app.py

# app.py (يدعم stream / non-stream من الواجهة)
import os
from typing import Any, Dict, List, Optional, Tuple, Union
import gradio as gr
from huggingface_hub import InferenceClient
import logging

# إعدادات
MODEL_ID = "HuggingFaceH4/zephyr-7b-beta"
MAX_CHAT_OUTPUT = 10000
HF_TOKEN_ENV_VARS = (
    "HF_TOKEN", "HUGGINGFACEHUB_API_TOKEN", "HUGGING_FACE_HUB_TOKEN", "HUGGINGFACE_TOKEN"
)
os.environ["GRADIO_ANALYTICS_ENABLED"] = "False"
os.environ["GRADIO_USE_SYMLINKS"] = "False"

# إعداد التسجيل
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

# قراءة التوكن
def _get_hf_token_from_env() -> Optional[str]:
    for key in HF_TOKEN_ENV_VARS:
        val = os.getenv(key)
        if val and val.strip():
            return val.strip()
    return None

def _select_hf_token(user_supplied: Optional[str]) -> Optional[str]:
    return user_supplied.strip() if user_supplied else _get_hf_token_from_env()

_CLIENT_CACHE: Dict[str, InferenceClient] = {}

def _get_client(token: Optional[str]) -> InferenceClient:
    key = token or "anon"
    if key not in _CLIENT_CACHE:
        _CLIENT_CACHE[key] = InferenceClient(model=MODEL_ID, token=token)
    return _CLIENT_CACHE[key]

# تطبيع سجل الدردشة
ChatHistoryType = Union[
    List[Dict[str, Any]],
    List[Tuple[Optional[str], Optional[str]]],
    List[List[Optional[str]]],
]

def _normalize_history(history: ChatHistoryType) -> List[Dict[str, str]]:
    out: List[Dict[str, str]] = []
    if not history:
        return out
    if isinstance(history[0], dict):
        for msg in history:
            role = msg.get("role", "user")
            content = msg.get("content", "")
            if isinstance(content, dict):
                path = content.get("path") or "file"
                content = f"[{path}]"
            out.append({"role": role, "content": str(content)})
    else:
        for pair in history:
            if isinstance(pair, (list, tuple)):
                if pair[0]: 
                    out.append({"role": "user", "content": str(pair[0])})
                if len(pair) > 1 and pair[1]:
                    out.append({"role": "assistant", "content": str(pair[1])})
    return out

def _truncate(txt: str, n: int = MAX_CHAT_OUTPUT) -> str:
    return txt if len(txt) <= n else txt[:n//2] + "\n...\n[output truncated]\n...\n" + txt[-n//2:]

# دالة الرد المعدلة
def respond(
    message: Union[str, Dict[str, Any]],
    history: ChatHistoryType,
    system_message: str,
    max_tokens: int,
    temperature: float,
    top_p: float,
    hf_token: str,
    enable_stream: bool,
):
    try:
        if isinstance(message, dict):
            message_text = message.get("text", "")
        else:
            message_text = message

        msgs = [{"role": "system", "content": system_message}]
        msgs.extend(_normalize_history(history))
        msgs.append({"role": "user", "content": message_text})

        token = _select_hf_token(hf_token)
        client = _get_client(token)

        accumulated = ""
        stream = client.chat_completion(
            messages=msgs,
            max_tokens=max_tokens,
            temperature=temperature,
            top_p=top_p,
            stream=True,
        )
        
        for chunk in stream:
            delta = chunk.choices[0].delta.content
            if delta:
                accumulated += delta
                if len(accumulated) > MAX_CHAT_OUTPUT:
                    accumulated = _truncate(accumulated)
                    break
                if enable_stream:
                    yield delta  # إرسال الجزء الجديد فقط
                    
        if not enable_stream:
            yield _truncate(accumulated)
            
    except Exception as e:
        logger.error(f"Error in respond: {str(e)}")
        yield f"[حدث خطأ: {str(e)}]"

# واجهة Gradio المعدلة
with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown("## 💬 واجهة دردشة متطورة")
    
    with gr.Row():
        with gr.Column(scale=3):
            chatbot = gr.Chatbot(height=500)
            msg = gr.Textbox(label="رسالتك")
        with gr.Column(scale=1):
            system_msg = gr.Textbox(value="You are a helpful assistant.", label="رسالة النظام")
            max_tokens = gr.Slider(1, 2048, value=512, step=1, label="الحد الأقصى للرموز")
            temperature = gr.Slider(0.1, 4.0, value=0.7, step=0.1, label="الدرجة الحرارية")
            top_p = gr.Slider(0.1, 1.0, value=0.95, step=0.05, label="Top-p")
            hf_token = gr.Textbox(label="رمز Hugging Face (اختياري)", type="password")
            enable_stream = gr.Checkbox(value=False, label="الرد المباشر")
            submit_btn = gr.Button("إرسال")
            clear_btn = gr.Button("مسح")

    def user(user_message, history):
        return "", history + [[user_message, None]]
    
    def bot(history, system_message, max_tokens, temperature, top_p, hf_token, enable_stream):
        response = respond(
            history[-1][0],
            history[:-1],
            system_message,
            max_tokens,
            temperature,
            top_p,
            hf_token,
            enable_stream,
        )
        history[-1][1] = ""
        for chunk in response:
            history[-1][1] += chunk
            yield history

    msg.submit(user, [msg, chatbot], [msg, chatbot], queue=False).then(
        bot, [chatbot, system_msg, max_tokens, temperature, top_p, hf_token, enable_stream], chatbot
    )
    submit_btn.click(user, [msg, chatbot], [msg, chatbot], queue=False).then(
        bot, [chatbot, system_msg, max_tokens, temperature, top_p, hf_token, enable_stream], chatbot
    )
    clear_btn.click(lambda: None, None, chatbot, queue=False)

# إعدادات التشغيل المعدلة
if __name__ == "__main__":
    demo.launch(
        server_name="0.0.0.0",
        server_port=7860,
        debug=False,  # تعطيل وضع التصحيح لتجنب المشاكل
        share=False,
        show_error=True,
        favicon_path=None,
        ssl_verify=True,
        ssl_keyfile=None,
        ssl_certfile=None,
        ssl_keyfile_password=None,
        quiet=True,
        show_api=False,
        max_threads=40,
        auth=None,
        auth_message=None,
        prevent_thread_lock=False,
        allowed_paths=None,
        blocked_paths=None,
        root_path="",
        http_headers={"Transfer-Encoding": "chunked"},  # حل مشكلة Content-Length
    )