
    auth = request.headers.get("Authorization", "")
    if not auth.startswith("Bearer "):
        return JSONResponse(status_code=401, content={"error": "Missing Bearer token in Authorization header."})

    token = auth.replace("Bearer ", "").strip()
    if token != API_KEY:
        return JSONResponse(status_code=401, content={"error": "Invalid API key."})

    # ✅ Parse user prompt
    body = await request.json()
    messages = body.get("messages", [])
    if not messages or not isinstance(messages, list):
        return JSONResponse(status_code=400, content={"error": "No messages provided."})

    user_prompt = messages[-1]["content"]

    # ✅ Format prompt for Qwen chat model
    prompt = f"<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n<|im_start|>user\n{user_prompt}<|im_end|>\n<|im_start|>assistant\n"
    inputs = tokenizer(prompt, return_tensors="pt").to("cpu")

    # ✅ Generate
    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_new_tokens=512,
            temperature=0.7,
            top_p=0.9,
            do_sample=True,
            pad_token_id=tokenizer.eos_token_id
        )

    full_output = tokenizer.decode(outputs[0], skip_special_tokens=True)
    answer = full_output.split("<|im_start|>assistant\n")[-1].strip()

    # ✅ Return in OpenAI-style format
    return {
        "id": "chatcmpl-custom-001",
        "object": "chat.completion",
        "model": "Qwen2.5-0.5B-Instruct-LoRA",
        "choices": [
            {
                "index": 0,
                "message": {
                    "role": "assistant",
                    "content": answer
                },
                "finish_reason": "stop"
            }
        ]
    }