Old Title

import os import re from http import HTTPStatus from typing import Dict, List, Optional, Tuple import base64 import mimetypes import PyPDF2 import docx import cv2 import numpy as np from PIL import Image import pytesseract import requests from urllib.parse import urlparse, urljoin from bs4 import BeautifulSoup import html2text import json import time import webbrowser import urllib.parse import copy import html import gradio as gr from huggingface_hub import InferenceClient from tavily import TavilyClient from huggingface_hub import HfApi import tempfile from openai import OpenAI from mistralai import Mistral import uuid import threading # Gradio supported languages for syntax highlighting GRADIO_SUPPORTED_LANGUAGES = [ "python", "c", "cpp", "markdown", "latex", "json", "html", "css", "javascript", "jinja2", "typescript", "yaml", "dockerfile", "shell", "r", "sql", "sql-msSQL", "sql-mySQL", "sql-mariaDB", "sql-sqlite", "sql-cassandra", "sql-plSQL", "sql-hive", "sql-pgSQL", "sql-gql", "sql-gpSQL", "sql-sparkSQL", "sql-esper", None ] def get_gradio_language(language): # Map composite options to a supported syntax highlighting if language == "streamlit": return "python" if language == "gradio": return "python" return language if language in GRADIO_SUPPORTED_LANGUAGES else None # Search/Replace Constants SEARCH_START = "<<<<<<< SEARCH" DIVIDER = "=======" REPLACE_END = ">>>>>>> REPLACE" # Configuration HTML_SYSTEM_PROMPT = """ONLY USE HTML, CSS AND JAVASCRIPT. If you want to use ICON make sure to import the library first. Try to create the best UI possible by using only HTML, CSS and JAVASCRIPT. MAKE IT RESPONSIVE USING MODERN CSS. Use as much as you can modern CSS for the styling, if you can't do something with modern CSS, then use custom CSS. Also, try to elaborate as much as you can, to create something unique. ALWAYS GIVE THE RESPONSE INTO A SINGLE HTML FILE For website redesign tasks: - Use the provided original HTML code as the starting point for redesign - Preserve all original content, structure, and functionality - Keep the same semantic HTML structure but enhance the styling - Reuse all original images and their URLs from the HTML code - Create a modern, responsive design with improved typography and spacing - Use modern CSS frameworks and design patterns - Ensure accessibility and mobile responsiveness - Maintain the same navigation and user flow - Enhance the visual design while keeping the original layout structure If an image is provided, analyze it and use the visual information to better understand the user's requirements. Always respond with code that can be executed or rendered directly. Always output only the HTML code inside a ```html ... ``` code block, and do not include any explanations or extra text. Do NOT add the language name at the top of the code output.""" # Stricter prompt for GLM-4.5V to ensure a complete, runnable HTML document with no escaped characters GLM45V_HTML_SYSTEM_PROMPT = """You are an expert front-end developer. Output a COMPLETE, STANDALONE HTML document that renders directly in a browser. Hard constraints: - DO NOT use React, ReactDOM, JSX, Babel, Vue, Angular, Svelte, or any SPA framework. - Use ONLY plain HTML, CSS, and vanilla JavaScript. - Allowed external resources: Tailwind CSS CDN, Font Awesome CDN, Google Fonts. - Do NOT escape characters (no \\n, \\t, or escaped quotes). Output raw HTML/JS/CSS. Structural requirements: - Include , , , and with proper nesting - Include required tags for any CSS you reference (e.g., Tailwind, Font Awesome, Google Fonts) - Keep everything in ONE file; inline CSS/JS as needed Return ONLY the code inside a single ```html ... ``` code block. No additional text before or after. """ # --------------------------------------------------------------------------- # Video temp-file management (per-session tracking and cleanup) # --------------------------------------------------------------------------- VIDEO_TEMP_DIR = os.path.join(tempfile.gettempdir(), "anycoder_videos") VIDEO_FILE_TTL_SECONDS = 6 * 60 * 60 # 6 hours _SESSION_VIDEO_FILES: Dict[str, List[str]] = {} _VIDEO_FILES_LOCK = threading.Lock() def _ensure_video_dir_exists() -> None: try: os.makedirs(VIDEO_TEMP_DIR, exist_ok=True) except Exception: pass def _register_video_for_session(session_id: Optional[str], file_path: str) -> None: if not session_id or not file_path: return with _VIDEO_FILES_LOCK: if session_id not in _SESSION_VIDEO_FILES: _SESSION_VIDEO_FILES[session_id] = [] _SESSION_VIDEO_FILES[session_id].append(file_path) def cleanup_session_videos(session_id: Optional[str]) -> None: if not session_id: return with _VIDEO_FILES_LOCK: file_list = _SESSION_VIDEO_FILES.pop(session_id, []) for path in file_list: try: if path and os.path.exists(path): os.unlink(path) except Exception: # Best-effort cleanup pass def reap_old_videos(ttl_seconds: int = VIDEO_FILE_TTL_SECONDS) -> None: """Delete old video files in the temp directory based on modification time.""" try: _ensure_video_dir_exists() now_ts = time.time() for name in os.listdir(VIDEO_TEMP_DIR): path = os.path.join(VIDEO_TEMP_DIR, name) try: if not os.path.isfile(path): continue mtime = os.path.getmtime(path) if now_ts - mtime > ttl_seconds: os.unlink(path) except Exception: pass except Exception: # Temp dir might not exist or be accessible; ignore pass # --------------------------------------------------------------------------- # Audio temp-file management (per-session tracking and cleanup) # --------------------------------------------------------------------------- AUDIO_TEMP_DIR = os.path.join(tempfile.gettempdir(), "anycoder_audio") AUDIO_FILE_TTL_SECONDS = 6 * 60 * 60 # 6 hours _SESSION_AUDIO_FILES: Dict[str, List[str]] = {} _AUDIO_FILES_LOCK = threading.Lock() def _ensure_audio_dir_exists() -> None: try: os.makedirs(AUDIO_TEMP_DIR, exist_ok=True) except Exception: pass def _register_audio_for_session(session_id: Optional[str], file_path: str) -> None: if not session_id or not file_path: return with _AUDIO_FILES_LOCK: if session_id not in _SESSION_AUDIO_FILES: _SESSION_AUDIO_FILES[session_id] = [] _SESSION_AUDIO_FILES[session_id].append(file_path) def cleanup_session_audio(session_id: Optional[str]) -> None: if not session_id: return with _AUDIO_FILES_LOCK: file_list = _SESSION_AUDIO_FILES.pop(session_id, []) for path in file_list: try: if path and os.path.exists(path): os.unlink(path) except Exception: pass def reap_old_audio(ttl_seconds: int = AUDIO_FILE_TTL_SECONDS) -> None: try: _ensure_audio_dir_exists() now_ts = time.time() for name in os.listdir(AUDIO_TEMP_DIR): path = os.path.join(AUDIO_TEMP_DIR, name) try: if not os.path.isfile(path): continue mtime = os.path.getmtime(path) if now_ts - mtime > ttl_seconds: os.unlink(path) except Exception: pass except Exception: pass TRANSFORMERS_JS_SYSTEM_PROMPT = """You are an expert web developer creating a transformers.js application. You will generate THREE separate files: index.html, index.js, and style.css. IMPORTANT: You MUST output ALL THREE files in the following format: ```html ``` ```javascript // index.js content here ``` ```css /* style.css content here */ ``` Requirements: 1. Create a modern, responsive web application using transformers.js 2. Use the transformers.js library for AI/ML functionality 3. Create a clean, professional UI with good user experience 4. Make the application fully responsive for mobile devices 5. Use modern CSS practices and JavaScript ES6+ features 6. Include proper error handling and loading states 7. Follow accessibility best practices Library import (required): Add the following snippet to index.html to import transformers.js: The index.html should contain the basic HTML structure and link to the CSS and JS files. The index.js should contain all the JavaScript logic including transformers.js integration. The style.css should contain all the styling for the application. Always output only the three code blocks as shown above, and do not include any explanations or extra text.""" SVELTE_SYSTEM_PROMPT = """You are an expert Svelte developer creating a modern Svelte application. You will generate ONLY the custom files that need user-specific content for the user's requested application. IMPORTANT: You MUST output files in the following format. Generate ONLY the files needed for the user's specific request: ```svelte ``` ```css /* src/app.css content here */ ``` If you need additional components for the user's specific app, add them like: ```svelte ``` Requirements: 1. Create a modern, responsive Svelte application based on the user's specific request 2. Use TypeScript for better type safety 3. Create a clean, professional UI with good user experience 4. Make the application fully responsive for mobile devices 5. Use modern CSS practices and Svelte best practices 6. Include proper error handling and loading states 7. Follow accessibility best practices 8. Use Svelte's reactive features effectively 9. Include proper component structure and organization 10. Generate ONLY components that are actually needed for the user's requested application Files you should generate: - src/App.svelte: Main application component (ALWAYS required) - src/app.css: Global styles (ALWAYS required) - src/lib/[ComponentName].svelte: Additional components (ONLY if needed for the user's specific app) The other files (index.html, package.json, vite.config.ts, tsconfig files, svelte.config.js, src/main.ts, src/vite-env.d.ts) are provided by the Svelte template and don't need to be generated. Always output only the two code blocks as shown above, and do not include any explanations or extra text.""" SVELTE_SYSTEM_PROMPT_WITH_SEARCH = """You are an expert Svelte developer creating a modern Svelte application. You have access to real-time web search. When needed, use web search to find the latest information, best practices, or specific Svelte technologies. You will generate ONLY the custom files that need user-specific content. IMPORTANT: You MUST output ONLY the custom files in the following format: ```svelte ``` ```css /* src/app.css content here --> ``` Requirements: 1. Create a modern, responsive Svelte application 2. Use TypeScript for better type safety 3. Create a clean, professional UI with good user experience 4. Make the application fully responsive for mobile devices 5. Use modern CSS practices and Svelte best practices 6. Include proper error handling and loading states 7. Follow accessibility best practices 8. Use Svelte's reactive features effectively 9. Include proper component structure and organization 10. Use web search to find the latest Svelte patterns, libraries, and best practices The files you generate are: - src/App.svelte: Main application component (your custom app logic) - src/app.css: Global styles (your custom styling) The other files (index.html, package.json, vite.config.ts, tsconfig files, svelte.config.js, src/main.ts, src/vite-env.d.ts) are provided by the Svelte template and don't need to be generated. Always output only the two code blocks as shown above, and do not include any explanations or extra text.""" TRANSFORMERS_JS_SYSTEM_PROMPT_WITH_SEARCH = """You are an expert web developer creating a transformers.js application. You have access to real-time web search. When needed, use web search to find the latest information, best practices, or specific technologies for transformers.js. You will generate THREE separate files: index.html, index.js, and style.css. IMPORTANT: You MUST output ALL THREE files in the following format: ```html ``` ```javascript // index.js content here ``` ```css /* style.css content here */ ``` Requirements: 1. Create a modern, responsive web application using transformers.js 2. Use the transformers.js library for AI/ML functionality 3. Use web search to find current best practices and latest transformers.js features 4. Create a clean, professional UI with good user experience 5. Make the application fully responsive for mobile devices 6. Use modern CSS practices and JavaScript ES6+ features 7. Include proper error handling and loading states 8. Follow accessibility best practices Library import (required): Add the following snippet to index.html to import transformers.js: The index.html should contain the basic HTML structure and link to the CSS and JS files. The index.js should contain all the JavaScript logic including transformers.js integration. The style.css should contain all the styling for the application. Always output only the three code blocks as shown above, and do not include any explanations or extra text.""" GENERIC_SYSTEM_PROMPT = """You are an expert {language} developer. Write clean, idiomatic, and runnable {language} code for the user's request. If possible, include comments and best practices. Output ONLY the code inside a ``` code block, and do not include any explanations or extra text. If the user provides a file or other context, use it as a reference. If the code is for a script or app, make it as self-contained as possible. Do NOT add the language name at the top of the code output.""" # System prompt with search capability HTML_SYSTEM_PROMPT_WITH_SEARCH = """You are an expert front-end developer. You have access to real-time web search. Output a COMPLETE, STANDALONE HTML document that renders directly in a browser. Requirements: - Include , , , and with proper nesting - Include all required and {REPLACE_END} ``` Example Fixing Dependencies (requirements.txt): ``` Adding missing dependency to fix ImportError... === requirements.txt === {SEARCH_START} gradio streamlit {DIVIDER} gradio streamlit mistral-common {REPLACE_END} ``` Example Deleting Code: ``` Removing the paragraph... {SEARCH_START}

This paragraph will be deleted.

{DIVIDER} {REPLACE_END} ```""" # Follow-up system prompt for modifying existing transformers.js applications TransformersJSFollowUpSystemPrompt = f"""You are an expert web developer modifying an existing transformers.js application. The user wants to apply changes based on their request. You MUST output ONLY the changes required using the following SEARCH/REPLACE block format. Do NOT output the entire file. Explain the changes briefly *before* the blocks if necessary, but the code changes THEMSELVES MUST be within the blocks. IMPORTANT: When the user reports an ERROR MESSAGE, analyze it carefully to determine which file needs fixing: - JavaScript errors/module loading issues → Fix index.js - HTML rendering/DOM issues → Fix index.html - Styling/visual issues → Fix style.css - CDN/library loading errors → Fix script tags in index.html The transformers.js application consists of three files: index.html, index.js, and style.css. When making changes, specify which file you're modifying by starting your search/replace blocks with the file name. Format Rules: 1. Start with {SEARCH_START} 2. Provide the exact lines from the current code that need to be replaced. 3. Use {DIVIDER} to separate the search block from the replacement. 4. Provide the new lines that should replace the original lines. 5. End with {REPLACE_END} 6. You can use multiple SEARCH/REPLACE blocks if changes are needed in different parts of the file. 7. To insert code, use an empty SEARCH block (only {SEARCH_START} and {DIVIDER} on their lines) if inserting at the very beginning, otherwise provide the line *before* the insertion point in the SEARCH block and include that line plus the new lines in the REPLACE block. 8. To delete code, provide the lines to delete in the SEARCH block and leave the REPLACE block empty (only {DIVIDER} and {REPLACE_END} on their lines). 9. IMPORTANT: The SEARCH block must *exactly* match the current code, including indentation and whitespace. Example Modifying HTML: ``` Changing the title in index.html... === index.html === {SEARCH_START} Old Title {DIVIDER} New Title {REPLACE_END} ``` Example Modifying JavaScript: ``` Adding a new function to index.js... === index.js === {SEARCH_START} // Existing code {DIVIDER} // Existing code function newFunction() {{ console.log("New function added"); }} {REPLACE_END} ``` Example Modifying CSS: ``` Changing background color in style.css... === style.css === {SEARCH_START} body {{ background-color: white; }} {DIVIDER} body {{ background-color: #f0f0f0; }} {REPLACE_END} ``` Example Fixing Library Loading Error: ``` Fixing transformers.js CDN loading error... === index.html === {SEARCH_START} {DIVIDER} {REPLACE_END} ```""" # Available models AVAILABLE_MODELS = [ { "name": "Moonshot Kimi-K2", "id": "moonshotai/Kimi-K2-Instruct", "description": "Moonshot AI Kimi-K2-Instruct model for code generation and general tasks" }, { "name": "Kimi K2 Turbo (Preview)", "id": "kimi-k2-turbo-preview", "description": "Moonshot AI Kimi K2 Turbo via OpenAI-compatible API" }, { "name": "DeepSeek V3", "id": "deepseek-ai/DeepSeek-V3-0324", "description": "DeepSeek V3 model for code generation" }, { "name": "DeepSeek R1", "id": "deepseek-ai/DeepSeek-R1-0528", "description": "DeepSeek R1 model for code generation" }, { "name": "ERNIE-4.5-VL", "id": "baidu/ERNIE-4.5-VL-424B-A47B-Base-PT", "description": "ERNIE-4.5-VL model for multimodal code generation with image support" }, { "name": "MiniMax M1", "id": "MiniMaxAI/MiniMax-M1-80k", "description": "MiniMax M1 model for code generation and general tasks" }, { "name": "Qwen3-235B-A22B", "id": "Qwen/Qwen3-235B-A22B", "description": "Qwen3-235B-A22B model for code generation and general tasks" }, { "name": "SmolLM3-3B", "id": "HuggingFaceTB/SmolLM3-3B", "description": "SmolLM3-3B model for code generation and general tasks" }, { "name": "GLM-4.5", "id": "zai-org/GLM-4.5", "description": "GLM-4.5 model with thinking capabilities for advanced code generation" }, { "name": "GLM-4.5V", "id": "zai-org/GLM-4.5V", "description": "GLM-4.5V multimodal model with image understanding for code generation" }, { "name": "GLM-4.1V-9B-Thinking", "id": "THUDM/GLM-4.1V-9B-Thinking", "description": "GLM-4.1V-9B-Thinking model for multimodal code generation with image support" }, { "name": "Qwen3-235B-A22B-Instruct-2507", "id": "Qwen/Qwen3-235B-A22B-Instruct-2507", "description": "Qwen3-235B-A22B-Instruct-2507 model for code generation and general tasks" }, { "name": "Qwen3-Coder-480B-A35B-Instruct", "id": "Qwen/Qwen3-Coder-480B-A35B-Instruct", "description": "Qwen3-Coder-480B-A35B-Instruct model for advanced code generation and programming tasks" }, { "name": "Qwen3-32B", "id": "Qwen/Qwen3-32B", "description": "Qwen3-32B model for code generation and general tasks" }, { "name": "Qwen3-4B-Instruct-2507", "id": "Qwen/Qwen3-4B-Instruct-2507", "description": "Qwen3-4B-Instruct-2507 model for code generation and general tasks" }, { "name": "Qwen3-4B-Thinking-2507", "id": "Qwen/Qwen3-4B-Thinking-2507", "description": "Qwen3-4B-Thinking-2507 model with advanced reasoning capabilities for code generation and general tasks" }, { "name": "Qwen3-235B-A22B-Thinking", "id": "Qwen/Qwen3-235B-A22B-Thinking-2507", "description": "Qwen3-235B-A22B-Thinking model with advanced reasoning capabilities" }, { "name": "Qwen3-30B-A3B-Instruct-2507", "id": "qwen3-30b-a3b-instruct-2507", "description": "Qwen3-30B-A3B-Instruct model via Alibaba Cloud DashScope API" }, { "name": "Qwen3-30B-A3B-Thinking-2507", "id": "qwen3-30b-a3b-thinking-2507", "description": "Qwen3-30B-A3B-Thinking model with advanced reasoning via Alibaba Cloud DashScope API" }, { "name": "Qwen3-Coder-30B-A3B-Instruct", "id": "qwen3-coder-30b-a3b-instruct", "description": "Qwen3-Coder-30B-A3B-Instruct model for advanced code generation via Alibaba Cloud DashScope API" }, { "name": "StepFun Step-3", "id": "step-3", "description": "StepFun Step-3 model - AI chat assistant by 阶跃星辰 with multilingual capabilities" }, { "name": "Codestral 2508", "id": "codestral-2508", "description": "Mistral Codestral model - specialized for code generation and programming tasks" }, { "name": "Mistral Medium 2508", "id": "mistral-medium-2508", "description": "Mistral Medium 2508 model via Mistral API for general tasks and coding" }, { "name": "Gemini 2.5 Flash", "id": "gemini-2.5-flash", "description": "Google Gemini 2.5 Flash via OpenAI-compatible API" }, { "name": "Gemini 2.5 Pro", "id": "gemini-2.5-pro", "description": "Google Gemini 2.5 Pro via OpenAI-compatible API" }, { "name": "GPT-OSS-120B", "id": "openai/gpt-oss-120b", "description": "OpenAI GPT-OSS-120B model for advanced code generation and general tasks" }, { "name": "GPT-OSS-20B", "id": "openai/gpt-oss-20b", "description": "OpenAI GPT-OSS-20B model for code generation and general tasks" }, { "name": "GPT-5", "id": "gpt-5", "description": "OpenAI GPT-5 model for advanced code generation and general tasks" }, { "name": "Grok-4", "id": "grok-4", "description": "Grok-4 model via Poe (OpenAI-compatible) for advanced tasks" }, { "name": "Claude-Opus-4.1", "id": "claude-opus-4.1", "description": "Anthropic Claude Opus 4.1 via Poe (OpenAI-compatible)" } ] # Default model selection DEFAULT_MODEL_NAME = "Qwen3-Coder-480B-A35B-Instruct" DEFAULT_MODEL = None for _m in AVAILABLE_MODELS: if _m.get("name") == DEFAULT_MODEL_NAME: DEFAULT_MODEL = _m break if DEFAULT_MODEL is None and AVAILABLE_MODELS: DEFAULT_MODEL = AVAILABLE_MODELS[0] DEMO_LIST = [ { "title": "Todo App", "description": "Create a simple todo application with add, delete, and mark as complete functionality" }, { "title": "Calculator", "description": "Build a basic calculator with addition, subtraction, multiplication, and division" }, { "title": "Chat Interface", "description": "Build a chat interface with message history and user input" }, { "title": "E-commerce Product Card", "description": "Create a product card component for an e-commerce website" }, { "title": "Login Form", "description": "Build a responsive login form with validation" }, { "title": "Dashboard Layout", "description": "Create a dashboard layout with sidebar navigation and main content area" }, { "title": "Data Table", "description": "Build a data table with sorting and filtering capabilities" }, { "title": "Image Gallery", "description": "Create an image gallery with lightbox functionality and responsive grid layout" }, { "title": "UI from Image", "description": "Upload an image of a UI design and I'll generate the HTML/CSS code for it" }, { "title": "Extract Text from Image", "description": "Upload an image containing text and I'll extract and process the text content" }, { "title": "Website Redesign", "description": "Enter a website URL to extract its content and redesign it with a modern, responsive layout" }, { "title": "Modify HTML", "description": "After generating HTML, ask me to modify it with specific changes using search/replace format" }, { "title": "Search/Replace Example", "description": "Generate HTML first, then ask: 'Change the title to My New Title' or 'Add a blue background to the body'" }, { "title": "Transformers.js App", "description": "Create a transformers.js application with AI/ML functionality using the transformers.js library" }, { "title": "Svelte App", "description": "Create a modern Svelte application with TypeScript, Vite, and responsive design" } ] # HF Inference Client HF_TOKEN = os.getenv('HF_TOKEN') if not HF_TOKEN: raise RuntimeError("HF_TOKEN environment variable is not set. Please set it to your Hugging Face API token.") def get_inference_client(model_id, provider="auto"): """Return an InferenceClient with provider based on model_id and user selection.""" if 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": # Use Poe (OpenAI-compatible) client for GPT-5 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 == "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 == "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 == "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 == "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 == "openai/gpt-oss-120b": provider = "groq" elif model_id == "openai/gpt-oss-20b": provider = "groq" elif model_id == "moonshotai/Kimi-K2-Instruct": provider = "groq" elif model_id == "Qwen/Qwen3-235B-A22B": provider = "cerebras" elif model_id == "Qwen/Qwen3-235B-A22B-Instruct-2507": provider = "cerebras" elif model_id == "Qwen/Qwen3-32B": provider = "cerebras" elif model_id == "Qwen/Qwen3-235B-A22B-Thinking-2507": provider = "cerebras" elif model_id == "Qwen/Qwen3-Coder-480B-A35B-Instruct": provider = "cerebras" return InferenceClient( provider=provider, api_key=HF_TOKEN, bill_to="huggingface" ) # Type definitions History = List[Tuple[str, str]] Messages = List[Dict[str, str]] # Tavily Search Client TAVILY_API_KEY = os.getenv('TAVILY_API_KEY') tavily_client = None if TAVILY_API_KEY: try: tavily_client = TavilyClient(api_key=TAVILY_API_KEY) except Exception as e: print(f"Failed to initialize Tavily client: {e}") tavily_client = None def history_to_messages(history: History, system: str) -> Messages: messages = [{'role': 'system', 'content': system}] for h in history: # Handle multimodal content in history user_content = h[0] if isinstance(user_content, list): # Extract text from multimodal content text_content = "" for item in user_content: if isinstance(item, dict) and item.get("type") == "text": text_content += item.get("text", "") user_content = text_content if text_content else str(user_content) messages.append({'role': 'user', 'content': user_content}) messages.append({'role': 'assistant', 'content': h[1]}) return messages def messages_to_history(messages: Messages) -> Tuple[str, History]: assert messages[0]['role'] == 'system' history = [] for q, r in zip(messages[1::2], messages[2::2]): # Extract text content from multimodal messages for history user_content = q['content'] if isinstance(user_content, list): text_content = "" for item in user_content: if isinstance(item, dict) and item.get("type") == "text": text_content += item.get("text", "") user_content = text_content if text_content else str(user_content) history.append([user_content, r['content']]) return history def history_to_chatbot_messages(history: History) -> List[Dict[str, str]]: """Convert history tuples to chatbot message format""" messages = [] for user_msg, assistant_msg in history: # Handle multimodal content if isinstance(user_msg, list): text_content = "" for item in user_msg: if isinstance(item, dict) and item.get("type") == "text": text_content += item.get("text", "") user_msg = text_content if text_content else str(user_msg) messages.append({"role": "user", "content": user_msg}) messages.append({"role": "assistant", "content": assistant_msg}) return messages def remove_code_block(text): # Try to match code blocks with language markers patterns = [ r'```(?:html|HTML)\n([\s\S]+?)\n```', # Match ```html or ```HTML r'```\n([\s\S]+?)\n```', # Match code blocks without language markers r'```([\s\S]+?)```' # Match code blocks without line breaks ] for pattern in patterns: match = re.search(pattern, text, re.DOTALL) if match: extracted = match.group(1).strip() # Remove a leading language marker line (e.g., 'python') if present if extracted.split('\n', 1)[0].strip().lower() in ['python', 'html', 'css', 'javascript', 'json', 'c', 'cpp', 'markdown', 'latex', 'jinja2', 'typescript', 'yaml', 'dockerfile', 'shell', 'r', 'sql', 'sql-mssql', 'sql-mysql', 'sql-mariadb', 'sql-sqlite', 'sql-cassandra', 'sql-plSQL', 'sql-hive', 'sql-pgsql', 'sql-gql', 'sql-gpsql', 'sql-sparksql', 'sql-esper']: return extracted.split('\n', 1)[1] if '\n' in extracted else '' # If HTML markup starts later in the block (e.g., Poe injected preface), trim to first HTML root html_root_idx = None for tag in [' 0: return extracted[html_root_idx:].strip() return extracted # If no code block is found, check if the entire text is HTML stripped = text.strip() if stripped.startswith('') or stripped.startswith(' 0: return stripped[idx:].strip() return stripped # Special handling for python: remove python marker if text.strip().startswith('```python'): return text.strip()[9:-3].strip() # Remove a leading language marker line if present (fallback) lines = text.strip().split('\n', 1) if lines[0].strip().lower() in ['python', 'html', 'css', 'javascript', 'json', 'c', 'cpp', 'markdown', 'latex', 'jinja2', 'typescript', 'yaml', 'dockerfile', 'shell', 'r', 'sql', 'sql-mssql', 'sql-mysql', 'sql-mariadb', 'sql-sqlite', 'sql-cassandra', 'sql-plSQL', 'sql-hive', 'sql-pgsql', 'sql-gql', 'sql-gpsql', 'sql-sparksql', 'sql-esper']: return lines[1] if len(lines) > 1 else '' return text.strip() ## React CDN compatibility fixer removed per user preference def strip_placeholder_thinking(text: str) -> str: """Remove placeholder 'Thinking...' status lines from streamed text.""" if not text: return text # Matches lines like: "Thinking..." or "Thinking... (12s elapsed)" return re.sub(r"(?mi)^[\t ]*Thinking\.\.\.(?:\s*$\d+s elapsed$)?[\t ]*$\n?", "", text) def is_placeholder_thinking_only(text: str) -> bool: """Return True if text contains only 'Thinking...' placeholder lines (with optional elapsed).""" if not text: return False stripped = text.strip() if not stripped: return False return re.fullmatch(r"(?s)(?:\s*Thinking\.\.\.(?:\s*$\d+s elapsed$)?\s*)+", stripped) is not None def extract_last_thinking_line(text: str) -> str: """Extract the last 'Thinking...' line to display as status.""" matches = list(re.finditer(r"Thinking\.\.\.(?:\s*$\d+s elapsed$)?", text)) return matches[-1].group(0) if matches else "Thinking..." def parse_transformers_js_output(text): """Parse transformers.js output and extract the three files (index.html, index.js, style.css)""" files = { 'index.html': '', 'index.js': '', 'style.css': '' } # Multiple patterns to match the three code blocks with different variations html_patterns = [ r'```html\s*\n([\s\S]+?)\n```', r'```htm\s*\n([\s\S]+?)\n```', r'```\s*(?:index\.html|html)\s*\n([\s\S]+?)\n```' ] js_patterns = [ r'```javascript\s*\n([\s\S]+?)\n```', r'```js\s*\n([\s\S]+?)\n```', r'```\s*(?:index\.js|javascript)\s*\n([\s\S]+?)\n```' ] css_patterns = [ r'```css\s*\n([\s\S]+?)\n```', r'```\s*(?:style\.css|css)\s*\n([\s\S]+?)\n```' ] # Extract HTML content for pattern in html_patterns: html_match = re.search(pattern, text, re.IGNORECASE) if html_match: files['index.html'] = html_match.group(1).strip() break # Extract JavaScript content for pattern in js_patterns: js_match = re.search(pattern, text, re.IGNORECASE) if js_match: files['index.js'] = js_match.group(1).strip() break # Extract CSS content for pattern in css_patterns: css_match = re.search(pattern, text, re.IGNORECASE) if css_match: files['style.css'] = css_match.group(1).strip() break # Fallback: support === index.html === format if any file is missing if not (files['index.html'] and files['index.js'] and files['style.css']): # Use regex to extract sections html_fallback = re.search(r'===\s*index\.html\s*===\s*\n([\s\S]+?)(?=\n===|$)', text, re.IGNORECASE) js_fallback = re.search(r'===\s*index\.js\s*===\s*\n([\s\S]+?)(?=\n===|$)', text, re.IGNORECASE) css_fallback = re.search(r'===\s*style\.css\s*===\s*\n([\s\S]+?)(?=\n===|$)', text, re.IGNORECASE) if html_fallback: files['index.html'] = html_fallback.group(1).strip() if js_fallback: files['index.js'] = js_fallback.group(1).strip() if css_fallback: files['style.css'] = css_fallback.group(1).strip() # Additional fallback: extract from numbered sections or file headers if not (files['index.html'] and files['index.js'] and files['style.css']): # Try patterns like "1. index.html:" or "**index.html**" patterns = [ (r'(?:^\d+\.\s*|^##\s*|^\*\*\s*)index\.html(?:\s*:|\*\*:?)\s*\n([\s\S]+?)(?=\n(?:\d+\.|##|\*\*|===)|$)', 'index.html'), (r'(?:^\d+\.\s*|^##\s*|^\*\*\s*)index\.js(?:\s*:|\*\*:?)\s*\n([\s\S]+?)(?=\n(?:\d+\.|##|\*\*|===)|$)', 'index.js'), (r'(?:^\d+\.\s*|^##\s*|^\*\*\s*)style\.css(?:\s*:|\*\*:?)\s*\n([\s\S]+?)(?=\n(?:\d+\.|##|\*\*|===)|$)', 'style.css') ] for pattern, file_key in patterns: if not files[file_key]: match = re.search(pattern, text, re.IGNORECASE | re.MULTILINE) if match: # Clean up the content by removing any code block markers content = match.group(1).strip() content = re.sub(r'^```\w*\s*\n', '', content) content = re.sub(r'\n```\s*$', '', content) files[file_key] = content.strip() return files def format_transformers_js_output(files): """Format the three files into a single display string""" output = [] output.append("=== index.html ===") output.append(files['index.html']) output.append("\n=== index.js ===") output.append(files['index.js']) output.append("\n=== style.css ===") output.append(files['style.css']) return '\n'.join(output) def build_transformers_inline_html(files: dict) -> str: """Merge transformers.js three-file output into a single self-contained HTML document. - Inlines style.css into a " if css else "" if style_tag: if '' in doc.lower(): # Preserve original casing by finding closing head case-insensitively match = _re.search(r"", doc, flags=_re.IGNORECASE) if match: idx = match.start() doc = doc[:idx] + style_tag + doc[idx:] else: # No head; insert at top of body match = _re.search(r"]*>", doc, flags=_re.IGNORECASE) if match: idx = match.end() doc = doc[:idx] + "\n" + style_tag + doc[idx:] else: # Append at beginning doc = style_tag + doc # Inline JS: insert before script_tag = f"" if js else "" # Cleanup script to clear Cache Storage and IndexedDB on unload to free model weights cleanup_tag = ( "" ) if script_tag: match = _re.search(r"", doc, flags=_re.IGNORECASE) if match: idx = match.start() doc = doc[:idx] + script_tag + cleanup_tag + doc[idx:] else: # Append at end doc = doc + script_tag + cleanup_tag return doc def send_transformers_to_sandbox(files: dict) -> str: """Build a self-contained HTML document from transformers.js files and return an iframe preview.""" merged_html = build_transformers_inline_html(files) return send_to_sandbox(merged_html) def parse_svelte_output(text): """Parse Svelte output to extract individual files""" files = { 'src/App.svelte': '', 'src/app.css': '' } import re # First try to extract using code block patterns svelte_pattern = r'```svelte\s*\n([\s\S]+?)\n```' css_pattern = r'```css\s*\n([\s\S]+?)\n```' # Extract svelte block for App.svelte svelte_match = re.search(svelte_pattern, text, re.IGNORECASE) css_match = re.search(css_pattern, text, re.IGNORECASE) if svelte_match: files['src/App.svelte'] = svelte_match.group(1).strip() if css_match: files['src/app.css'] = css_match.group(1).strip() # Fallback: support === filename === format if any file is missing if not (files['src/App.svelte'] and files['src/app.css']): # Use regex to extract sections app_svelte_fallback = re.search(r'===\s*src/App\.svelte\s*===\n([\s\S]+?)(?=\n===|$)', text, re.IGNORECASE) app_css_fallback = re.search(r'===\s*src/app\.css\s*===\n([\s\S]+?)(?=\n===|$)', text, re.IGNORECASE) if app_svelte_fallback: files['src/App.svelte'] = app_svelte_fallback.group(1).strip() if app_css_fallback: files['src/app.css'] = app_css_fallback.group(1).strip() return files def format_svelte_output(files): """Format Svelte files into a single display string""" output = [] output.append("=== src/App.svelte ===") output.append(files['src/App.svelte']) output.append("\n=== src/app.css ===") output.append(files['src/app.css']) return '\n'.join(output) def history_render(history: History): return gr.update(visible=True), history def clear_history(): return [], [], None, "" # Empty lists for both tuple format and chatbot messages, None for file, empty string for website URL def update_image_input_visibility(model): """Update image input visibility based on selected model""" is_ernie_vl = model.get("id") == "baidu/ERNIE-4.5-VL-424B-A47B-Base-PT" is_glm_vl = model.get("id") == "THUDM/GLM-4.1V-9B-Thinking" is_glm_45v = model.get("id") == "zai-org/GLM-4.5V" return gr.update(visible=is_ernie_vl or is_glm_vl or is_glm_45v) def process_image_for_model(image): """Convert image to base64 for model input""" if image is None: return None # Convert numpy array to PIL Image if needed import io import base64 import numpy as np from PIL import Image # Handle numpy array from Gradio if isinstance(image, np.ndarray): image = Image.fromarray(image) buffer = io.BytesIO() image.save(buffer, format='PNG') img_str = base64.b64encode(buffer.getvalue()).decode() return f"data:image/png;base64,{img_str}" def generate_image_with_qwen(prompt: str, image_index: int = 0) -> str: """Generate image using Qwen image model via Hugging Face InferenceClient with optimized data URL""" try: # Check if HF_TOKEN is available if not os.getenv('HF_TOKEN'): return "Error: HF_TOKEN environment variable is not set. Please set it to your Hugging Face API token." # Create InferenceClient for Qwen image generation client = InferenceClient( provider="auto", api_key=os.getenv('HF_TOKEN'), bill_to="huggingface", ) # Generate image using Qwen/Qwen-Image model image = client.text_to_image( prompt, model="Qwen/Qwen-Image", ) # Resize image to reduce size while maintaining quality max_size = 512 if image.width > max_size or image.height > max_size: image.thumbnail((max_size, max_size), Image.Resampling.LANCZOS) # Convert PIL Image to optimized base64 for HTML embedding import io import base64 buffer = io.BytesIO() # Save as JPEG with compression for smaller file size image.convert('RGB').save(buffer, format='JPEG', quality=85, optimize=True) img_str = base64.b64encode(buffer.getvalue()).decode() # Return HTML img tag with optimized data URL return f' {prompt}

' except Exception as e: print(f"Image generation error: {str(e)}") return f"Error generating image: {str(e)}" def generate_image_to_image(input_image_data, prompt: str) -> str: """Generate an image using image-to-image with Qwen-Image-Edit via Hugging Face InferenceClient. Returns an HTML tag with optimized base64 JPEG data, similar to text-to-image output. """ try: # Check token if not os.getenv('HF_TOKEN'): return "Error: HF_TOKEN environment variable is not set. Please set it to your Hugging Face API token." # Prepare client client = InferenceClient( provider="auto", api_key=os.getenv('HF_TOKEN'), bill_to="huggingface", ) # Normalize input image to bytes import io from PIL import Image try: import numpy as np except Exception: np = None if hasattr(input_image_data, 'read'): # File-like object raw = input_image_data.read() pil_image = Image.open(io.BytesIO(raw)) elif hasattr(input_image_data, 'mode') and hasattr(input_image_data, 'size'): # PIL Image pil_image = input_image_data elif np is not None and isinstance(input_image_data, np.ndarray): pil_image = Image.fromarray(input_image_data) elif isinstance(input_image_data, (bytes, bytearray)): pil_image = Image.open(io.BytesIO(input_image_data)) else: # Fallback: try to convert via bytes pil_image = Image.open(io.BytesIO(bytes(input_image_data))) # Ensure RGB if pil_image.mode != 'RGB': pil_image = pil_image.convert('RGB') # Resize input image to avoid request body size limits max_input_size = 1024 if pil_image.width > max_input_size or pil_image.height > max_input_size: pil_image.thumbnail((max_input_size, max_input_size), Image.Resampling.LANCZOS) buf = io.BytesIO() pil_image.save(buf, format='JPEG', quality=85, optimize=True) input_bytes = buf.getvalue() # Call image-to-image image = client.image_to_image( input_bytes, prompt=prompt, model="Qwen/Qwen-Image-Edit", ) # Resize/optimize max_size = 512 if image.width > max_size or image.height > max_size: image.thumbnail((max_size, max_size), Image.Resampling.LANCZOS) out_buf = io.BytesIO() image.convert('RGB').save(out_buf, format='JPEG', quality=85, optimize=True) import base64 img_str = base64.b64encode(out_buf.getvalue()).decode() return f" $\"{prompt}\"$ " except Exception as e: print(f"Image-to-image generation error: {str(e)}") return f"Error generating image (image-to-image): {str(e)}" def generate_video_from_image(input_image_data, prompt: str, session_id: Optional[str] = None) -> str: """Generate a video from an input image and prompt using Hugging Face InferenceClient. Returns an HTML