app.py

import os
import json
import time
import asyncio
import logging
import hashlib
from datetime import datetime
from typing import List, Dict, Optional, Union
from concurrent.futures import ThreadPoolExecutor

from fastapi import FastAPI, HTTPException, Request, UploadFile, File, WebSocket, WebSocketDisconnect, Depends
from fastapi.responses import StreamingResponse, HTMLResponse, FileResponse, JSONResponse
from fastapi.staticfiles import StaticFiles
from fastapi.templating import Jinja2Templates
from fastapi.security import APIKeyHeader
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel, Field
from langdetect import detect, DetectorFactory
import numpy as np
import pandas as pd
import cv2
import torch
from PIL import Image
import moviepy.editor as mp

# تحميل نماذج الذكاء الاصطناعي
from transformers import pipeline, AutoModelForCausalLM, AutoTokenizer
from diffusers import StableDiffusionPipeline, EulerDiscreteScheduler
from tensorflow.keras.models import load_model
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import cosine_similarity
from transformers import BitsAndBytesConfig

# التهيئة الأساسية
DetectorFactory.seed = 0
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger("MarkAI")

app = FastAPI(
    title="MarkAI - الذكاء الاصطناعي المتكامل",
    version="2.0",
    description="منصة متكاملة للذكاء الاصطناعي تدعم توليد النصوص، الأكواد، الصور والفيديوهات مع نظام ذاكرة متقدم",
    contact={
        "name": "Ibrahim Lasfar",
        "email": "ibrahim@markai.com"
    },
    license_info={
        "name": "MIT License",
    }
)

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

# تهيئة مجلدات التخزين
os.makedirs("uploads", exist_ok=True)
os.makedirs("memory/conversations", exist_ok=True)
os.makedirs("memory/projects", exist_ok=True)
os.makedirs("memory/code", exist_ok=True)
os.makedirs("memory/backups", exist_ok=True)

# 1. نماذج اللغات المدعومة (محدثة مع النماذج الكبيرة)
LANGUAGE_MODELS = {
    # النماذج الصغيرة (افتراضية)
    "en": "gpt2-medium",
    "ar": "arbml/gpt2-arabic-poetry",
    "zh": "bert-base-chinese",
    "ja": "colorfulscoop/gpt2-small-ja",
    "fr": "dbmdz/gpt2-french",
    "de": "dbmdz/gpt2-german",
    "it": "LorenzoDeMattei/GePpeTto",
    "hi": "surajpai/GPT2-Hindi",
    "code": "codeparrot/codeparrot-small",
    
    # النماذج الكبيرة
    "en-large": "EleutherAI/gpt-j-6B",
    "ar-large": "bigscience/bloom-7b1",
    "code-large": "tiiuae/falcon-7b"
}

# 2. نظام الأمان والمفاتيح
API_KEY_HEADER = APIKeyHeader(name="X-API-KEY")

def load_api_keys():
    try:
        with open("memory/api_keys.json", "r") as f:
            return json.load(f)
    except:
        return {"demo_key": "demo123"}  # مفتاح تجريبي افتراضي

def save_api_keys(keys):
    with open("memory/api_keys.json", "w") as f:
        json.dump(keys, f)

def authenticate(api_key: str):
    keys = load_api_keys()
    return api_key in keys.values()

# 3. نظام الذاكرة المتقدم
class AIMemory:
    def __init__(self):
        self.conversations = {}
        self.projects = {}
        self.code_repository = {}
        self.load_all_data()
        
    def load_all_data(self):
        """تحميل جميع البيانات من الملفات"""
        try:
            # تحميل المحادثات
            for conv_file in os.listdir("memory/conversations"):
                if conv_file.endswith(".json"):
                    conv_id = conv_file.split(".")[0]
                    with open(f"memory/conversations/{conv_file}", "r", encoding="utf-8") as f:
                        self.conversations[conv_id] = json.load(f)
            
            # تحميل المشاريع
            if os.path.exists("memory/projects/projects.json"):
                with open("memory/projects/projects.json", "r", encoding="utf-8") as f:
                    self.projects = json.load(f)
            
            # تحميل مستودع الأكواد
            if os.path.exists("memory/code/code_repository.json"):
                with open("memory/code/code_repository.json", "r", encoding="utf-8") as f:
                    self.code_repository = json.load(f)
        except Exception as e:
            logger.error(f"Error loading data: {str(e)}")
    
    def create_conversation(self, initial_prompt: str) -> str:
        """إنشاء محادثة جديدة مع تسمية تلقائية"""
        conv_id = hashlib.md5(f"{initial_prompt}{datetime.now()}".encode()).hexdigest()[:10]
        conv_name = initial_prompt[:30] + "..." if len(initial_prompt) > 30 else initial_prompt
        
        conversation = {
            "id": conv_id,
            "name": conv_name,
            "created_at": str(datetime.now()),
            "updated_at": str(datetime.now()),
            "messages": [],
            "context": [],
            "status": "active"
        }
        
        self.conversations[conv_id] = conversation
        self.save_conversation(conv_id)
        return conv_id
    
    def save_conversation(self, conv_id: str):
        """حفظ محادثة معينة"""
        if conv_id in self.conversations:
            try:
                with open(f"memory/conversations/{conv_id}.json", "w", encoding="utf-8") as f:
                    json.dump(self.conversations[conv_id], f, ensure_ascii=False, indent=2)
            except Exception as e:
                logger.error(f"Error saving conversation {conv_id}: {str(e)}")
    
    def add_message(self, conv_id: str, role: str, content: str, metadata: dict = {}):
        """إضافة رسالة إلى المحادثة"""
        if conv_id not in self.conversations:
            raise ValueError("المحادثة غير موجودة")
            
        message = {
            "role": role,
            "content": content,
            "timestamp": str(datetime.now()),
            "metadata": metadata
        }
        
        self.conversations[conv_id]["messages"].append(message)
        self.conversations[conv_id]["updated_at"] = str(datetime.now())
        self.save_conversation(conv_id)
    
    def get_conversation_context(self, conv_id: str, max_messages: int = 10) -> List[dict]:
        """الحصول على سياق المحادثة"""
        if conv_id not in self.conversations:
            return []
            
        return self.conversations[conv_id]["messages"][-max_messages:]
    
    def create_project(self, name: str, description: str, project_type: str) -> str:
        """إنشاء مشروع جديد"""
        project_id = hashlib.md5(f"{name}{datetime.now()}".encode()).hexdigest()[:8]
        
        project = {
            "id": project_id,
            "name": name,
            "description": description,
            "type": project_type,
            "created_at": str(datetime.now()),
            "updated_at": str(datetime.now()),
            "status": "active",
            "files": [],
            "conversations": []
        }
        
        self.projects[project_id] = project
        self.save_projects()
        return project_id
    
    def save_projects(self):
        """حفظ جميع المشاريع"""
        try:
            with open("memory/projects/projects.json", "w", encoding="utf-8") as f:
                json.dump(self.projects, f, ensure_ascii=False, indent=2)
        except Exception as e:
            logger.error(f"Error saving projects: {str(e)}")
    
    def save_code_snippet(self, code: str, language: str, purpose: str, metadata: dict = {}):
        """حفظ جزء من الكود في المستودع"""
        code_id = hashlib.md5(f"{code}{datetime.now()}".encode()).hexdigest()[:8]
        
        snippet = {
            "id": code_id,
            "code": code,
            "language": language,
            "purpose": purpose,
            "metadata": metadata,
            "created_at": str(datetime.now()),
            "usage_count": 0
        }
        
        self.code_repository[code_id] = snippet
        self.save_code_repository()
        return code_id
    
    def save_code_repository(self):
        """حفظ مستودع الأكواد"""
        try:
            with open("memory/code/code_repository.json", "w", encoding="utf-8") as f:
                json.dump(self.code_repository, f, ensure_ascii=False, indent=2)
        except Exception as e:
            logger.error(f"Error saving code repository: {str(e)}")
    
    def backup_data(self):
        """إنشاء نسخة احتياطية لجميع البيانات"""
        timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
        backup_dir = f"memory/backups/{timestamp}"
        os.makedirs(backup_dir, exist_ok=True)
        
        try:
            # نسخ المحادثات
            os.makedirs(f"{backup_dir}/conversations", exist_ok=True)
            for conv_id, conv_data in self.conversations.items():
                with open(f"{backup_dir}/conversations/{conv_id}.json", "w", encoding="utf-8") as f:
                    json.dump(conv_data, f, ensure_ascii=False, indent=2)
            
            # نسخ المشاريع
            with open(f"{backup_dir}/projects.json", "w", encoding="utf-8") as f:
                json.dump(self.projects, f, ensure_ascii=False, indent=2)
            
            # نسخ الأكواد
            with open(f"{backup_dir}/code_repository.json", "w", encoding="utf-8") as f:
                json.dump(self.code_repository, f, ensure_ascii=False, indent=2)
            
            return backup_dir
        except Exception as e:
            logger.error(f"Error during backup: {str(e)}")
            return None

memory = AIMemory()

# 4. نظام التقييم والتحليل
class AnalyticsEngine:
    def __init__(self):
        self.sentiment_model = pipeline("sentiment-analysis")
        self.tfidf = TfidfVectorizer()
        
    def analyze_sentiment(self, text: str) -> dict:
        """تحليل المشاعر للنص"""
        try:
            result = self.sentiment_model(text)[0]
            return {
                "sentiment": result["label"],
                "score": result["score"],
                "positive": result["label"] == "POSITIVE",
                "negative": result["label"] == "NEGATIVE"
            }
        except Exception as e:
            logger.warning(f"Sentiment analysis failed, using fallback: {str(e)}")
            # Fallback basic sentiment analysis
            positive_words = ["good", "great", "excellent", "happy", "جيد", "رائع", "ممتاز", "سعيد"]
            negative_words = ["bad", "terrible", "awful", "sad", "سيء", "فظيع", "مزعج", "حزين"]
            
            positive_count = sum(text.lower().count(word) for word in positive_words)
            negative_count = sum(text.lower().count(word) for word in negative_words)
            
            if positive_count > negative_count:
                return {"sentiment": "POSITIVE", "score": positive_count/(positive_count+negative_count+1)}
            elif negative_count > positive_count:
                return {"sentiment": "NEGATIVE", "score": negative_count/(positive_count+negative_count+1)}
            else:
                return {"sentiment": "NEUTRAL", "score": 0.5}
    
    def evaluate_response(self, prompt: str, response: str) -> dict:
        """تقييم جودة الرد"""
        # تحليل طول الرد
        length_score = min(len(response.split()) / 100, 1.0)
        
        # تحليل التنوع
        unique_words = len(set(response.split()))
        diversity_score = min(unique_words / 50, 1.0)
        
        # تحليل الصلة بالموضوع
        try:
            vectors = self.tfidf.fit_transform([prompt, response])
            relevance_score = cosine_similarity(vectors[0:1], vectors[1:2])[0][0]
        except Exception as e:
            logger.warning(f"TF-IDF analysis failed: {str(e)}")
            relevance_score = 0.7  # قيمة افتراضية في حالة الخطأ
            
        # تحليل المشاعر
        sentiment = self.analyze_sentiment(response)
        
        return {
            "length_score": length_score,
            "diversity_score": diversity_score,
            "relevance_score": relevance_score,
            "sentiment": sentiment,
            "overall_score": (length_score + diversity_score + relevance_score + sentiment["score"]) / 4
        }

analytics = AnalyticsEngine()

# 5. المحرك الأساسي للذكاء الاصطناعي
class AIEngine:
    def __init__(self):
        self.executor = ThreadPoolExecutor(max_workers=8)
        self.models = {}
        self.device = "cuda" if torch.cuda.is_available() else "cpu"
        self.quantization_config = BitsAndBytesConfig(
            load_in_4bit=True,
            bnb_4bit_compute_dtype=torch.float16,
            bnb_4bit_quant_type="nf4"
        )
        
    async def load_model(self, model_type: str, model_name: str = None, use_large: bool = False):
        """تحميل نموذج معين"""
        model_key = f"{model_type}-large" if use_large else model_type
        
        if model_key not in self.models:
            try:
                if model_type == "text":
                    model_name = model_name or (LANGUAGE_MODELS.get(f"{model_type}-large") if use_large else LANGUAGE_MODELS.get("en"))
                    
                    if use_large:
                        model = AutoModelForCausalLM.from_pretrained(
                            model_name,
                            quantization_config=self.quantization_config,
                            device_map="auto",
                            torch_dtype=torch.float16
                        )
                    else:
                        model = AutoModelForCausalLM.from_pretrained(model_name).to(self.device)
                    
                    tokenizer = AutoTokenizer.from_pretrained(model_name)
                    self.models[model_key] = {"tokenizer": tokenizer, "model": model}
                
                elif model_type == "image":
                    scheduler = EulerDiscreteScheduler.from_pretrained("stabilityai/stable-diffusion-2", subfolder="scheduler")
                    model = StableDiffusionPipeline.from_pretrained(
                        "stabilityai/stable-diffusion-2",
                        scheduler=scheduler,
                        torch_dtype=torch.float16
                    ).to(self.device)
                    self.models[model_key] = model
                
                elif model_type == "code":
                    if use_large:
                        model = AutoModelForCausalLM.from_pretrained(
                            LANGUAGE_MODELS["code-large"],
                            quantization_config=self.quantization_config,
                            device_map="auto",
                            torch_dtype=torch.float16
                        )
                    else:
                        model = AutoModelForCausalLM.from_pretrained(LANGUAGE_MODELS["code"]).to(self.device)
                    
                    tokenizer = AutoTokenizer.from_pretrained(LANGUAGE_MODELS["code-large"] if use_large else LANGUAGE_MODELS["code"])
                    self.models[model_key] = {"tokenizer": tokenizer, "model": model}
                
                logger.info(f"تم تحميل النموذج بنجاح: {model_key}")
            except Exception as e:
                logger.error(f"خطأ في تحميل النموذج: {str(e)}")
                raise
        
        return self.models[model_key]
    
    async def generate_text(self, prompt: str, lang: str = None, max_length: int = 300, use_large: bool = False) -> str:
        """توليد نص بناء على المطالبة"""
        if not lang:
            try:
                lang = detect(prompt)
            except:
                lang = "en"
        
        model_name = LANGUAGE_MODELS.get(f"{lang}-large" if use_large else lang, 
                                      LANGUAGE_MODELS.get("en-large" if use_large else "en"))
        
        model = await self.load_model("text", model_name, use_large)
        
        inputs = model["tokenizer"](prompt, return_tensors="pt").to(self.device)
        outputs = model["model"].generate(**inputs, max_length=max_length, do_sample=True, top_k=50, top_p=0.95)
        
        return model["tokenizer"].decode(outputs[0], skip_special_tokens=True)
    
    async def generate_code(self, prompt: str, language: str = "python", max_length: int = 500, use_large: bool = False) -> str:
        """توليد كود برمجي"""
        model = await self.load_model("code", use_large=use_large)
        
        prompt = f"# Language: {language}\n# Description: {prompt}\n# Code:\n"
        inputs = model["tokenizer"](prompt, return_tensors="pt").to(self.device)
        outputs = model["model"].generate(**inputs, max_length=max_length, do_sample=True, top_k=50, top_p=0.95)
        
        generated_code = model["tokenizer"].decode(outputs[0], skip_special_tokens=True)
        
        # حفظ الكود في المستودع
        code_id = memory.save_code_snippet(
            code=generated_code,
            language=language,
            purpose=prompt[:100],
            metadata={
                "generated_at": str(datetime.now()),
                "model_used": "large" if use_large else "base"
            }
        )
        
        return generated_code
    
    async def generate_image(self, prompt: str, save_path: str = None) -> str:
        """توليد صورة من النص"""
        model = await self.load_model("image")
        
        if not save_path:
            save_path = f"uploads/generated_image_{int(time.time())}.png"
            
        image = model(prompt).images[0]
        image.save(save_path)
        
        return save_path
    
    async def generate_video(self, prompt: str, duration: int = 5, fps: int = 24) -> str:
        """توليد فيديو من النص (محاكاة)"""
        save_path = f"uploads/generated_video_{int(time.time())}.mp4"
        
        # إنشاء فيديو مع نص (استخدام صورة سوداء كخلفية)
        clip = mp.ColorClip(size=(640, 480), color=(0, 0, 0), duration=duration)
        txt_clip = mp.TextClip(prompt, fontsize=24, color='white', size=clip.size).set_position('center').set_duration(duration)
        video = mp.CompositeVideoClip([clip, txt_clip])
        video.write_videofile(save_path, fps=fps)
        
        return save_path
    
    async def analyze_code(self, code: str, language: str = "python") -> dict:
        """تحليل الكود وإعطاء تقييم"""
        # تحليل أساسي للكود
        analysis = {
            "length": len(code.split("\n")),
            "complexity": "low",
            "quality": "medium",
            "issues": [],
            "suggestions": []
        }
        
        # تحليل أولي
        if len(code.split("\n")) > 50:
            analysis["complexity"] = "high"
            analysis["suggestions"].append("Consider breaking this into smaller functions/modules")
        
        if "TODO" in code or "FIXME" in code:
            analysis["issues"].append("Contains unfinished tasks (TODO/FIXME)")
            analysis["quality"] = "low"
        
        if language == "python" and "print(" in code:
            analysis["suggestions"].append("Consider using logging instead of print statements for production code")
        
        return analysis
    
    async def improve_code(self, code: str, language: str, improvements: List[str]) -> str:
        """تحسين الكود بناء على طلبات محددة"""
        improved_code = code
        
        # تطبيق التحسينات الأساسية
        if "add_comments" in improvements:
            improved_code = f"# Improved by MarkAI at {datetime.now()}\n# Original code with enhancements\n\n{improved_code}"
        
        if "optimize" in improvements:
            improved_code = improved_code.replace("for i in range(len(", "for item in ")
            improved_code = improved_code.replace(".append(", " += [")
        
        if "add_error_handling" in improvements and language == "python":
            improved_code = f"try:\n    {improved_code.replace('\n', '\n    ')}\nexcept Exception as e:\n    print(f\"An error occurred: {e}\")"
        
        return improved_code

engine = AIEngine()

# 6. نظام التفكير والتخطيط
class ThinkingEngine:
    def __init__(self):
        self.planning_steps = {
            "text": {
                "ar": [
                    "🔍 تحليل الطلب والمتطلبات...",
                    "🧠 معالجة البيانات والبحث...",
                    "📚 استرجاع المعلومات ذات الصلة...",
                    "✨ توليد الإجابة المثلى..."
                ],
                "en": [
                    "🔍 Analyzing request and requirements...",
                    "🧠 Processing data and researching...",
                    "📚 Retrieving relevant information...",
                    "✨ Generating optimal response..."
                ]
            },
            "code": {
                "ar": [
                    "🔍 تحليل متطلبات الكود...",
                    "🧠 تصميم الخوارزمية...",
                    "📚 البحث عن الحلول المثلى...",
                    "✨ كتابة وتوليد الكود..."
                ],
                "en": [
                    "🔍 Analyzing code requirements...",
                    "🧠 Designing algorithm...",
                    "📚 Researching optimal solutions...",
                    "✨ Writing and generating code..."
                ]
            },
            "image": {
                "ar": [
                    "🔍 تحليل وصف الصورة...",
                    "🧠 تكوين المفاهيم الفنية...",
                    "🎨 رسم العناصر الأساسية...",
                    "✨ إضافة اللمسات النهائية..."
                ],
                "en": [
                    "🔍 Analyzing image description...",
                    "🧠 Composing artistic concepts...",
                    "🎨 Sketching basic elements...",
                    "✨ Adding final touches..."
                ]
            },
            "video": {
                "ar": [
                    "🔍 تحليل السيناريو...",
                    "🎬 إعداد القصة والمشاهد...",
                    "🎞️ تركيب العناصر المرئية...",
                    "✨ إضافة المؤثرات والصوت..."
                ],
                "en": [
                    "🔍 Analyzing scenario...",
                    "🎬 Preparing storyboard and scenes...",
                    "🎞️ Composing visual elements...",
                    "✨ Adding effects and sound..."
                ]
            },
            "project": {
                "ar": [
                    "🔍 تحليل متطلبات المشروع...",
                    "📝 تحديد الهيكل الأساسي...",
                    "🛠️ إعداد الملفات والموارد...",
                    "✨ إنشاء المشروع الجديد..."
                ],
                "en": [
                    "🔍 Analyzing project requirements...",
                    "📝 Defining basic structure...",
                    "🛠️ Preparing files and resources...",
                    "✨ Creating new project..."
                ]
            }
        }
    
    def get_thinking_steps(self, task_type: str, lang: str = "en") -> List[str]:
        """الحصول على خطوات التفكير حسب نوع المهمة واللغة"""
        return self.planning_steps.get(task_type, self.planning_steps["text"]).get(lang, self.planning_steps["text"]["en"])
    
    async def generate_plan(self, prompt: str, task_type: str = "text") -> dict:
        """إنشاء خطة تنفيذية للمهمة"""
        try:
            lang = detect(prompt)
        except:
            lang = "en"
            
        steps = self.get_thinking_steps(task_type, lang)
        
        plan = {
            "task": prompt,
            "type": task_type,
            "language": lang,
            "steps": steps,
            "estimated_time": "30 seconds",  # يمكن جعل هذا أكثر دقة
            "required_resources": ["CPU", "GPU"] if task_type in ["image", "video"] else ["CPU"],
            "created_at": str(datetime.now())
        }
        
        return plan

thinker = ThinkingEngine()

# 7. نماذج طلبات API
class GenerationRequest(BaseModel):
    prompt: str
    content_type: str = "text"  # text, code, image, video, project
    language: Optional[str] = None
    conversation_id: Optional[str] = None
    improvements: Optional[List[str]] = None
    use_large_model: bool = False  # إضافة خيار استخدام النماذج الكبيرة

class ConversationRequest(BaseModel):
    initial_prompt: str
    project_id: Optional[str] = None
    use_large_model: bool = False

class ProjectRequest(BaseModel):
    name: str
    description: str
    project_type: str  # web, mobile, desktop, ai, other

class CodeImprovementRequest(BaseModel):
    code: str
    language: str
    improvements: List[str] = Field(..., example=["add_comments", "optimize", "add_error_handling"])
    use_large_model: bool = False

# 8. نظام إدارة المحادثات عبر WebSocket
class ConnectionManager:
    def __init__(self):
        self.active_connections: Dict[str, WebSocket] = {}
    
    async def connect(self, conversation_id: str, websocket: WebSocket):
        await websocket.accept()
        self.active_connections[conversation_id] = websocket
    
    def disconnect(self, conversation_id: str):
        if conversation_id in self.active_connections:
            del self.active_connections[conversation_id]
    
    async def send_message(self, conversation_id: str, message: str):
        if conversation_id in self.active_connections:
            await self.active_connections[conversation_id].send_text(message)

manager = ConnectionManager()

# 9. نقاط النهاية الأساسية
@app.post("/api/conversation/start")
async def start_conversation(request: ConversationRequest):
    """بدء محادثة جديدة"""
    conv_id = memory.create_conversation(request.initial_prompt)
    
    if request.project_id and request.project_id in memory.projects:
        memory.projects[request.project_id]["conversations"].append(conv_id)
        memory.save_projects()
    
    # إضافة الرسالة الأولى
    memory.add_message(
        conv_id=conv_id,
        role="user",
        content=request.initial_prompt,
        metadata={
            "type": "text", 
            "project_id": request.project_id,
            "use_large_model": request.use_large_model
        }
    )
    
    return {"conversation_id": conv_id, "name": memory.conversations[conv_id]["name"]}

@app.websocket("/api/conversation/ws/{conversation_id}")
async def websocket_conversation(websocket: WebSocket, conversation_id: str):
    """محادثة في الوقت الحقيقي عبر WebSocket"""
    await manager.connect(conversation_id, websocket)
    
    try:
        while True:
            data = await websocket.receive_text()
            message = json.loads(data)
            
            if message["type"] == "user_message":
                # حفظ رسالة المستخدم
                memory.add_message(
                    conv_id=conversation_id,
                    role="user",
                    content=message["content"],
                    metadata={
                        "type": message.get("content_type", "text"),
                        "use_large_model": message.get("use_large_model", False)
                    }
                )
                
                # إنشاء خطة للرد
                content_type = message.get("content_type", "text")
                plan = await thinker.generate_plan(message["content"], content_type)
                
                # إرسال خطوات التفكير
                for step in plan["steps"]:
                    await manager.send_message(conversation_id, json.dumps({
                        "type": "thinking",
                        "content": step
                    }))
                    await asyncio.sleep(1)
                
                # توليد الرد
                use_large = message.get("use_large_model", False)
                
                if content_type == "text":
                    response = await engine.generate_text(
                        message["content"],
                        use_large=use_large
                    )
                elif content_type == "code":
                    response = await engine.generate_code(
                        message["content"],
                        message.get("language", "python"),
                        use_large=use_large
                    )
                elif content_type == "image":
                    image_path = await engine.generate_image(message["content"])
                    response = f"IMAGE_GENERATED:{image_path}"
                elif content_type == "video":
                    video_path = await engine.generate_video(message["content"])
                    response = f"VIDEO_GENERATED:{video_path}"
                else:
                    response = "نوع المحتوى غير مدعوم"
                
                # تحليل الرد
                evaluation = analytics.evaluate_response(message["content"], response)
                
                # حفظ الرد
                memory.add_message(
                    conv_id=conversation_id,
                    role="assistant",
                    content=response,
                    metadata={
                        "type": content_type,
                        "evaluation": evaluation,
                        "plan": plan,
                        "model_used": "large" if use_large else "base"
                    }
                )
                
                # إرسال الرد النهائي
                await manager.send_message(conversation_id, json.dumps({
                    "type": "assistant_response",
                    "content": response,
                    "evaluation": evaluation
                }))
                
    except WebSocketDisconnect:
        manager.disconnect(conversation_id)
    except Exception as e:
        logger.error(f"WebSocket error: {str(e)}")
        await manager.send_message(conversation_id, json.dumps({
            "type": "error",
            "content": f"حدث خطأ: {str(e)}"
        }))

@app.post("/api/project/create")
async def create_project(request: ProjectRequest, api_key: str = Depends(API_KEY_HEADER)):
    """إنشاء مشروع جديد"""
    if not authenticate(api_key):
        raise HTTPException(status_code=403, detail="غير مصرح به")
    
    project_id = memory.create_project(request.name, request.description, request.project_type)
    
    # إنشاء مجلد المشروع
    project_dir = f"projects/{project_id}"
    os.makedirs(project_dir, exist_ok=True)
    
    # إنشاء ملفات أساسية
    with open(f"{project_dir}/README.md", "w", encoding="utf-8") as f:
        f.write(f"# {request.name}\n\n{request.description}\n\nCreated by MarkAI at {datetime.now()}")
    
    return {"project_id": project_id, "path": project_dir}

@app.post("/api/code/improve")
async def improve_code(request: CodeImprovementRequest):
    """تحسين الكود المقدم"""
    analysis = await engine.analyze_code(request.code, request.language)
    improved_code = await engine.improve_code(request.code, request.language, request.improvements)
    
    # حفظ الكود المحسن
    code_id = memory.save_code_snippet(
        code=improved_code,
        language=request.language,
        purpose="Improved code",
        metadata={
            "original_code": request.code,
            "improvements": request.improvements,
            "analyzed_at": str(datetime.now()),
            "model_used": "large" if request.use_large_model else "base"
        }
    )
    
    return {
        "improved_code": improved_code,
        "analysis": analysis,
        "code_id": code_id
    }

@app.get("/api/conversation/list")
async def list_conversations(project_id: Optional[str] = None):
    """الحصول على قائمة المحادثات"""
    if project_id and project_id in memory.projects:
        convs = [memory.conversations[cid] for cid in memory.projects[project_id]["conversations"] if cid in memory.conversations]
    else:
        convs = list(memory.conversations.values())
    
    return {"conversations": convs}

@app.get("/api/project/list")
async def list_projects():
    """الحصول على قائمة المشاريع"""
    return {"projects": list(memory.projects.values())}

@app.get("/api/code/list")
async def list_code_snippets(language: Optional[str] = None):
    """الحصول على قائمة الأكواد المحفوظة"""
    snippets = list(memory.code_repository.values())
    
    if language:
        snippets = [s for s in snippets if s["language"].lower() == language.lower()]
    
    return {"snippets": snippets}

# 10. نظام النسخ الاحتياطي التلقائي
async def backup_scheduler():
    while True:
        await asyncio.sleep(3600)  # كل ساعة
        try:
            backup_dir = memory.backup_data()
            if backup_dir:
                logger.info(f"تم إنشاء نسخة احتياطية في: {backup_dir}")
        except Exception as e:
            logger.error(f"فشل النسخ الاحتياطي: {str(e)}")

# 11. واجهة المستخدم
app.mount("/static", StaticFiles(directory="static"), name="static")
app.mount("/uploads", StaticFiles(directory="uploads"), name="uploads")
templates = Jinja2Templates(directory="templates")

@app.get("/", response_class=HTMLResponse)
async def read_root(request: Request):
    return templates.TemplateResponse("index.html", {"request": request})

@app.get("/chat/{conversation_id}", response_class=HTMLResponse)
async def chat_interface(request: Request, conversation_id: str):
    if conversation_id not in memory.conversations:
        raise HTTPException(status_code=404, detail="المحادثة غير موجودة")
    
    return templates.TemplateResponse("chat.html", {
        "request": request,
        "conversation": memory.conversations[conversation_id]
    })

# 12. بدء المهام الجانبية
@app.on_event("startup")
async def startup_event():
    asyncio.create_task(backup_scheduler())
    
    # تحميل النماذج الأساسية مسبقاً
    await engine.load_model("text")
    await engine.load_model("code")
    
    logger.info("تم بدء تشغيل MarkAI بنجاح")

# 13. ملفات إضافية لتهيئة Hugging Face Spaces
@app.get("/app")
async def serve_app():
    return FileResponse("static/index.html")

@app.get("/favicon.ico")
async def favicon():
    return FileResponse("static/favicon.ico")

# 14. تشغيل التطبيق
if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=7860, reload=True)