Update design and generation logic

api/endpoints.py

@@ -15,8 +15,10 @@ MODEL_NAME = os.getenv("MODEL_NAME", "openai/gpt-oss-20b:fireworks-ai")
 def model_info():
     return {
         "model_name": MODEL_NAME,
-        "secondary_model": os.getenv("SECONDARY_MODEL_NAME", "MGZON/Veltrix"),
+        "secondary_model": os.getenv("SECONDARY_MODEL_NAME", "deepseek-ai/DeepSeek-R1-Distill-Qwen-7B"),
         "tertiary_model": os.getenv("TERTIARY_MODEL_NAME", "mistralai/Mixtral-8x7B-Instruct-v0.1"),
+        "clip_base_model": os.getenv("CLIP_BASE_MODEL", "openai/clip-vit-base-patch32"),
+        "clip_large_model": os.getenv("CLIP_LARGE_MODEL", "openai/clip-vit-large-patch14"),
         "api_base": API_ENDPOINT,
         "status": "online"
     }
@@ -46,6 +48,42 @@ async def chat_endpoint(req: QueryRequest):
     response = "".join(list(stream))
     return {"response": response}
 
+
+# في api/endpoints.py
+@router.post("/api/audio-transcription")
+async def audio_transcription_endpoint(file: UploadFile = File(...)):
+    model_name, api_endpoint = select_model("transcribe audio", input_type="audio")
+    audio_data = await file.read()
+    response = "".join(list(request_generation(
+        api_key=HF_TOKEN,
+        api_base=api_endpoint,
+        message="Transcribe audio",
+        system_prompt="Transcribe the provided audio using Whisper.",
+        model_name=model_name,
+        temperature=0.7,
+        max_new_tokens=128000,
+        input_type="audio",
+        audio_data=audio_data,
+    )))
+    return {"transcription": response}
+
+@router.post("/api/text-to-speech")
+async def text_to_speech_endpoint(req: dict):
+    text = req.get("text", "")
+    model_name, api_endpoint = select_model("text to speech", input_type="text")
+    response = request_generation(
+        api_key=HF_TOKEN,
+        api_base=api_endpoint,
+        message=text,
+        system_prompt="Convert the provided text to speech using Parler-TTS.",
+        model_name=model_name,
+        temperature=0.7,
+        max_new_tokens=128000,
+        input_type="text",
+    )
+    audio_data = b"".join(list(response))
+    return StreamingResponse(io.BytesIO(audio_data), media_type="audio/wav")
+
 @router.post("/api/code")
 async def code_endpoint(req: dict):
     framework = req.get("framework")
@@ -57,7 +95,7 @@ async def code_endpoint(req: dict):
         api_key=HF_TOKEN,
         api_base=api_endpoint,
         message=prompt,
-        system_prompt="You are a coding expert.",
+        system_prompt="You are a coding expert. Provide detailed, well-commented code with examples and explanations.",
         model_name=model_name,
         temperature=0.7,
         max_new_tokens=128000,
@@ -72,13 +110,30 @@ async def analysis_endpoint(req: dict):
         api_key=HF_TOKEN,
         api_base=api_endpoint,
         message=message,
-        system_prompt="You are an expert analyst. Provide detailed analysis with step-by-step reasoning.",
+        system_prompt="You are an expert analyst. Provide detailed analysis with step-by-step reasoning and examples.",
         model_name=model_name,
         temperature=0.7,
         max_new_tokens=128000,
     )))
     return {"analysis": response}
 
+@router.post("/api/image-analysis")
+async def image_analysis_endpoint(req: dict):
+    image_url = req.get("image_url", "")
+    task = req.get("task", "describe")
+    prompt = f"Perform the following task on the image at {image_url}: {task}"
+    model_name, api_endpoint = select_model(prompt)
+    response = "".join(list(request_generation(
+        api_key=HF_TOKEN,
+        api_base=api_endpoint,
+        message=prompt,
+        system_prompt="You are an expert in image analysis. Provide detailed descriptions or classifications based on the query.",
+        model_name=model_name,
+        temperature=0.7,
+        max_new_tokens=128000,
+    )))
+    return {"image_analysis": response}
+
 @router.get("/api/test-model")
 async def test_model(model: str = MODEL_NAME, endpoint: str = API_ENDPOINT):
     try:

api/models.py

@@ -3,8 +3,8 @@ from typing import List, Optional
 
 class QueryRequest(BaseModel):
     message: str
-    system_prompt: str = "You are a helpful assistant capable of code generation, analysis, review, and more."
+    system_prompt: str = "You are an expert assistant providing detailed, comprehensive, and well-structured responses. For code, include comments, examples, and complete implementations. For image-related queries, provide detailed analysis or descriptions. For general queries, provide in-depth explanations with examples and additional context where applicable. Continue generating content until the query is fully addressed, leveraging the full capacity of the model."
     history: Optional[List[dict]] = None
-    temperature: float = 0.9
+    temperature: float = 0.7
     max_new_tokens: int = 128000
     enable_browsing: bool = False

main.py

@@ -5,9 +5,11 @@ from fastapi.responses import HTMLResponse, RedirectResponse
 from fastapi.staticfiles import StaticFiles
 from fastapi.templating import Jinja2Templates
 from starlette.middleware.base import BaseHTTPMiddleware
+from fastapi.openapi.docs import get_swagger_ui_html
 import gradio as gr
 from api.endpoints import router as api_router
 from utils.generation import generate, LATEX_DELIMS
+import io
 
 # إعداد التسجيل
 logging.basicConfig(level=logging.INFO)
@@ -18,6 +20,7 @@ logger.info("Files in /app/: %s", os.listdir("/app"))
 
 # إعداد العميل لـ Hugging Face Inference API
 HF_TOKEN = os.getenv("HF_TOKEN")
+BACKUP_HF_TOKEN = os.getenv("BACKUP_HF_TOKEN")  # إضافة التوكن الاحتياطي
 if not HF_TOKEN:
     logger.error("HF_TOKEN is not set in environment variables.")
     raise ValueError("HF_TOKEN is required for Inference API.")
@@ -28,43 +31,129 @@ CONCURRENCY_LIMIT = int(os.getenv("CONCURRENCY_LIMIT", 20))
 
 # إعداد CSS
 css = """
-.gradio-container { max-width: 800px; margin: auto; }
-.chatbot { border: 1px solid #ccc; border-radius: 10px; }
-.input-textbox { font-size: 16px; }
+.gradio-container { max-width: 1200px; margin: auto; }
+.chatbot { border: 1px solid #ccc; border-radius: 10px; padding: 15px; background-color: #f9f9f9; }
+.input-textbox { font-size: 16px; padding: 10px; }
+.upload-button::before {
+    content: '📷';
+    margin-right: 8px;
+    font-size: 22px;
+}
+.audio-input::before {
+    content: '🎤';
+    margin-right: 8px;
+    font-size: 22px;
+}
+.audio-output::before {
+    content: '🔊';
+    margin-right: 8px;
+    font-size: 22px;
+}
+.loading::after {
+    content: '';
+    display: inline-block;
+    width: 16px;
+    height: 16px;
+    border: 2px solid #333;
+    border-top-color: transparent;
+    border-radius: 50%;
+    animation: spin 1s linear infinite;
+    margin-left: 8px;
+}
+@keyframes spin {
+    to { transform: rotate(360deg); }
+}
+.output-container {
+    margin-top: 20px;
+    padding: 10px;
+    border: 1px solid #ddd;
+    border-radius: 8px;
+}
+.audio-output-container {
+    display: flex;
+    align-items: center;
+    gap: 10px;
+    margin-top: 10px;
+}
 """
 
+# دالة لمعالجة الإدخال (نص، صوت، صور، ملفات)
+def process_input(message, audio_input=None, file_input=None, history=None, system_prompt=None, temperature=0.7, reasoning_effort="medium", enable_browsing=True, max_new_tokens=128000):
+    input_type = "text"
+    audio_data = None
+    image_data = None
+    if audio_input:
+        input_type = "audio"
+        with open(audio_input, "rb") as f:
+            audio_data = f.read()
+        message = "Transcribe this audio"
+    elif file_input:
+        input_type = "file"
+        if file_input.endswith(('.png', '.jpg', '.jpeg')):
+            input_type = "image"
+            with open(file_input, "rb") as f:
+                image_data = f.read()
+            message = f"Analyze image: {file_input}"
+        else:
+            message = f"Analyze file: {file_input}"
+    
+    response_text = ""
+    audio_response = None
+    for chunk in generate(
+        message=message,
+        history=history,
+        system_prompt=system_prompt,
+        temperature=temperature,
+        reasoning_effort=reasoning_effort,
+        enable_browsing=enable_browsing,
+        max_new_tokens=max_new_tokens,
+        input_type=input_type,
+        audio_data=audio_data,
+        image_data=image_data
+    ):
+        if isinstance(chunk, bytes):
+            audio_response = io.BytesIO(chunk)
+            audio_response.name = "response.wav"
+        else:
+            response_text += chunk
+        yield response_text, audio_response
+
 # إعداد واجهة Gradio
 chatbot_ui = gr.ChatInterface(
-    fn=generate,
+    fn=process_input,
     type="messages",
     chatbot=gr.Chatbot(
         label="MGZon Chatbot",
         type="messages",
-        height=600,
+        height=800,
         latex_delimiters=LATEX_DELIMS,
     ),
     additional_inputs_accordion=gr.Accordion("⚙️ Settings", open=True),
     additional_inputs=[
-        gr.Textbox(label="System prompt", value="You are a helpful assistant capable of code generation, analysis, review, and more.", lines=2),
-        gr.Slider(label="Temperature", minimum=0.0, maximum=1.0, step=0.1, value=0.9),
+        gr.Textbox(
+            label="System Prompt",
+            value="You are an expert assistant providing detailed, comprehensive, and well-structured responses. Support text, audio, image, and file inputs. For audio, transcribe using Whisper. For text-to-speech, use Parler-TTS. For images and files, analyze content appropriately. Continue generating content until the query is fully addressed, leveraging the full capacity of the model.",
+            lines=4
+        ),
+        gr.Slider(label="Temperature", minimum=0.0, maximum=1.0, step=0.1, value=0.7),
         gr.Radio(label="Reasoning Effort", choices=["low", "medium", "high"], value="medium"),
         gr.Checkbox(label="Enable DeepSearch (web browsing)", value=True),
-        gr.Slider(label="Max New Tokens", minimum=50, maximum=128000, step=50, value=4096),
+        gr.Slider(label="Max New Tokens", minimum=50, maximum=128000, step=50, value=128000),
+        gr.Audio(label="Voice Input", type="filepath", elem_classes="audio-input"),
+        gr.File(label="Upload Image/File", file_types=["image", ".pdf", ".txt"], elem_classes="upload-button"),
     ],
+    additional_outputs=[gr.Audio(label="Voice Output", type="filepath", elem_classes="audio-output", autoplay=True)],
     stop_btn="Stop",
     examples=[
-        ["Explain the difference between supervised and unsupervised learning."],
-        ["Generate a React component for a login form."],
-        ["Review this Python code: print('Hello World')"],
-        ["Analyze the performance of a Django REST API."],
-        ["Tell me about MGZon products and services."],
-        ["Create a Flask route for user authentication."],
-        ["What are the latest trends in AI?"],
-        ["Provide guidelines for publishing a technical blog post."],
-        ["Who is the founder of MGZon?"],
+        ["Explain the difference between supervised and unsupervised learning in detail with examples."],
+        ["Generate a complete React component for a login form with form validation and error handling."],
+        ["Describe this image: https://example.com/image.jpg"],
+        ["Transcribe this audio: [upload audio file]."],
+        ["Convert this text to speech: Hello, welcome to MGZon!"],
+        ["Analyze this file: [upload PDF or text file]."],
     ],
     title="MGZon Chatbot",
-    description="A versatile chatbot powered by MGZON/Veltrix for MGZon queries. Supports code generation, analysis, review, web search, and MGZon-specific queries. Licensed under Apache 2.0.",
+    description="A versatile chatbot powered by DeepSeek, CLIP, Whisper, and Parler-TTS for text, image, audio, and file queries. Supports long responses, voice input/output, file uploads with custom icons, and backup token switching. Licensed under Apache 2.0.",
     theme="gradio/soft",
     css=css,
 )
@@ -79,20 +168,18 @@ app = gr.mount_gradio_app(app, chatbot_ui, path="/gradio")
 app.mount("/static", StaticFiles(directory="static"), name="static")
 templates = Jinja2Templates(directory="templates")
 
-# تضمين API endpoints
-app.include_router(api_router)
-
 # Middleware لمعالجة 404
 class NotFoundMiddleware(BaseHTTPMiddleware):
     async def dispatch(self, request: Request, call_next):
         try:
             response = await call_next(request)
             if response.status_code == 404:
+                logger.warning(f"404 Not Found: {request.url}")
                 return templates.TemplateResponse("404.html", {"request": request}, status_code=404)
             return response
         except Exception as e:
-            logger.exception(f"Error processing request: {e}")
-            return templates.TemplateResponse("404.html", {"request": request}, status_code=404)
+            logger.exception(f"Error processing request {request.url}: {e}")
+            return templates.TemplateResponse("500.html", {"request": request, "error": str(e)}, status_code=500)
 
 app.add_middleware(NotFoundMiddleware)
 
@@ -106,10 +193,15 @@ async def root(request: Request):
 async def docs(request: Request):
     return templates.TemplateResponse("docs.html", {"request": request})
 
+# Swagger UI endpoint
+@app.get("/swagger", response_class=HTMLResponse)
+async def swagger_ui():
+    return get_swagger_ui_html(openapi_url="/openapi.json", title="MGZon API Documentation")
+
 # Redirect لـ /gradio
 @app.get("/launch-chatbot", response_class=RedirectResponse)
 async def launch_chatbot():
-    return RedirectResponse(url="/gradio")
+    return RedirectResponse(url="/gradio", status_code=302)
 
 # تشغيل الخادم
 if __name__ == "__main__":

requirements.txt

@@ -1,13 +1,24 @@
-fastapi==0.115.0
+fastapi==0.115.2
 uvicorn==0.30.6
-gradio==4.44.1
+gradio==4.48.0
 openai==1.42.0
 httpx==0.27.0
 python-dotenv==1.0.1
-pydocstyle==6.3.0
+pydocstyle==6.4.0
 requests==2.32.3
 beautifulsoup4==4.12.3
 tenacity==8.5.0
 selenium==4.25.0
+webdriver-manager==4.0.2
 jinja2==3.1.4
 cachetools==5.5.0
+pydub==0.25.1
+ffmpeg-python==0.2.0
+numpy==1.26.4
+parler-tts==0.2.0
+torch==2.4.1
+torchaudio==2.4.1
+transformers==4.45.1
+webrtcvad==2.0.10
+Pillow==10.4.0
+

templates/docs.html

@@ -1,176 +1,72 @@
 <!DOCTYPE html>
 <html lang="en">
 <head>
-  <meta charset="UTF-8" />
-  <meta name="viewport" content="width=device-width, initial-scale=1.0" />
-  <title>MGZon API Documentation</title>
-  <!-- Tailwind CSS v3 -->
-  <script src="https://cdn.tailwindcss.com"></script>
-  <!-- Boxicons -->
-  <link href="https://cdn.jsdelivr.net/npm/[email protected]/css/boxicons.min.css" rel="stylesheet" />
-  <!-- Prism for code highlighting -->
-  <link href="https://cdn.jsdelivr.net/npm/[email protected]/themes/prism.min.css" rel="stylesheet" />
-  <script src="https://cdn.jsdelivr.net/npm/[email protected].0/prism.min.js"></script>
-  <style>
-    /* Animated gradient background */
-    @keyframes gradientShift {
-      0% { background-position: 0% 50%; }
-      50% { background-position: 100% 50%; }
-      100% { background-position: 0% 50%; }
-    }
-    body {
-      background: linear-gradient(135deg, #0f172a, #0e7490, #065f46, #064e3b);
-      background-size: 400% 400%;
-      animation: gradientShift 15s ease infinite;
-      font-family: system-ui, sans-serif;
-    }
-    /* Glass-morphism helpers */
-    .glass {
-      background: rgba(255, 255, 255, 0.07);
-      border-radius: 1rem;
-      border: 1px solid rgba(255, 255, 255, 0.12);
-      backdrop-filter: blur(12px);
-      -webkit-backdrop-filter: blur(12px);
-    }
-    /* Sidebar transition */
-    .sidebar {
-      transition: transform 0.3s ease-in-out;
-    }
-    .sidebar.collapsed .logo {
-      opacity: 0;
-      transition: opacity 0.2s ease;
-    }
-    /* Prevent sidebar from covering footer */
-    .main-content {
-      min-height: calc(100vh - 4rem);
-    }
-    /* Hover effects for cards and tabs */
-    .glass:hover {
-      transform: scale(1.05);
-      box-shadow: 0 8px 16px rgba(0, 0, 0, 0.3);
-      background: rgba(255, 255, 255, 0.15);
-    }
-    /* Tab content transition */
-    .tab-content {
-      display: none;
-      transition: opacity 0.3s ease;
-    }
-    .tab-content.active {
-      display: block;
-      opacity: 1;
-    }
-    /* Responsive sidebar */
-    @media (max-width: 768px) {
-      .sidebar {
-        transform: translateX(-100%);
-      }
-      .sidebar.active {
-        transform: translateX(0);
-      }
-    }
-    /* Copy button feedback */
-    .copy-btn.copied::after {
-      content: 'Copied!';
-      position: absolute;
-      top: -2rem;
-      right: 1rem;
-      background: #10b981;
-      color: white;
-      padding: 0.5rem;
-      border-radius: 0.5rem;
-      font-size: 0.875rem;
-    }
-  </style>
+    <meta charset="UTF-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <title>MGZon API Documentation</title>
+    <link href="https://cdn.jsdelivr.net/npm/[email protected]/dist/tailwind.min.css" rel="stylesheet">
+    <link href="https://cdn.jsdelivr.net/npm/[email protected]/themes/prism.min.css" rel="stylesheet">
+    <link href="https://cdn.jsdelivr.net/npm/[email protected]/css/boxicons.min.css" rel="stylesheet">
+    <style>
+        @keyframes slideIn {
+            from { transform: translateX(-100%); }
+            to { transform: translateX(0); }
+        }
+        .sidebar.active { transform: translateX(0); }
+        @media (max-width: 768px) {
+            .sidebar { transform: translateX(-100%); }
+            .sidebar.active { transform: translateX(0); animation: slideIn 0.3s ease; }
+        }
+        .tab-content { display: none; }
+        .tab-content.active { display: block; }
+    </style>
 </head>
-<body class="text-white flex flex-col min-h-screen">
-  <!-- Mobile toggle button -->
-  <button
-    id="sidebarToggle"
-    class="md:hidden fixed top-4 left-4 z-50 p-2 text-2xl text-white rounded bg-gray-800/60 hover:bg-gray-700/80 transition"
-    aria-label="Toggle navigation"
-  >
-    <i class="bx bx-menu"></i>
-  </button>
-
-  <!-- Sidebar -->
-  <aside
-    id="sidebar"
-    class="sidebar fixed inset-y-0 left-0 w-64 bg-gradient-to-b from-teal-800 to-emerald-900 p-6 flex flex-col overflow-y-auto z-40"
-  >
-    <button
-      id="closeSidebar"
-      class="md:hidden text-white text-2xl absolute top-4 right-4"
-      aria-label="Close sidebar"
-    >
-      ✕
-    </button>
-    <img
-      src="https://raw.githubusercontent.com/Mark-Lasfar/MGZon/main/public/icons/mg.svg"
-      alt="MGZon Logo"
-      class="logo w-20 h-20 mx-auto mb-8 animate-pulse"
-    />
-    <nav class="flex flex-col gap-3">
-      <a href="/" class="px-4 py-2 rounded-lg hover:bg-emerald-600 transition">Home</a>
-      <a href="/docs" class="px-4 py-2 rounded-lg bg-emerald-600">API Documentation</a>
-      <a href="https://hager-zon.vercel.app/about" class="px-4 py-2 rounded-lg hover:bg-emerald-600 transition">About MGZon</a>
-      <a href="https://hager-zon.vercel.app/community" class="px-4 py-2 rounded-lg hover:bg-emerald-600 transition">Community</a>
-      <a href="https://mark-elasfar.web.app/" target="_blank" class="px-4 py-2 rounded-lg hover:bg-emerald-600 transition">Mark Al-Asfar</a>
-    </nav>
-  </aside>
-
-  <!-- Main content -->
-  <main class="flex-1 md:ml-64 p-6 main-content">
-    <section class="container max-w-6xl mx-auto">
-      <a
-        href="/"
-        class="inline-block bg-gray-700/80 text-white px-4 py-2 rounded-lg mb-4 hover:bg-emerald-600 transition"
-      >
-        Back to Home
-      </a>
-      <div class="text-center py-12">
-        <img
-          src="https://raw.githubusercontent.com/Mark-Lasfar/MGZon/main/public/icons/mg.svg"
-          alt="MGZon Logo"
-          class="w-32 h-32 mx-auto mb-6 animate-bounce"
-        />
-        <h1 class="text-5xl font-bold mb-4 bg-clip-text text-transparent bg-gradient-to-r from-teal-300 to-emerald-400">
-          MGZon API Documentation 📚
-        </h1>
-        <p class="text-lg max-w-2xl mx-auto mb-8">
-          Integrate MGZon Chatbot into your projects with our API. Supports Python (Django, Flask, FastAPI), JavaScript (React, Node.js, Express), Ruby (Rails), PHP (Laravel), and more.
-        </p>
-        <a
-          href="/swagger"
-          class="inline-block bg-emerald-500 text-white px-4 py-2 rounded-lg mt-4 hover:bg-emerald-600 transition"
-        >
-          View Interactive API Docs
-        </a>
-      </div>
-      <div class="docs my-12">
-        <h2 class="text-3xl font-bold text-center mb-8">Endpoints</h2>
-        <div class="flex flex-wrap gap-4 mb-8">
-          <button class="tab-btn glass px-4 py-2 rounded-lg text-white bg-emerald-500" data-tab="chat">Chat</button>
-          <button class="tab-btn glass px-4 py-2 rounded-lg text-white" data-tab="code">Code</button>
-          <button class="tab-btn glass px-4 py-2 rounded-lg text-white" data-tab="analysis">Analysis</button>
-          <button class="tab-btn glass px-4 py-2 rounded-lg text-white" data-tab="oauth">OAuth</button>
-        </div>
-        <div id="chat" class="tab-content glass p-6 active">
-          <div class="code-block glass p-6 mb-6 relative">
-            <h3 class="text-xl font-semibold text-emerald-300 mb-2">POST /api/chat</h3>
-            <p>Send a chat message to the MGZon Chatbot and get a response.</p>
-            <pre><code class="language-json">{
-  "message": "Your query here",
-  "system_prompt": "You are a helpful assistant",
-  "history": [{"role": "user", "content": "Previous message"}, {"role": "assistant", "content": "Previous response"}],
-  "temperature": 0.7,
-  "max_new_tokens": 4096,
-  "enable_browsing": false
+<body class="bg-gray-900 text-white min-h-screen flex flex-col">
+    <button class="md:hidden fixed top-4 left-4 z-50 text-white text-2xl sidebar-toggle">☰</button>
+    <div class="sidebar fixed top-0 left-0 h-full w-64 bg-gradient-to-b from-blue-800 to-orange-500 text-white p-6 flex flex-col md:translate-x-0 transition-transform duration-300">
+        <img src="https://raw.githubusercontent.com/Mark-Lasfar/MGZon/main/public/icons/mg.svg" alt="MGZon Logo" class="w-20 h-20 mb-6 animate-pulse">
+        <nav class="flex flex-col gap-4">
+            <a href="/" class="px-4 py-2 rounded-lg hover:bg-orange-600 transition">Home</a>
+            <a href="/docs" class="px-4 py-2 rounded-lg bg-orange-600">API Documentation</a>
+            <a href="https://hager-zon.vercel.app/about" class="px-4 py-2 rounded-lg hover:bg-orange-600 transition">About MGZon</a>
+            <a href="https://hager-zon.vercel.app/community" class="px-4 py-2 rounded-lg hover:bg-orange-600 transition">Community</a>
+            <a href="https://mark-elasfar.web.app/" target="_blank" class="px-4 py-2 rounded-lg hover:bg-orange-600 transition">Mark Al-Asfar</a>
+        </nav>
+    </div>
+    <div class="flex-1 md:ml-64 p-6">
+        <div class="container max-w-6xl mx-auto">
+            <div class="text-center py-12">
+                <img src="https://raw.githubusercontent.com/Mark-Lasfar/MGZon/main/public/icons/mg.svg" alt="MGZon Logo" class="w-32 h-32 mx-auto mb-6 animate-bounce">
+                <h1 class="text-5xl font-bold mb-4 animate-fade-in">MGZon API Documentation 📚</h1>
+                <p class="text-lg mb-8">
+                    Integrate MGZon Chatbot into your projects with our API. Supports Python (Django, Flask, FastAPI), JavaScript (React, Node.js, Express), Ruby (Rails), PHP (Laravel), and more.
+                </p>
+            </div>
+            <div class="docs my-12">
+                <h2 class="text-3xl font-bold text-center mb-8">Endpoints</h2>
+                <div class="flex flex-wrap gap-4 mb-8">
+                    <button class="tab-btn bg-orange-500 text-white px-4 py-2 rounded-lg" data-tab="chat">Chat</button>
+                    <button class="tab-btn bg-gray-700 text-white px-4 py-2 rounded-lg" data-tab="code">Code</button>
+                    <button class="tab-btn bg-gray-700 text-white px-4 py-2 rounded-lg" data-tab="analysis">Analysis</button>
+                    <button class="tab-btn bg-gray-700 text-white px-4 py-2 rounded-lg" data-tab="oauth">OAuth</button>
+                </div>
+                <div id="chat" class="tab-content active">
+                    <div class="code-block bg-gray-800 p-6 rounded-lg shadow-lg mb-6">
+                        <h3 class="text-xl font-semibold text-orange-500 mb-2">POST /api/chat</h3>
+                        <p>Send a chat message to the MGZon Chatbot and get a response.</p>
+                        <pre><code class="language-json">{
+    "message": "Your query here",
+    "system_prompt": "You are a helpful assistant",
+    "history": [{"role": "user", "content": "Previous message"}, {"role": "assistant", "content": "Previous response"}],
+    "temperature": 0.7,
+    "max_new_tokens": 4096,
+    "enable_browsing": false
 }</code></pre>
-            <button class="copy-btn bg-emerald-500 text-white px-4 py-2 rounded-lg mt-4 relative">Copy</button>
-          </div>
-          <div class="code-block glass p-6 mb-6 relative">
-            <h3 class="text-xl font-semibold text-emerald-300 mb-2">Python (requests)</h3>
-            <pre><code class="language-python">import requests
+                        <button class="copy-btn bg-orange-500 text-white px-4 py-2 rounded-lg mt-4">Copy</button>
+                    </div>
+                    <div class="code-block bg-gray-800 p-6 rounded-lg shadow-lg mb-6">
+                        <h3 class="text-xl font-semibold text-orange-500 mb-2">Python (requests)</h3>
+                        <pre><code class="language-python">import requests
 
 response = requests.post(
     "https://hager-zon.vercel.app/api/chat",
@@ -182,11 +78,11 @@ response = requests.post(
     }
 )
 print(response.json())</code></pre>
-            <button class="copy-btn bg-emerald-500 text-white px-4 py-2 rounded-lg mt-4 relative">Copy</button>
-          </div>
-          <div class="code-block glass p-6 mb-6 relative">
-            <h3 class="text-xl font-semibold text-emerald-300 mb-2">Python (Django)</h3>
-            <pre><code class="language-python">import requests
+                        <button class="copy-btn bg-orange-500 text-white px-4 py-2 rounded-lg mt-4">Copy</button>
+                    </div>
+                    <div class="code-block bg-gray-800 p-6 rounded-lg shadow-lg mb-6">
+                        <h3 class="text-xl font-semibold text-orange-500 mb-2">Python (Django)</h3>
+                        <pre><code class="language-python">import requests
 from django.http import JsonResponse
 from django.views import View
 
@@ -198,11 +94,11 @@ class ChatView(View):
             json={"message": data, "system_prompt": "You are a coding expert", "temperature": 0.7}
         )
         return JsonResponse(response.json())</code></pre>
-            <button class="copy-btn bg-emerald-500 text-white px-4 py-2 rounded-lg mt-4 relative">Copy</button>
-          </div>
-          <div class="code-block glass p-6 mb-6 relative">
-            <h3 class="text-xl font-semibold text-emerald-300 mb-2">Python (Flask)</h3>
-            <pre><code class="language-python">from flask import Flask, request, jsonify
+                        <button class="copy-btn bg-orange-500 text-white px-4 py-2 rounded-lg mt-4">Copy</button>
+                    </div>
+                    <div class="code-block bg-gray-800 p-6 rounded-lg shadow-lg mb-6">
+                        <h3 class="text-xl font-semibold text-orange-500 mb-2">Python (Flask)</h3>
+                        <pre><code class="language-python">from flask import Flask, request, jsonify
 import requests
 
 app = Flask(__name__)
@@ -215,11 +111,11 @@ def chat():
         json={"message": data, "system_prompt": "You are a coding expert", "temperature": 0.7}
     )
     return jsonify(response.json())</code></pre>
-            <button class="copy-btn bg-emerald-500 text-white px-4 py-2 rounded-lg mt-4 relative">Copy</button>
-          </div>
-          <div class="code-block glass p-6 mb-6 relative">
-            <h3 class="text-xl font-semibold text-emerald-300 mb-2">JavaScript (React)</h3>
-            <pre><code class="language-javascript">import React, { useState } from 'react';
+                        <button class="copy-btn bg-orange-500 text-white px-4 py-2 rounded-lg mt-4">Copy</button>
+                    </div>
+                    <div class="code-block bg-gray-800 p-6 rounded-lg shadow-lg mb-6">
+                        <h3 class="text-xl font-semibold text-orange-500 mb-2">JavaScript (React)</h3>
+                        <pre><code class="language-javascript">import React, { useState } from 'react';
 
 function ChatComponent() {
     const [message, setMessage] = useState('');
@@ -250,11 +146,11 @@ function ChatComponent() {
 }
 
 export default ChatComponent;</code></pre>
-            <button class="copy-btn bg-emerald-500 text-white px-4 py-2 rounded-lg mt-4 relative">Copy</button>
-          </div>
-          <div class="code-block glass p-6 mb-6 relative">
-            <h3 class="text-xl font-semibold text-emerald-300 mb-2">JavaScript (Node.js/Express)</h3>
-            <pre><code class="language-javascript">const express = require('express');
+                        <button class="copy-btn bg-orange-500 text-white px-4 py-2 rounded-lg mt-4">Copy</button>
+                    </div>
+                    <div class="code-block bg-gray-800 p-6 rounded-lg shadow-lg mb-6">
+                        <h3 class="text-xl font-semibold text-orange-500 mb-2">JavaScript (Node.js/Express)</h3>
+                        <pre><code class="language-javascript">const express = require('express');
 const fetch = require('node-fetch');
 const app = express();
 
@@ -277,11 +173,11 @@ app.post('/chat', async (req, res) => {
 });
 
 app.listen(3000, () => console.log('Server running on port 3000'));</code></pre>
-            <button class="copy-btn bg-emerald-500 text-white px-4 py-2 rounded-lg mt-4 relative">Copy</button>
-          </div>
-          <div class="code-block glass p-6 mb-6 relative">
-            <h3 class="text-xl font-semibold text-emerald-300 mb-2">Ruby (Rails)</h3>
-            <pre><code class="language-ruby">require 'httparty'
+                        <button class="copy-btn bg-orange-500 text-white px-4 py-2 rounded-lg mt-4">Copy</button>
+                    </div>
+                    <div class="code-block bg-gray-800 p-6 rounded-lg shadow-lg mb-6">
+                        <h3 class="text-xl font-semibold text-orange-500 mb-2">Ruby (Rails)</h3>
+                        <pre><code class="language-ruby">require 'httparty'
 
 class ChatController < ApplicationController
     def create
@@ -298,11 +194,11 @@ class ChatController < ApplicationController
         render json: response.parsed_response
     end
 end</code></pre>
-            <button class="copy-btn bg-emerald-500 text-white px-4 py-2 rounded-lg mt-4 relative">Copy</button>
-          </div>
-          <div class="code-block glass p-6 mb-6 relative">
-            <h3 class="text-xl font-semibold text-emerald-300 mb-2">PHP (Laravel)</h3>
-            <pre><code class="language-php">use Illuminate\Support\Facades\Http;
+                        <button class="copy-btn bg-orange-500 text-white px-4 py-2 rounded-lg mt-4">Copy</button>
+                    </div>
+                    <div class="code-block bg-gray-800 p-6 rounded-lg shadow-lg mb-6">
+                        <h3 class="text-xl font-semibold text-orange-500 mb-2">PHP (Laravel)</h3>
+                        <pre><code class="language-php">use Illuminate\Support\Facades\Http;
 
 class ChatController extends Controller
 {
@@ -317,206 +213,107 @@ class ChatController extends Controller
         return $response->json();
     }
 }</code></pre>
-            <button class="copy-btn bg-emerald-500 text-white px-4 py-2 rounded-lg mt-4 relative">Copy</button>
-          </div>
-          <div class="code-block glass p-6 mb-6 relative">
-            <h3 class="text-xl font-semibold text-emerald-300 mb-2">Bash (cURL)</h3>
-            <pre><code class="language-bash">curl -X POST https://hager-zon.vercel.app/api/chat \
+                        <button class="copy-btn bg-orange-500 text-white px-4 py-2 rounded-lg mt-4">Copy</button>
+                    </div>
+                    <div class="code-block bg-gray-800 p-6 rounded-lg shadow-lg mb-6">
+                        <h3 class="text-xl font-semibold text-orange-500 mb-2">Bash (cURL)</h3>
+                        <pre><code class="language-bash">curl -X POST https://hager-zon.vercel.app/api/chat \
 -H "Content-Type: application/json" \
 -d '{"message":"Generate a React component","system_prompt":"You are a coding expert","temperature":0.7,"max_new_tokens":4096}'</code></pre>
-            <button class="copy-btn bg-emerald-500 text-white px-4 py-2 rounded-lg mt-4 relative">Copy</button>
-          </div>
-        </div>
-        <div id="code" class="tab-content glass p-6">
-          <div class="code-block glass p-6 mb-6 relative">
-            <h3 class="text-xl font-semibold text-emerald-300 mb-2">POST /api/code</h3>
-            <p>Generate or modify code for various frameworks.</p>
-            <pre><code class="language-json">{
-  "framework": "React",
-  "task": "Create a login form component",
-  "code": "// Existing code (optional)"
+                        <button class="copy-btn bg-orange-500 text-white px-4 py-2 rounded-lg mt-4">Copy</button>
+                    </div>
+                </div>
+                <div id="code" class="tab-content">
+                    <div class="code-block bg-gray-800 p-6 rounded-lg shadow-lg mb-6">
+                        <h3 class="text-xl font-semibold text-orange-500 mb-2">POST /api/code</h3>
+                        <p>Generate or modify code for various frameworks.</p>
+                        <pre><code class="language-json">{
+    "framework": "React",
+    "task": "Create a login form component",
+    "code": "// Existing code (optional)"
 }</code></pre>
-            <button class="copy-btn bg-emerald-500 text-white px-4 py-2 rounded-lg mt-4 relative">Copy</button>
-          </div>
-        </div>
-        <div id="analysis" class="tab-content glass p-6">
-          <div class="code-block glass p-6 mb-6 relative">
-            <h3 class="text-xl font-semibold text-emerald-300 mb-2">POST /api/analysis</h3>
-            <p>Analyze code or data with detailed insights.</p>
-            <pre><code class="language-json">{
-  "text": "Analyze this Python code: print('Hello World')"
+                        <button class="copy-btn bg-orange-500 text-white px-4 py-2 rounded-lg mt-4">Copy</button>
+                    </div>
+                </div>
+                <div id="analysis" class="tab-content">
+                    <div class="code-block bg-gray-800 p-6 rounded-lg shadow-lg mb-6">
+                        <h3 class="text-xl font-semibold text-orange-500 mb-2">POST /api/analysis</h3>
+                        <p>Analyze code or data with detailed insights.</p>
+                        <pre><code class="language-json">{
+    "text": "Analyze this Python code: print('Hello World')"
 }</code></pre>
-            <button class="copy-btn bg-emerald-500 text-white px-4 py-2 rounded-lg mt-4 relative">Copy</button>
-          </div>
-        </div>
-        <div id="oauth" class="tab-content glass p-6">
-          <div class="code-block glass p-6 mb-6 relative">
-            <h3 class="text-xl font-semibold text-emerald-300 mb-2">OAuth 2.0 Authentication</h3>
-            <p>Authenticate with MGZon using OAuth 2.0.</p>
-            <pre><code class="language-bash">curl -X POST https://hager-zon.vercel.app/oauth/token \
+                        <button class="copy-btn bg-orange-500 text-white px-4 py-2 rounded-lg mt-4">Copy</button>
+                    </div>
+                </div>
+                <div id="oauth" class="tab-content">
+                    <div class="code-block bg-gray-800 p-6 rounded-lg shadow-lg mb-6">
+                        <h3 class="text-xl font-semibold text-orange-500 mb-2">OAuth 2.0 Authentication</h3>
+                        <p>Authenticate with MGZon using OAuth 2.0.</p>
+                        <pre><code class="language-bash">curl -X POST https://hager-zon.vercel.app/oauth/token \
 -H "Content-Type: application/json" \
 -d '{"client_id": "YOUR_CLIENT_ID", "client_secret": "YOUR_CLIENT_SECRET", "grant_type": "client_credentials"}'</code></pre>
-            <button class="copy-btn bg-emerald-500 text-white px-4 py-2 rounded-lg mt-4 relative">Copy</button>
-          </div>
-        </div>
-      </div>
-    </section>
-  </main>
-
-  <!-- Footer -->
-  <footer class="bg-gradient-to-r from-teal-900 to-emerald-900 py-12 mt-8">
-    <div class="container max-w-6xl mx-auto text-center">
-      <img
-        src="https://raw.githubusercontent.com/Mark-Lasfar/MGZon/main/public/icons/mg.svg"
-        alt="MGZon Logo"
-        class="w-24 h-24 mx-auto mb-6 animate-pulse"
-      />
-      <p class="mb-4">
-        Developed by
-        <a href="https://mark-elasfar.web.app/" target="_blank" class="text-emerald-300 hover:underline">Mark Al-Asfar</a>
-        | Powered by
-        <a href="https://hager-zon.vercel.app/" target="_blank" class="text-emerald-300 hover:underline">MGZon AI</a>
-      </p>
-      <div class="grid grid-cols-1 sm:grid-cols-2 md:grid-cols-3 gap-6">
-        <div class="glass p-4 cursor-pointer" onclick="showCardDetails('email')">
-          <i class="bx bx-mail-send text-3xl text-emerald-300 mb-2"></i>
-          <h4 class="font-semibold mb-1">Email Us</h4>
-          <p>
-            <a href="mailto:[email protected]" class="text-emerald-300 hover:underline">[email protected]</a>
-          </p>
-          <div id="email-details" class="hidden mt-4 p-4 bg-gray-700/80 rounded-lg">
-            <p>Reach out to our support team for any inquiries or assistance.</p>
-            <button onclick="closeCardDetails('email')" class="bg-emerald-500 text-white px-4 py-2 rounded-lg mt-2">Close</button>
-          </div>
+                        <button class="copy-btn bg-orange-500 text-white px-4 py-2 rounded-lg mt-4">Copy</button>
+                    </div>
+                </div>
+                <a href="/" class="inline-block bg-gradient-to-r from-orange-500 to-red-500 text-white px-8 py-4 rounded-full text-lg font-semibold hover:scale-105 transition transform mt-8">Back to Home</a>
+            </div>
         </div>
-        <div class="glass p-4 cursor-pointer" onclick="showCardDetails('phone')">
-          <i class="bx bx-phone text-3xl text-emerald-300 mb-2"></i>
-          <h4 class="font-semibold mb-1">Phone Support</h4>
-          <p>+1-800-123-4567</p>
-          <div id="phone-details" class="hidden mt-4 p-4 bg-gray-700/80 rounded-lg">
-            <p>Contact our support team via phone for immediate assistance.</p>
-            <button onclick="closeCardDetails('phone')" class="bg-emerald-500 text-white px-4 py-2 rounded-lg mt-2">Close</button>
-          </div>
-        </div>
-        <div class="glass p-4 cursor-pointer" onclick="showCardDetails('community')">
-          <i class="bx bx-group text-3xl text-emerald-300 mb-2"></i>
-          <h4 class="font-semibold mb-1">Community</h4>
-          <p>
-            <a href="https://hager-zon.vercel.app/community" class="text-emerald-300 hover:underline">Join us</a>
-          </p>
-          <div id="community-details" class="hidden mt-4 p-4 bg-gray-700/80 rounded-lg">
-            <p>Join our vibrant community to share ideas and collaborate.</p>
-            <button onclick="closeCardDetails('community')" class="bg-emerald-500 text-white px-4 py-2 rounded-lg mt-2">Close</button>
-          </div>
-        </div>
-        <div class="glass p-4 cursor-pointer" onclick="showCardDetails('api-docs')">
-          <i class="bx bx-code-alt text-3xl text-emerald-300 mb-2"></i>
-          <h4 class="font-semibold mb-1">API Docs</h4>
-          <p>
-            <a href="/docs" class="text-emerald-300 hover:underline">Explore Docs</a>
-          </p>
-          <div id="api-docs-details" class="hidden mt-4 p-4 bg-gray-700/80 rounded-lg">
-            <p>Explore our API documentation for seamless integration.</p>
-            <button onclick="closeCardDetails('api-docs')" class="bg-emerald-500 text-white px-4 py-2 rounded-lg mt-2">Close</button>
-          </div>
-        </div>
-        <div class="glass p-4 cursor-pointer" onclick="showCardDetails('faq')">
-          <i class="bx bx-help-circle text-3xl text-emerald-300 mb-2"></i>
-          <h4 class="font-semibold mb-1">FAQ</h4>
-          <p>
-            <a href="https://hager-zon.vercel.app/faq" target="_blank" class="text-emerald-300 hover:underline">Read FAQ</a>
-          </p>
-          <div id="faq-details" class="hidden mt-4 p-4 bg-gray-700/80 rounded-lg">
-            <p>Find answers to common questions in our FAQ section.</p>
-            <button onclick="closeCardDetails('faq')" class="bg-emerald-500 text-white px-4 py-2 rounded-lg mt-2">Close</button>
-          </div>
-        </div>
-        <div class="glass p-4 cursor-pointer" onclick="showCardDetails('docs')">
-          <i class="bx bx-book text-3xl text-emerald-300 mb-2"></i>
-          <h4 class="font-semibold mb-1">Documentation</h4>
-          <p>
-            <a href="/docs" class="text-emerald-300 hover:underline">Full Docs</a>
-          </p>
-          <div id="docs-details" class="hidden mt-4 p-4 bg-gray-700/80 rounded-lg">
-            <p>Access comprehensive documentation for MGZon Chatbot.</p>
-            <button onclick="closeCardDetails('docs')" class="bg-emerald-500 text-white px-4 py-2 rounded-lg mt-2">Close</button>
-          </div>
-        </div>
-      </div>
-      <div class="flex justify-center gap-6 mt-6">
-        <a href="https://github.com/Mark-Lasfar/MGZon" class="text-2xl text-white hover:text-emerald-300 transition">
-          <i class="bx bxl-github"></i>
-        </a>
-        <a href="https://x.com/MGZon" class="text-2xl text-white hover:text-emerald-300 transition">
-          <i class="bx bxl-twitter"></i>
-        </a>
-        <a href="https://www.facebook.com/people/Mark-Al-Asfar/pfbid02GMisUQ8AqWkNZjoKtWFHH1tbdHuVscN1cjcFnZWy9HkRaAsmanBfT6mhySAyqpg4l/" class="text-2xl text-white hover:text-emerald-300 transition">
-          <i class="bx bxl-facebook"></i>
-        </a>
-      </div>
-      <p class="mt-6">© 2025 Mark Al-Asfar & MGZon AI. All rights reserved.</p>
     </div>
-  </footer>
-
-  <!-- Scripts -->
-  <script>
-    // Sidebar toggle
-    const sidebar = document.getElementById('sidebar');
-    const toggleBtn = document.getElementById('sidebarToggle');
-    const closeSidebarBtn = document.getElementById('closeSidebar');
-
-    toggleBtn.addEventListener('click', () => {
-      sidebar.classList.toggle('active');
-      sidebar.classList.toggle('-translate-x-full');
-      sidebar.classList.toggle('translate-x-0');
-    });
-
-    closeSidebarBtn.addEventListener('click', () => {
-      sidebar.classList.remove('active');
-      sidebar.classList.add('-translate-x-full');
-      sidebar.classList.remove('translate-x-0');
-    });
-
-    document.addEventListener('click', (e) => {
-      if (
-        !sidebar.contains(e.target) &&
-        !toggleBtn.contains(e.target) &&
-        window.innerWidth < 768
-      ) {
-        sidebar.classList.remove('active');
-        sidebar.classList.add('-translate-x-full');
-        sidebar.classList.remove('translate-x-0');
-      }
-    });
-
-    // Tab switching
-    document.querySelectorAll('.tab-btn').forEach(btn => {
-      btn.addEventListener('click', () => {
-        document.querySelectorAll('.tab-btn').forEach(b => b.classList.remove('bg-emerald-500'));
-        btn.classList.add('bg-emerald-500');
-        document.querySelectorAll('.tab-content').forEach(tab => tab.classList.remove('active'));
-        document.getElementById(btn.dataset.tab).classList.add('active');
-      });
-    });
-
-    // Copy button logic
-    document.querySelectorAll('.copy-btn').forEach(btn => {
-      btn.addEventListener('click', () => {
-        const code = btn.previousElementSibling.querySelector('code').textContent;
-        navigator.clipboard.writeText(code).then(() => {
-          btn.classList.add('copied');
-          setTimeout(() => btn.classList.remove('copied'), 2000);
+    <footer class="bg-gradient-to-r from-blue-900 to-gray-900 py-12 mt-auto">
+        <div class="container max-w-6xl mx-auto text-center">
+            <img src="https://raw.githubusercontent.com/Mark-Lasfar/MGZon/main/public/icons/mg.svg" alt="MGZon Logo" class="w-24 h-24 mx-auto mb-6 animate-pulse">
+            <p class="mb-4">Developed by <a href="https://mark-elasfar.web.app/" target="_blank" class="text-orange-500 hover:underline">Mark Al-Asfar</a> | Powered by <a href="https://hager-zon.vercel.app/" target="_blank" class="text-orange-500 hover:underline">MGZon AI</a></p>
+            <div class="grid grid-cols-2 md:grid-cols-3 gap-6">
+                <div class="bg-gray-800 p-6 rounded-lg shadow-lg">
+                    <i class="bx bx-mail-send text-3xl text-orange-500 mb-4"></i>
+                    <h3 class="text-lg font-semibold mb-2">Email Us</h3>
+                    <p>Contact: <a href="mailto:[email protected]" class="text-orange-500 hover:underline">[email protected]</a></p>
+                </div>
+                <div class="bg-gray-800 p-6 rounded-lg shadow-lg">
+                    <i class="bx bx-phone text-3xl text-orange-500 mb-4"></i>
+                    <h3 class="text-lg font-semibold mb-2">Phone Support</h3>
+                    <p>Call: +1-800-123-4567</p>
+                </div>
+                <div class="bg-gray-800 p-6 rounded-lg shadow-lg">
+                    <i class="bx bx-group text-3xl text-orange-500 mb-4"></i>
+                    <h3 class="text-lg font-semibold mb-2">Community</h3>
+                    <p>Join: <a href="https://hager-zon.vercel.app/community" class="text-orange-500 hover:underline">mgzon.com/community</a></p>
+                </div>
+                <div class="bg-gray-800 p-6 rounded-lg shadow-lg">
+                    <i class="bx bx-code-alt text-3xl text-orange-500 mb-4"></i>
+                    <h3 class="text-lg font-semibold mb-2">API Docs</h3>
+                    <p>Explore: <a href="/docs" class="text-orange-500 hover:underline">mgzon.com/docs</a></p>
+                </div>
+                <div class="bg-gray-800 p-6 rounded-lg shadow-lg">
+                    <i class="bx bx-help-circle text-3xl text-orange-500 mb-4"></i>
+                    <h3 class="text-lg font-semibold mb-2">FAQ</h3>
+                    <p>Read: <a href="https://hager-zon.vercel.app/faq" class="text-orange-500 hover:underline">mgzon.com/faq</a></p>
+                </div>
+                <div class="bg-gray-800 p-6 rounded-lg shadow-lg">
+                    <i class="bx bx-book text-3xl text-orange-500 mb-4"></i>
+                    <h3 class="text-lg font-semibold mb-2">Documentation</h3>
+                    <p>Learn: <a href="/docs" class="text-orange-500 hover:underline">mgzon.com/docs</a></p>
+                </div>
+            </div>
+            <div class="flex justify-center gap-6 mt-6">
+                <a href="https://github.com/Mark-Lasfar/MGZon" class="text-2xl text-white hover:text-orange-500 transition"><i class="bx bxl-github"></i></a>
+                <a href="https://x.com/MGZon" class="text-2xl text-white hover:text-orange-500 transition"><i class="bx bxl-twitter"></i></a>
+                <a href="https://www.facebook.com/people/Mark-Al-Asfar/pfbid02GMisUQ8AqWkNZjoKtWFHH1tbdHuVscN1cjcFnZWy9HkRaAsmanBfT6mhySAyqpg4l/" class="text-2xl text-white hover:text-orange-500 transition"><i class="bx bxl-facebook"></i></a>
+            </div>
+            <p class="mt-6">© 2025 Mark Al-Asfar & MGZon AI. All rights reserved.</p>
+        </div>
+    </footer>
+    <script src="/static/js/scripts.js"></script>
+    <script src="https://cdn.jsdelivr.net/npm/[email protected]/prism.min.js"></script>
+    <script>
+        document.querySelectorAll('.tab-btn').forEach(btn => {
+            btn.addEventListener('click', () => {
+                document.querySelectorAll('.tab-btn').forEach(b => b.classList.replace('bg-orange-500', 'bg-gray-700'));
+                btn.classList.replace('bg-gray-700', 'bg-orange-500');
+                document.querySelectorAll('.tab-content').forEach(tab => tab.classList.remove('active'));
+                document.getElementById(btn.dataset.tab).classList.add('active');
+            });
         });
-      });
-    });
-
-    // Card details toggle
-    function showCardDetails(cardId) {
-      document.getElementById(`${cardId}-details`).classList.remove('hidden');
-    }
-    function closeCardDetails(cardId) {
-      document.getElementById(`${cardId}-details`).classList.add('hidden');
-    }
-  </script>
+    </script>
 </body>
 </html>

utils/generation.py

@@ -8,6 +8,12 @@ from tenacity import retry, stop_after_attempt, wait_exponential
 import logging
 from cachetools import TTLCache
 import hashlib
+import requests
+import pydub
+import io
+import torchaudio
+from PIL import Image
+from transformers import CLIPModel, CLIPProcessor, AutoProcessor, ParlerTTSForConditionalGeneration
 
 logger = logging.getLogger(__name__)
 
@@ -24,14 +30,59 @@ LATEX_DELIMS = [
 
 # إعداد العميل لـ Hugging Face Inference API
 HF_TOKEN = os.getenv("HF_TOKEN")
+BACKUP_HF_TOKEN = os.getenv("BACKUP_HF_TOKEN")  # توكن احتياطي
 API_ENDPOINT = os.getenv("API_ENDPOINT", "https://router.huggingface.co/v1")
 FALLBACK_API_ENDPOINT = "https://api-inference.huggingface.co/v1"
 MODEL_NAME = os.getenv("MODEL_NAME", "openai/gpt-oss-20b:fireworks-ai")
-SECONDARY_MODEL_NAME = os.getenv("SECONDARY_MODEL_NAME", "MGZON/Veltrix")
+SECONDARY_MODEL_NAME = os.getenv("SECONDARY_MODEL_NAME", "deepseek-ai/DeepSeek-R1-Distill-Qwen-7B")
 TERTIARY_MODEL_NAME = os.getenv("TERTIARY_MODEL_NAME", "mistralai/Mixtral-8x7B-Instruct-v0.1")
+CLIP_BASE_MODEL = os.getenv("CLIP_BASE_MODEL", "openai/clip-vit-base-patch32")
+CLIP_LARGE_MODEL = os.getenv("CLIP_LARGE_MODEL", "openai/clip-vit-large-patch14")
+ASR_MODEL = os.getenv("ASR_MODEL", "openai/whisper-large-v3-turbo")
+TTS_MODEL = os.getenv("TTS_MODEL", "parler-tts/parler-tts-mini-v1")
 
-def select_model(query: str) -> tuple[str, str]:
+def check_model_availability(model_name: str, api_base: str, api_key: str) -> tuple[bool, str]:
+    """التحقق من توفر النموذج عبر API مع دعم التوكن الاحتياطي"""
+    try:
+        response = requests.get(
+            f"{api_base}/models/{model_name}",
+            headers={"Authorization": f"Bearer {api_key}"},
+            timeout=10
+        )
+        if response.status_code == 200:
+            return True, api_key
+        elif response.status_code == 429 and BACKUP_HF_TOKEN and api_key != BACKUP_HF_TOKEN:
+            logger.warning(f"Rate limit reached for token {api_key}. Switching to backup token.")
+            return check_model_availability(model_name, api_base, BACKUP_HF_TOKEN)
+        logger.error(f"Model {model_name} not available: {response.status_code}")
+        return False, api_key
+    except Exception as e:
+        logger.error(f"Failed to check model availability for {model_name}: {e}")
+        if BACKUP_HF_TOKEN and api_key != BACKUP_HF_TOKEN:
+            logger.warning(f"Retrying with backup token for {model_name}")
+            return check_model_availability(model_name, api_base, BACKUP_HF_TOKEN)
+        return False, api_key
+
+def select_model(query: str, input_type: str = "text") -> tuple[str, str]:
     query_lower = query.lower()
+    # دعم الصوت
+    if input_type == "audio" or any(keyword in query_lower for keyword in ["voice", "audio", "speech", "صوت", "تحويل صوت"]):
+        logger.info(f"Selected {ASR_MODEL} with endpoint {FALLBACK_API_ENDPOINT} for audio input")
+        return ASR_MODEL, FALLBACK_API_ENDPOINT
+    # دعم تحويل النص إلى صوت
+    if any(keyword in query_lower for keyword in ["text-to-speech", "tts", "تحويل نص إلى صوت"]):
+        logger.info(f"Selected {TTS_MODEL} with endpoint {FALLBACK_API_ENDPOINT} for text-to-speech")
+        return TTS_MODEL, FALLBACK_API_ENDPOINT
+    # نماذج CLIP للاستعلامات المتعلقة بالصور
+    image_patterns = [
+        r"\bimage\b", r"\bpicture\b", r"\bphoto\b", r"\bvisual\b", r"\bصورة\b", r"\bتحليل\s+صورة\b",
+        r"\bimage\s+analysis\b", r"\bimage\s+classification\b", r"\bimage\s+description\b"
+    ]
+    for pattern in image_patterns:
+        if re.search(pattern, query_lower, re.IGNORECASE):
+            logger.info(f"Selected {CLIP_BASE_MODEL} with endpoint {FALLBACK_API_ENDPOINT} for image-related query: {query}")
+            return CLIP_BASE_MODEL, FALLBACK_API_ENDPOINT
+    # نموذج DeepSeek للاستعلامات المتعلقة بـ MGZon
     mgzon_patterns = [
         r"\bmgzon\b", r"\bmgzon\s+(products|services|platform|features|mission|technology|solutions|oauth)\b",
         r"\bميزات\s+mgzon\b", r"\bخدمات\s+mgzon\b", r"\boauth\b"
@@ -40,10 +91,11 @@ def select_model(query: str) -> tuple[str, str]:
         if re.search(pattern, query_lower, re.IGNORECASE):
             logger.info(f"Selected {SECONDARY_MODEL_NAME} with endpoint {FALLBACK_API_ENDPOINT} for MGZon-related query: {query}")
             return SECONDARY_MODEL_NAME, FALLBACK_API_ENDPOINT
+    # النموذج الافتراضي للاستعلامات العامة
     logger.info(f"Selected {MODEL_NAME} with endpoint {API_ENDPOINT} for general query: {query}")
     return MODEL_NAME, API_ENDPOINT
 
-@retry(stop=stop_after_attempt(5), wait=wait_exponential(multiplier=2, min=4, max=30))
+@retry(stop=stop_after_attempt(3), wait=wait_exponential(multiplier=2, min=4, max=60))
 def request_generation(
     api_key: str,
     api_base: str,
@@ -51,16 +103,25 @@ def request_generation(
     system_prompt: str,
     model_name: str,
     chat_history: Optional[List[dict]] = None,
-    temperature: float = 0.9,
+    temperature: float = 0.7,
     max_new_tokens: int = 128000,
     reasoning_effort: str = "off",
     tools: Optional[List[dict]] = None,
     tool_choice: Optional[str] = None,
     deep_search: bool = False,
-) -> Generator[str, None, None]:
+    input_type: str = "text",
+    audio_data: Optional[bytes] = None,
+    image_data: Optional[bytes] = None,
+) -> Generator[bytes | str, None, None]:
     from utils.web_search import web_search  # تأخير الاستيراد
 
-    # إنشاء مفتاح للـ cache بناءً على المدخلات
+    # التحقق من توفر النموذج مع دعم التوكن الاحتياطي
+    is_available, selected_api_key = check_model_availability(model_name, api_base, api_key)
+    if not is_available:
+        yield f"Error: Model {model_name} is not available. Please check the model endpoint or token."
+        return
+
+    # إنشاء مفتاح للـ cache
     cache_key = hashlib.md5(json.dumps({
         "message": message,
         "system_prompt": system_prompt,
@@ -76,22 +137,93 @@ def request_generation(
             yield chunk
         return
 
-    client = OpenAI(api_key=api_key, base_url=api_base, timeout=60.0)
+    client = OpenAI(api_key=selected_api_key, base_url=api_base, timeout=120.0)
     task_type = "general"
-    if any(keyword in message.lower() for keyword in ["code", "programming", "python", "javascript", "react", "django", "flask"]):
+    enhanced_system_prompt = system_prompt
+
+    # معالجة الصوت (ASR)
+    if model_name == ASR_MODEL and audio_data:
+        task_type = "audio_transcription"
+        try:
+            audio_file = io.BytesIO(audio_data)
+            audio = pydub.AudioSegment.from_file(audio_file)
+            audio = audio.set_channels(1).set_frame_rate(16000)
+            audio_file = io.BytesIO()
+            audio.export(audio_file, format="wav")
+            audio_file.name = "audio.wav"
+            transcription = client.audio.transcriptions.create(
+                model=model_name,
+                file=audio_file,
+                response_format="text"
+            )
+            yield transcription
+            cache[cache_key] = [transcription]
+            return
+        except Exception as e:
+            logger.error(f"Audio transcription failed: {e}")
+            yield f"Error: Audio transcription failed: {e}"
+            return
+
+    # معالجة تحويل النص إلى صوت (TTS)
+    if model_name == TTS_MODEL:
+        task_type = "text_to_speech"
+        try:
+            model = ParlerTTSForConditionalGeneration.from_pretrained(model_name)
+            processor = AutoProcessor.from_pretrained(model_name)
+            inputs = processor(text=message, return_tensors="pt")
+            audio = model.generate(**inputs)
+            audio_file = io.BytesIO()
+            torchaudio.save(audio_file, audio[0], sample_rate=22050, format="wav")
+            audio_file.seek(0)
+            yield audio_file.read()
+            cache[cache_key] = [audio_file.read()]
+            return
+        except Exception as e:
+            logger.error(f"Text-to-speech failed: {e}")
+            yield f"Error: Text-to-speech failed: {e}"
+            return
+
+    # معالجة الصور
+    if model_name in [CLIP_BASE_MODEL, CLIP_LARGE_MODEL] and image_data:
+        task_type = "image_analysis"
+        try:
+            model = CLIPModel.from_pretrained(model_name)
+            processor = CLIPProcessor.from_pretrained(model_name)
+            image = Image.open(io.BytesIO(image_data)).convert("RGB")
+            inputs = processor(text=message, images=image, return_tensors="pt", padding=True)
+            outputs = model(**inputs)
+            logits_per_image = outputs.logits_per_image
+            probs = logits_per_image.softmax(dim=1)
+            yield f"Image analysis result: {probs.tolist()}"
+            cache[cache_key] = [f"Image analysis result: {probs.tolist()}"]
+            return
+        except Exception as e:
+            logger.error(f"Image analysis failed: {e}")
+            yield f"Error: Image analysis failed: {e}"
+            return
+
+    # تحسين system_prompt بناءً على نوع المهمة
+    if model_name in [CLIP_BASE_MODEL, CLIP_LARGE_MODEL]:
+        task_type = "image"
+        enhanced_system_prompt = f"{system_prompt}\nYou are an expert in image analysis and description. Provide detailed descriptions, classifications, or analysis of images based on the query. Continue until the query is fully addressed."
+    elif any(keyword in message.lower() for keyword in ["code", "programming", "python", "javascript", "react", "django", "flask"]):
         task_type = "code"
-        enhanced_system_prompt = f"{system_prompt}\nYou are an expert programmer. Provide accurate, well-commented code with examples and explanations. Support frameworks like React, Django, Flask, and others as needed. Format code with triple backticks (```) and specify the language."
+        enhanced_system_prompt = f"{system_prompt}\nYou are an expert programmer. Provide accurate, well-commented code with comprehensive examples and detailed explanations. Support frameworks like React, Django, Flask, and others. Format code with triple backticks (```) and specify the language. Continue until the task is fully addressed."
     elif any(keyword in message.lower() for keyword in ["analyze", "analysis", "تحليل"]):
         task_type = "analysis"
-        enhanced_system_prompt = f"{system_prompt}\nProvide detailed analysis with step-by-step reasoning, examples, and data-driven insights."
+        enhanced_system_prompt = f"{system_prompt}\nProvide detailed analysis with step-by-step reasoning, examples, and data-driven insights. Continue until all aspects of the query are thoroughly covered."
     elif any(keyword in message.lower() for keyword in ["review", "مراجعة"]):
         task_type = "review"
-        enhanced_system_prompt = f"{system_prompt}\nReview the provided content thoroughly, identify issues, and suggest improvements with detailed explanations."
+        enhanced_system_prompt = f"{system_prompt}\nReview the provided content thoroughly, identify issues, and suggest improvements with detailed explanations. Ensure the response is complete and detailed."
     elif any(keyword in message.lower() for keyword in ["publish", "نشر"]):
         task_type = "publish"
-        enhanced_system_prompt = f"{system_prompt}\nPrepare content for publishing, ensuring clarity, professionalism, and adherence to best practices."
+        enhanced_system_prompt = f"{system_prompt}\nPrepare content for publishing, ensuring clarity, professionalism, and adherence to best practices. Provide a complete and detailed response."
     else:
-        enhanced_system_prompt = system_prompt
+        enhanced_system_prompt = f"{system_prompt}\nFor general queries, provide comprehensive, detailed responses with examples and explanations where applicable. Continue generating content until the query is fully answered, leveraging the full capacity of the model."
+
+    # إذا كان الاستعلام قصيرًا، شجع على التفصيل
+    if len(message.split()) < 5:
+        enhanced_system_prompt += "\nEven for short or general queries, provide a detailed, in-depth response with examples, explanations, and additional context to ensure completeness."
 
     logger.info(f"Task type detected: {task_type}")
     input_messages: List[dict] = [{"role": "system", "content": enhanced_system_prompt}]
@@ -111,8 +243,8 @@ def request_generation(
     else:
         input_messages.append({"role": "user", "content": message})
 
-    tools = tools if tools and "gpt-oss" in model_name else []
-    tool_choice = tool_choice if tool_choice in ["auto", "none", "any", "required"] and "gpt-oss" in model_name else "none"
+    tools = tools if tools and model_name in [MODEL_NAME, SECONDARY_MODEL_NAME, TERTIARY_MODEL_NAME] else []
+    tool_choice = tool_choice if tool_choice in ["auto", "none", "any", "required"] and model_name in [MODEL_NAME, SECONDARY_MODEL_NAME, TERTIARY_MODEL_NAME] else "none"
 
     cached_chunks = []
     try:
@@ -152,13 +284,13 @@ def request_generation(
                 saw_visible_output = True
                 buffer += content
 
-                if "\n" in buffer or len(buffer) > 1000:  # تقليل حجم الـ buffer
+                if "\n" in buffer or len(buffer) > 5000:
                     cached_chunks.append(buffer)
                     yield buffer
                     buffer = ""
                 continue
 
-            if chunk.choices[0].delta.tool_calls and "gpt-oss" in model_name:
+            if chunk.choices[0].delta.tool_calls and model_name in [MODEL_NAME, SECONDARY_MODEL_NAME, TERTIARY_MODEL_NAME]:
                 tool_call = chunk.choices[0].delta.tool_calls[0]
                 name = getattr(tool_call, "function", {}).get("name", None)
                 args = getattr(tool_call, "function", {}).get("arguments", None)
@@ -168,7 +300,7 @@ def request_generation(
                     last_tool_args = args
                 continue
 
-            if chunk.choices[0].finish_reason in ("stop", "tool_calls", "error"):
+            if chunk.choices[0].finish_reason in ("stop", "tool_calls", "error", "length"):
                 if buffer:
                     cached_chunks.append(buffer)
                     yield buffer
@@ -193,22 +325,50 @@ def request_generation(
                 if chunk.choices[0].finish_reason == "error":
                     cached_chunks.append(f"Error: Unknown error")
                     yield f"Error: Unknown error"
+                elif chunk.choices[0].finish_reason == "length":
+                    cached_chunks.append("Response truncated due to token limit. Please refine your query or request continuation.")
+                    yield "Response truncated due to token limit. Please refine your query or request continuation."
                 break
 
         if buffer:
             cached_chunks.append(buffer)
             yield buffer
 
-        cache[cache_key] = cached_chunks  # حفظ في الـ cache
+        cache[cache_key] = cached_chunks
 
     except Exception as e:
         logger.exception(f"[Gateway] Streaming failed for model {model_name}: {e}")
+        if selected_api_key != BACKUP_HF_TOKEN and BACKUP_HF_TOKEN:
+            logger.warning(f"Retrying with backup token for model {model_name}")
+            for chunk in request_generation(
+                api_key=BACKUP_HF_TOKEN,
+                api_base=api_base,
+                message=message,
+                system_prompt=system_prompt,
+                model_name=model_name,
+                chat_history=chat_history,
+                temperature=temperature,
+                max_new_tokens=max_new_tokens,
+                reasoning_effort=reasoning_effort,
+                tools=tools,
+                tool_choice=tool_choice,
+                deep_search=deep_search,
+                input_type=input_type,
+                audio_data=audio_data,
+                image_data=image_data,
+            ):
+                yield chunk
+            return
         if model_name == MODEL_NAME:
             fallback_model = SECONDARY_MODEL_NAME
             fallback_endpoint = FALLBACK_API_ENDPOINT
             logger.info(f"Retrying with fallback model: {fallback_model} on {fallback_endpoint}")
             try:
-                client = OpenAI(api_key=api_key, base_url=fallback_endpoint, timeout=60.0)
+                is_available, selected_api_key = check_model_availability(fallback_model, fallback_endpoint, selected_api_key)
+                if not is_available:
+                    yield f"Error: Fallback model {fallback_model} is not available."
+                    return
+                client = OpenAI(api_key=selected_api_key, base_url=fallback_endpoint, timeout=120.0)
                 stream = client.chat.completions.create(
                     model=fallback_model,
                     messages=input_messages,
@@ -237,13 +397,13 @@ def request_generation(
                         saw_visible_output = True
                         buffer += content
 
-                        if "\n" in buffer or len(buffer) > 1000:
+                        if "\n" in buffer or len(buffer) > 5000:
                             cached_chunks.append(buffer)
                             yield buffer
                             buffer = ""
                         continue
 
-                    if chunk.choices[0].finish_reason in ("stop", "error"):
+                    if chunk.choices[0].finish_reason in ("stop", "error", "length"):
                         if buffer:
                             cached_chunks.append(buffer)
                             yield buffer
@@ -260,6 +420,9 @@ def request_generation(
                         if chunk.choices[0].finish_reason == "error":
                             cached_chunks.append(f"Error: Unknown error with fallback model {fallback_model}")
                             yield f"Error: Unknown error with fallback model {fallback_model}"
+                        elif chunk.choices[0].finish_reason == "length":
+                            cached_chunks.append("Response truncated due to token limit. Please refine your query or request continuation.")
+                            yield "Response truncated due to token limit. Please refine your query or request continuation."
                         break
 
                 if buffer:
@@ -270,10 +433,12 @@ def request_generation(
 
             except Exception as e2:
                 logger.exception(f"[Gateway] Streaming failed for fallback model {fallback_model}: {e2}")
-                cached_chunks.append(f"Error: Failed to load both models ({model_name} and {fallback_model}): {e2}")
-                yield f"Error: Failed to load both models ({model_name} and {fallback_model}): {e2}"
                 try:
-                    client = OpenAI(api_key=api_key, base_url=FALLBACK_API_ENDPOINT, timeout=60.0)
+                    is_available, selected_api_key = check_model_availability(TERTIARY_MODEL_NAME, FALLBACK_API_ENDPOINT, selected_api_key)
+                    if not is_available:
+                        yield f"Error: Tertiary model {TERTIARY_MODEL_NAME} is not available."
+                        return
+                    client = OpenAI(api_key=selected_api_key, base_url=FALLBACK_API_ENDPOINT, timeout=120.0)
                     stream = client.chat.completions.create(
                         model=TERTIARY_MODEL_NAME,
                         messages=input_messages,
@@ -288,12 +453,12 @@ def request_generation(
                             content = chunk.choices[0].delta.content
                             saw_visible_output = True
                             buffer += content
-                            if "\n" in buffer or len(buffer) > 1000:
+                            if "\n" in buffer or len(buffer) > 5000:
                                 cached_chunks.append(buffer)
                                 yield buffer
                                 buffer = ""
                             continue
-                        if chunk.choices[0].finish_reason in ("stop", "error"):
+                        if chunk.choices[0].finish_reason in ("stop", "error", "length"):
                             if buffer:
                                 cached_chunks.append(buffer)
                                 yield buffer
@@ -304,6 +469,9 @@ def request_generation(
                             if chunk.choices[0].finish_reason == "error":
                                 cached_chunks.append(f"Error: Unknown error with tertiary model {TERTIARY_MODEL_NAME}")
                                 yield f"Error: Unknown error with tertiary model {TERTIARY_MODEL_NAME}"
+                            elif chunk.choices[0].finish_reason == "length":
+                                cached_chunks.append("Response truncated due to token limit. Please refine your query or request continuation.")
+                                yield "Response truncated due to token limit. Please refine your query or request continuation."
                             break
                     if buffer:
                         cached_chunks.append(buffer)
@@ -311,30 +479,32 @@ def request_generation(
                     cache[cache_key] = cached_chunks
                 except Exception as e3:
                     logger.exception(f"[Gateway] Streaming failed for tertiary model {TERTIARY_MODEL_NAME}: {e3}")
-                    cached_chunks.append(f"Error: Failed to load all models: {e3}")
-                    yield f"Error: Failed to load all models: {e3}"
+                    yield f"Error: Failed to load all models: Primary ({model_name}), Secondary ({fallback_model}), Tertiary ({TERTIARY_MODEL_NAME}). Please check your model configurations."
+                    return
         else:
-            cached_chunks.append(f"Error: Failed to load model {model_name}: {e}")
             yield f"Error: Failed to load model {model_name}: {e}"
+            return
 
 def format_final(analysis_text: str, visible_text: str) -> str:
     reasoning_safe = html.escape((analysis_text or "").strip())
     response = (visible_text or "").strip()
+    if not reasoning_safe and not response:
+        return "No response generated."
     return (
         "<details><summary><strong>🤔 Analysis</strong></summary>\n"
         "<pre style='white-space:pre-wrap;'>"
         f"{reasoning_safe}"
         "</pre>\n</details>\n\n"
         "**💬 Response:**\n\n"
-        f"{response}"
+        f"{response}" if response else "No final response available."
     )
 
-def generate(message, history, system_prompt, temperature, reasoning_effort, enable_browsing, max_new_tokens):
-    if not message.strip():
-        yield "Please enter a prompt."
+def generate(message, history, system_prompt, temperature, reasoning_effort, enable_browsing, max_new_tokens, input_type="text", audio_data=None, image_data=None):
+    if not message.strip() and not audio_data and not image_data:
+        yield "Please enter a prompt or upload a file."
         return
 
-    model_name, api_endpoint = select_model(message)
+    model_name, api_endpoint = select_model(message, input_type=input_type)
     chat_history = []
     for h in history:
         if isinstance(h, dict):
@@ -389,9 +559,24 @@ def generate(message, history, system_prompt, temperature, reasoning_effort, ena
                     "required": ["code", "language"],
                 },
             },
+        },
+        {
+            "type": "function",
+            "function": {
+                "name": "image_analysis",
+                "description": "Analyze or describe an image based on the provided query",
+                "parameters": {
+                    "type": "object",
+                    "properties": {
+                        "image_url": {"type": "string", "description": "URL of the image to analyze"},
+                        "task": {"type": "string", "description": "Task description (e.g., describe, classify)"},
+                    },
+                    "required": ["task"],
+                },
+            },
         }
-    ] if "gpt-oss" in model_name else []
-    tool_choice = "auto" if "gpt-oss" in model_name else "none"
+    ] if model_name in [MODEL_NAME, SECONDARY_MODEL_NAME, TERTIARY_MODEL_NAME] else []
+    tool_choice = "auto" if model_name in [MODEL_NAME, SECONDARY_MODEL_NAME, TERTIARY_MODEL_NAME] else "none"
 
     in_analysis = False
     in_visible = False
@@ -423,9 +608,15 @@ def generate(message, history, system_prompt, temperature, reasoning_effort, ena
             tools=tools,
             tool_choice=tool_choice,
             deep_search=enable_browsing,
+            input_type=input_type,
+            audio_data=audio_data,
+            image_data=image_data,
         )
 
         for chunk in stream:
+            if isinstance(chunk, bytes):
+                yield chunk
+                continue
             if chunk == "analysis":
                 in_analysis, in_visible = True, False
                 if not raw_started: