remove non HF models

backend_api.py

@@ -24,7 +24,6 @@ import httpx
 from backend_models import (
     get_inference_client, 
     get_real_model_id,
-    create_gemini3_messages,
     is_native_sdk_model,
     is_mistral_model
 )
@@ -100,22 +99,11 @@ def get_cached_client(model_id: str, provider: str = "auto"):
 # Define models and languages here to avoid importing Gradio UI
 AVAILABLE_MODELS = [
     {"name": "GLM-4.6", "id": "zai-org/GLM-4.6", "description": "GLM-4.6 model via HuggingFace with Cerebras provider (Default)", "supports_images": False},
-    {"name": "Devstral Medium 2512", "id": "devstral-medium-2512", "description": "Mistral Devstral 2512 - Expert code generation model via OpenRouter", "supports_images": False},
     {"name": "GLM-4.6V 👁️", "id": "zai-org/GLM-4.6V:zai-org", "description": "GLM-4.6V vision model - supports image uploads for visual understanding", "supports_images": True},
-    {"name": "DeepSeek V3.2", "id": "deepseek-ai/DeepSeek-V3.2-Exp", "description": "DeepSeek V3.2 Experimental - Fast model for code generation via HuggingFace Router with Novita provider", "supports_images": False},
-    {"name": "DeepSeek R1", "id": "deepseek-ai/DeepSeek-R1-0528", "description": "DeepSeek R1 model for code generation", "supports_images": False},
-    {"name": "Gemini 3.0 Flash Preview", "id": "gemini-3-flash-preview", "description": "Google Gemini 3.0 Flash Preview with Thinking Mode (High)", "supports_images": False},
-    {"name": "Gemini 3.0 Pro", "id": "gemini-3.0-pro", "description": "Google Gemini 3.0 Pro via Poe with advanced reasoning", "supports_images": False},
-    {"name": "Grok 4.1 Fast", "id": "x-ai/grok-4.1-fast", "description": "Grok 4.1 Fast model via OpenRouter (20 req/min on free tier)", "supports_images": False},
+    {"name": "DeepSeek V3", "id": "deepseek-ai/DeepSeek-V3", "description": "DeepSeek V3 - Fast model for code generation via HuggingFace Router with Novita provider", "supports_images": False},
+    {"name": "DeepSeek R1", "id": "deepseek-ai/DeepSeek-R1", "description": "DeepSeek R1 model for code generation via HuggingFace", "supports_images": False},
     {"name": "MiniMax M2", "id": "MiniMaxAI/MiniMax-M2", "description": "MiniMax M2 model via HuggingFace InferenceClient with Novita provider", "supports_images": False},
-    {"name": "GPT-5.1", "id": "gpt-5.1", "description": "OpenAI GPT-5.1 model via Poe for advanced code generation and general tasks", "supports_images": False},
-    {"name": "GPT-5.1 Instant", "id": "gpt-5.1-instant", "description": "OpenAI GPT-5.1 Instant model via Poe for fast responses", "supports_images": False},
-    {"name": "GPT-5.1 Codex", "id": "gpt-5.1-codex", "description": "OpenAI GPT-5.1 Codex model via Poe optimized for code generation", "supports_images": False},
-    {"name": "Claude-Opus-4.5", "id": "claude-opus-4.5", "description": "Anthropic Claude Opus 4.5 via Poe (OpenAI-compatible)", "supports_images": False},
-    {"name": "Claude-Sonnet-4.5", "id": "claude-sonnet-4.5", "description": "Anthropic Claude Sonnet 4.5 via Poe (OpenAI-compatible)", "supports_images": False},
-    {"name": "Claude-Haiku-4.5", "id": "claude-haiku-4.5", "description": "Anthropic Claude Haiku 4.5 via Poe (OpenAI-compatible)", "supports_images": False},
     {"name": "Kimi K2 Thinking", "id": "moonshotai/Kimi-K2-Thinking", "description": "Moonshot Kimi K2 Thinking model via HuggingFace with Together AI provider", "supports_images": False},
-    {"name": "GLM-4.6", "id": "zai-org/GLM-4.6", "description": "GLM-4.6 model via HuggingFace with Cerebras provider", "supports_images": False},
 ]
 
 # Cache model lookup for faster access (built after AVAILABLE_MODELS is defined)
@@ -843,71 +831,29 @@ async def generate_code(
             
             # Stream the response
             try:
-                # Handle Mistral models with different API
-                if is_mistral_model(selected_model_id):
-                    print(f"[Generate] Using Mistral SDK for {selected_model_id}")
-                    
-                    # Mistral models use the standard chat.stream API
-                    stream = client.chat.stream(
-                        model=actual_model_id,
-                        messages=messages,
-                        max_tokens=10000
-                    )
-
-                # Handle Native SDK models (Gemini 3)
-                elif is_native_sdk_model(selected_model_id):
-                    print(f"[Generate] Using Native SDK (Gemini) for {selected_model_id}")
-                    
-                    if selected_model_id == "gemini-3-flash-preview":
-                        contents, config = create_gemini3_messages(messages)
-                        stream = client.models.generate_content_stream(
-                           model=selected_model_id,
-                           contents=contents,
-                           config=config
-                        )
-                    else:
-                        raise ValueError(f"Unknown native SDK model: {selected_model_id}")
-                
-                # All other models use OpenAI-compatible API
-                else:
-                    stream = client.chat.completions.create(
-                        model=actual_model_id,
-                        messages=messages,
-                        temperature=0.7,
-                        max_tokens=10000,
-                        stream=True
-                    )
+                # All models now use OpenAI-compatible API via HF Router or Inference API
+                stream = client.chat.completions.create(
+                    model=actual_model_id,
+                    messages=messages,
+                    temperature=0.7,
+                    max_tokens=10000,
+                    stream=True
+                )
                 
                 chunk_count = 0
-                is_mistral = is_mistral_model(selected_model_id)
-                is_native = is_native_sdk_model(selected_model_id)
                 
-                # Only process stream if it exists (not None for Conversations API)
+                # Only process stream if it exists
                 if stream:
-                    # Optimized chunk processing - reduce attribute lookups
+                    # Optimized chunk processing
                     for chunk in stream:
                         chunk_content = None
                         
-                        if is_native:
-                            # Native SDK format (Gemini)
-                            try:
-                                chunk_content = chunk.text
-                            except (AttributeError, ValueError):
-                                continue
-                        elif is_mistral:
-                            # Mistral format: chunk.data.choices[0].delta.content
-                            try:
-                                if chunk.data and chunk.data.choices and chunk.data.choices[0].delta.content:
-                                    chunk_content = chunk.data.choices[0].delta.content
-                            except (AttributeError, IndexError):
-                                continue
-                        else:
-                            # OpenAI format: chunk.choices[0].delta.content
-                            try:
-                                if chunk.choices and chunk.choices[0].delta.content:
-                                    chunk_content = chunk.choices[0].delta.content
-                            except (AttributeError, IndexError):
-                                continue
+                        # OpenAI format: chunk.choices[0].delta.content
+                        try:
+                            if chunk.choices and chunk.choices[0].delta.content:
+                                chunk_content = chunk.choices[0].delta.content
+                        except (AttributeError, IndexError):
+                            continue
                         
                         if chunk_content:
                             generated_code += chunk_content

backend_models.py

@@ -6,15 +6,6 @@ import os
 from typing import Optional
 
 from openai import OpenAI
-from mistralai import Mistral
-
-# Import genai for Gemini (legacy - no longer used with Poe API)
-try:
-    from google import genai
-    from google.genai import types
-    GEMINI_AVAILABLE = True
-except ImportError:
-    GEMINI_AVAILABLE = False
 
 def get_inference_client(model_id: str, provider: str = "auto"):
     """
@@ -22,119 +13,7 @@ def get_inference_client(model_id: str, provider: str = "auto"):
     
     Returns OpenAI-compatible client for all models or raises error if not configured.
     """
-    if model_id == "gemini-3.0-pro":
-        # Use Poe (OpenAI-compatible) client for Gemini 3.0 Pro
-        return OpenAI(
-            api_key=os.getenv("POE_API_KEY"),
-            base_url="https://api.poe.com/v1"
-        )
-    
-    elif model_id == "qwen3-30b-a3b-instruct-2507":
-        # Use DashScope OpenAI client
-        return OpenAI(
-            api_key=os.getenv("DASHSCOPE_API_KEY"),
-            base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
-        )
-    
-    elif model_id == "qwen3-30b-a3b-thinking-2507":
-        # Use DashScope OpenAI client for Thinking model
-        return OpenAI(
-            api_key=os.getenv("DASHSCOPE_API_KEY"),
-            base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
-        )
-    
-    elif model_id == "qwen3-coder-30b-a3b-instruct":
-        # Use DashScope OpenAI client for Coder model
-        return OpenAI(
-            api_key=os.getenv("DASHSCOPE_API_KEY"),
-            base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
-        )
-    
-    elif model_id == "gpt-5.1":
-        # Use Poe (OpenAI-compatible) client for GPT-5.1 model
-        return OpenAI(
-            api_key=os.getenv("POE_API_KEY"),
-            base_url="https://api.poe.com/v1"
-        )
-    
-    elif model_id == "gpt-5.1-instant":
-        # Use Poe (OpenAI-compatible) client for GPT-5.1 Instant model
-        return OpenAI(
-            api_key=os.getenv("POE_API_KEY"),
-            base_url="https://api.poe.com/v1"
-        )
-    
-    elif model_id == "gpt-5.1-codex":
-        # Use Poe (OpenAI-compatible) client for GPT-5.1 Codex model
-        return OpenAI(
-            api_key=os.getenv("POE_API_KEY"),
-            base_url="https://api.poe.com/v1"
-        )
-    
-    elif model_id == "gpt-5.1-codex-mini":
-        # Use Poe (OpenAI-compatible) client for GPT-5.1 Codex Mini model
-        return OpenAI(
-            api_key=os.getenv("POE_API_KEY"),
-            base_url="https://api.poe.com/v1"
-        )
-    
-    elif model_id == "grok-4":
-        # Use Poe (OpenAI-compatible) client for Grok-4 model
-        return OpenAI(
-            api_key=os.getenv("POE_API_KEY"),
-            base_url="https://api.poe.com/v1"
-        )
-    
-    elif model_id == "Grok-Code-Fast-1":
-        # Use Poe (OpenAI-compatible) client for Grok-Code-Fast-1 model
-        return OpenAI(
-            api_key=os.getenv("POE_API_KEY"),
-            base_url="https://api.poe.com/v1"
-        )
-    
-    elif model_id == "claude-opus-4.1":
-        # Use Poe (OpenAI-compatible) client for Claude-Opus-4.1
-        return OpenAI(
-            api_key=os.getenv("POE_API_KEY"),
-            base_url="https://api.poe.com/v1"
-        )
-    
-    elif model_id == "claude-opus-4.5":
-        # Use Poe (OpenAI-compatible) client for Claude-Opus-4.5
-        return OpenAI(
-            api_key=os.getenv("POE_API_KEY"),
-            base_url="https://api.poe.com/v1"
-        )
-    
-    elif model_id == "claude-sonnet-4.5":
-        # Use Poe (OpenAI-compatible) client for Claude-Sonnet-4.5
-        return OpenAI(
-            api_key=os.getenv("POE_API_KEY"),
-            base_url="https://api.poe.com/v1"
-        )
-    
-    elif model_id == "claude-haiku-4.5":
-        # Use Poe (OpenAI-compatible) client for Claude-Haiku-4.5
-        return OpenAI(
-            api_key=os.getenv("POE_API_KEY"),
-            base_url="https://api.poe.com/v1"
-        )
-    
-    elif model_id == "x-ai/grok-4.1-fast":
-        # Use OpenRouter client for Grok 4.1 Fast model
-        return OpenAI(
-            api_key=os.getenv("OPENROUTER_API_KEY"),
-            base_url="https://openrouter.ai/api/v1",
-        )
-    
-    elif model_id.startswith("openrouter/"):
-        # OpenRouter models
-        return OpenAI(
-            api_key=os.getenv("OPENROUTER_API_KEY"),
-            base_url="https://openrouter.ai/api/v1",
-        )
-    
-    elif model_id == "MiniMaxAI/MiniMax-M2":
+    if model_id == "MiniMaxAI/MiniMax-M2":
         # Use HuggingFace Router with Novita provider for MiniMax M2 model
         return OpenAI(
             base_url="https://router.huggingface.co/v1",
@@ -142,56 +21,6 @@ def get_inference_client(model_id: str, provider: str = "auto"):
             default_headers={"X-HF-Bill-To": "huggingface"}
         )
     
-    elif model_id == "step-3":
-        # Use StepFun API client for Step-3 model
-        return OpenAI(
-            api_key=os.getenv("STEP_API_KEY"),
-            base_url="https://api.stepfun.com/v1"
-        )
-    
-    elif model_id == "devstral-medium-2512":
-        # Use OpenRouter for Devstral
-        return OpenAI(
-            base_url="https://openrouter.ai/api/v1",
-            api_key=os.getenv("OPENROUTER_API_KEY"),
-            default_headers={
-                "HTTP-Referer": "https://anycoder.app",
-                "X-Title": "AnyCoder",
-            }
-        )
-    
-    elif model_id == "codestral-2508" or model_id == "mistral-medium-2508":
-        # Use Mistral client for Mistral models
-        return Mistral(api_key=os.getenv("MISTRAL_API_KEY"))
-    
-    elif model_id == "gemini-2.5-flash":
-        # Use Google Gemini (OpenAI-compatible) client
-        return OpenAI(
-            api_key=os.getenv("GEMINI_API_KEY"),
-            base_url="https://generativelanguage.googleapis.com/v1beta/openai/",
-        )
-    
-    elif model_id == "gemini-2.5-pro":
-        # Use Google Gemini Pro (OpenAI-compatible) client
-        return OpenAI(
-            api_key=os.getenv("GEMINI_API_KEY"),
-            base_url="https://generativelanguage.googleapis.com/v1beta/openai/",
-        )
-    
-    elif model_id == "gemini-flash-lite-latest":
-        # Use Google Gemini Flash Lite Latest (OpenAI-compatible) client
-        return OpenAI(
-            api_key=os.getenv("GEMINI_API_KEY"),
-            base_url="https://generativelanguage.googleapis.com/v1beta/openai/",
-        )
-    
-    elif model_id == "kimi-k2-turbo-preview":
-        # Use Moonshot AI (OpenAI-compatible) client for Kimi K2 Turbo (Preview)
-        return OpenAI(
-            api_key=os.getenv("MOONSHOT_API_KEY"),
-            base_url="https://api.moonshot.ai/v1",
-        )
-    
     elif model_id == "moonshotai/Kimi-K2-Thinking":
         # Use HuggingFace Router with Novita provider
         return OpenAI(
@@ -232,27 +61,6 @@ def get_inference_client(model_id: str, provider: str = "auto"):
             default_headers={"X-HF-Bill-To": "huggingface"}
         )
     
-    elif model_id == "stealth-model-1":
-        # Use stealth model with generic configuration
-        api_key = os.getenv("STEALTH_MODEL_1_API_KEY")
-        if not api_key:
-            raise ValueError("STEALTH_MODEL_1_API_KEY environment variable is required")
-        
-        base_url = os.getenv("STEALTH_MODEL_1_BASE_URL")
-        if not base_url:
-            raise ValueError("STEALTH_MODEL_1_BASE_URL environment variable is required")
-        
-        return OpenAI(
-            api_key=api_key,
-            base_url=base_url,
-        )
-    
-    elif model_id == "gemini-3-flash-preview":
-        # Use native Google GenAI client for Gemini 3.0 Flash Preview
-        if not GEMINI_AVAILABLE:
-            raise ImportError("google-genai package required for Gemini 3")
-        return genai.Client(api_key=os.getenv("GEMINI_API_KEY"))
-
     else:
         # Unknown model - try HuggingFace Inference API
         return OpenAI(
@@ -263,17 +71,7 @@ def get_inference_client(model_id: str, provider: str = "auto"):
 
 def get_real_model_id(model_id: str) -> str:
     """Get the real model ID with provider suffixes if needed"""
-    if model_id == "devstral-medium-2512":
-        return "mistralai/devstral-2512:free"
-    
-    elif model_id == "stealth-model-1":
-        # Get the real model ID from environment variable
-        real_model_id = os.getenv("STEALTH_MODEL_1_ID")
-        if not real_model_id:
-            raise ValueError("STEALTH_MODEL_1_ID environment variable is required")
-        return real_model_id
-    
-    elif model_id == "zai-org/GLM-4.6":
+    if model_id == "zai-org/GLM-4.6":
         # GLM-4.6 requires Cerebras provider suffix in model string for API calls
         return "zai-org/GLM-4.6:cerebras"
     
@@ -300,52 +98,12 @@ def get_real_model_id(model_id: str) -> str:
     return model_id
 
 
-def create_gemini3_messages(messages: list) -> tuple:
-    """
-    Convert OpenAI-style messages to Gemini 3 format.
-    Returns (contents, tools, config)
-    """
-    if not GEMINI_AVAILABLE:
-        raise ImportError("google-genai package required for Gemini 3")
-    
-    contents = []
-    system_prompt = None
-    
-    for msg in messages:
-        if msg['role'] == 'system':
-            system_prompt = msg['content']
-        elif msg['role'] in ['user', 'assistant']:
-            contents.append(
-                types.Content(
-                    role="user" if msg['role'] == 'user' else "model",
-                    parts=[types.Part.from_text(text=msg['content'])]
-                )
-            )
-    
-    # Add system prompt as first user message if exists
-    if system_prompt:
-        contents.insert(0, types.Content(
-            role="user",
-            parts=[types.Part.from_text(text=f"System instructions: {system_prompt}")]
-        ))
-    
-    # Configure tools and thinking
-    tools = [types.Tool(googleSearch=types.GoogleSearch())]
-    config = types.GenerateContentConfig(
-        thinking_config=types.ThinkingConfig(thinking_level="HIGH"),
-        tools=tools,
-        max_output_tokens=16384
-    )
-    
-    return contents, config
-
-
 def is_native_sdk_model(model_id: str) -> bool:
     """Check if model uses native SDK (not OpenAI-compatible)"""
-    return model_id == "gemini-3-flash-preview"
+    return False
 
 
 def is_mistral_model(model_id: str) -> bool:
     """Check if model uses Mistral SDK"""
-    return model_id in ["codestral-2508", "mistral-medium-2508"]
+    return False
 

frontend/src/components/LandingPage.tsx

@@ -513,8 +513,8 @@ ${isGradio ? '\n\nIMPORTANT: Only output app.py with the redesigned UI (themes,
               if (onStart) {
                 // Pass duplicated space ID so auto-deploy updates it
                 console.log('[Redesign] Calling onStart with duplicated repo ID:', duplicatedRepoId);
-                console.log('[Redesign] Using Claude-Sonnet-4.5 for redesign');
-                onStart(redesignPrompt, result.language || 'html', 'claude-sonnet-4.5', undefined, duplicatedRepoId);
+                console.log('[Redesign] Using GLM-4.6 for redesign');
+                onStart(redesignPrompt, result.language || 'html', 'zai-org/GLM-4.6', undefined, duplicatedRepoId);
               }
             }, 100);
 
@@ -558,8 +558,8 @@ Note: After generating the redesign, I will create a Pull Request on the origina
 
             if (onStart) {
               console.log('[Redesign] Will create PR - not passing repo ID');
-              console.log('[Redesign] Using Claude-Sonnet-4.5 for redesign');
-              onStart(redesignPrompt, result.language || 'html', 'claude-sonnet-4.5', undefined, repoId, true); // Pass true for shouldCreatePR
+              console.log('[Redesign] Using GLM-4.6 for redesign');
+              onStart(redesignPrompt, result.language || 'html', 'zai-org/GLM-4.6', undefined, repoId, true); // Pass true for shouldCreatePR
             }
 
             console.log('[Redesign] Will create PR after code generation completes');