app.py

from fastapi import FastAPI, Request
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel
from llama_cpp import Llama
import logging
import os
import ssl
import requests

# 🔧 إعداد السجل
logging.basicConfig(
    level=logging.DEBUG,
    format="🪵 [%(asctime)s] [%(levelname)s] %(message)s",
)
logger = logging.getLogger(__name__)

MODEL_REPO = "QuantFactory/Qwen2.5-7B-Instruct-GGUF"
MODEL_FILE = "Qwen2.5-7B-Instruct.Q4_K_M.gguf"
MODEL_PATH = f"/home/user/app/data/cache/{MODEL_FILE}"

# تحقق من SSL لـ Hugging Face Hub
def check_ssl_connection():
    try:
        response = requests.get("https://huggingface.co", verify=True, timeout=5)
        logger.info("✅ الاتصال بـ Hugging Face باستخدام SSL ناجح")
    except requests.exceptions.SSLError as e:
        logger.error(f"❌ خطأ SSL عند الاتصال بـ Hugging Face: {str(e)}")
    except Exception as e:
        logger.error(f"❌ خطأ في الاتصال بـ Hugging Face: {str(e)}")

check_ssl_connection()

# تحميل النموذج
if not os.path.exists(MODEL_PATH):
    from huggingface_hub import hf_hub_download
    os.makedirs("/home/user/app/data/cache", exist_ok=True)
    logger.info("📦 تحميل النموذج من Hugging Face Hub...")
    try:
        hf_hub_download(
            repo_id=MODEL_REPO,
            filename=MODEL_FILE,
            local_dir="/home/user/app/data/cache",
        )
        logger.info(f"✅ النموذج تم تحميله: {MODEL_PATH}")
    except Exception as e:
        logger.error(f"❌ فشل تحميل النموذج: {str(e)}")
        raise
else:
    logger.info(f"✅ النموذج موجود: {MODEL_PATH}")

# تحميل النموذج
try:
    llm = Llama(
        model_path=MODEL_PATH,
        n_ctx=4096,
        n_threads=6,
        n_gpu_layers=0,
        verbose=False
    )
    logger.info("🛠 النموذج تم تحميله بنجاح")
except Exception as e:
    logger.error(f"❌ فشل تحميل النموذج: {str(e)}")
    raise

# 🔍 اختبار النموذج
try:
    logger.info("🔍 يجري اختبار النموذج...")
    test_output = llm("مرحبا", max_tokens=10)
    logger.info(f"✅ اختبار النموذج ناجح: {test_output}")
except Exception as e:
    logger.error(f"❌ فشل اختبار النموذج: {str(e)}")
    raise RuntimeError("فشل تحميل النموذج") from e

SYSTEM_PROMPT = """<|im_start|>system
You are Qwen, created by Alibaba Cloud. You are an AI development assistant. Follow these rules:
1. If request is simple (single file, <50 lines), handle it directly
2. For complex requests (multiple files, >50 lines), just respond with \"CODER\"
3. Always check code for errors before sending
4. Never execute unsafe code<|im_end|>"""

# API setup
app = FastAPI()

# تفعيل CORS
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],  # يمكن تخصيصه لاحقًا (مثل ["http://localhost:3000"])
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

# تحقق من CORS أثناء الطلب
async def check_cors(request: Request):
    origin = request.headers.get("origin")
    if not origin:
        logger.warning("⚠ لا يوجد رأس Origin، قد تكون هناك مشكلة CORS")
    else:
        logger.info(f"🌐 تمتلك الطلب Origin: {origin}")

@app.on_event("startup")
async def startup_event():
    logger.info("🚀 الخادم بدأ بنجاح – /chat جاهز")
    logger.info("🌍 العنوان: https://<اسمك>.hf.space/chat (عند النشر على Hugging Face)")

class ChatRequest(BaseModel):
    message: str
    history: list[tuple[str, str]] = []  # استخدام tuple كما في الكود الأصلي

class ChatResponse(BaseModel):
    response: str
    updated_history: list[tuple[str, str]]

def format_prompt(messages):
    formatted = []
    for role, content in messages:
        if role == "system":
            formatted.append(f"<|im_start|>system\n{content}<|im_end|>")
        elif role == "user":
            formatted.append(f"<|im_start|>user\n{content}<|im_end|>")
        else:
            formatted.append(f"<|im_start|>assistant\n{content}<|im_end|>")
    formatted.append("<|im_start|>assistant\n")
    return "\n".join(formatted)

@app.post("/test")
async def test_endpoint():
    logger.info("📩 طلب تم استلامه على /test")
    return {"status": "success", "message": "النقطة النهائية تعمل"}

@app.post("/chat", response_model=ChatResponse)
async def chat(req: ChatRequest, request: Request):
    await check_cors(request)
    logger.info(f"🟡 تم استقبال الطلب من المتصفح: {req.message}")
    
    messages = [("system", SYSTEM_PROMPT)]
    for user_msg, bot_msg in req.history:
        messages.append(("user", user_msg))
        messages.append(("assistant", bot_msg))
    messages.append(("user", req.message))

    prompt = format_prompt(messages)
    logger.debug(f"📝 prompt المُرسل للنموذج:\n{prompt[:500]}...")

    try:
        output = llm(
            prompt,
            max_tokens=1024,
            temperature=0.7,
            top_p=0.9,
            stop=["<|im_end|>", "<|im_start|>"],
            echo=False
        )
        reply = output["choices"][0]["text"].strip()
        logger.info(f"🤖 رد النموذج: {reply}")
    except Exception as e:
        logger.error(f"حدث خطأ: {str(e)}")
        raise

    updated_history = req.history + [(req.message, reply)]
    return ChatResponse(response=reply, updated_history=updated_history)

@app.get("/")
def root():
    return {"message": "الخادم يعمل", "status": "ok"}