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 # 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): 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.""" 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 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 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 = """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 You have access to real-time web search. When needed, use web search to find the latest information, best practices, or specific technologies. 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 - Use web search to find current design trends and best practices for the specific type of website - 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.""" GENERIC_SYSTEM_PROMPT_WITH_SEARCH = """You are an expert {language} developer. You have access to real-time web search. When needed, use web search to find the latest information, best practices, or specific technologies for {language}. 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.""" # Follow-up system prompt for modifying existing HTML files FollowUpSystemPrompt = f"""You are an expert web developer modifying an existing project. 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: - ImportError/ModuleNotFoundError → Fix requirements.txt by adding missing packages - Syntax errors in Python code → Fix app.py or the main Python file - HTML/CSS/JavaScript errors → Fix the respective HTML/CSS/JS files - Configuration errors → Fix config files, Docker files, etc. For Python applications (Gradio/Streamlit), the project structure typically includes: - app.py (main application file) - requirements.txt (dependencies) - Other supporting files as needed 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. 10. For multi-file projects, specify which file you're modifying by starting with the filename before the search/replace block. Example Modifying Code: ``` Some explanation... {SEARCH_START}

Old Title

{DIVIDER}

New Title

{REPLACE_END} {SEARCH_START} {DIVIDER} {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": "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-4.1V-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": "Qwen3-235B-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": "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" } ] 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 == "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": # Use Mistral client for Codestral model return Mistral(api_key=os.getenv("MISTRAL_API_KEY")) elif model_id == "openai/gpt-oss-120b": provider = "cerebras" 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 '' return extracted # If no code block is found, check if the entire text is HTML if text.strip().startswith('') or text.strip().startswith(' 1 else '' return text.strip() 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*$', '', content) files[file_key] = content.strip() return files def format_transformers_js_output(files): """Format the three files into a single display string""" output = [] output.append("=== index.html ===") output.append(files['index.html']) output.append("\n=== index.js ===") output.append(files['index.js']) output.append("\n=== style.css ===") output.append(files['style.css']) return '\n'.join(output) def parse_svelte_output(text): """Parse Svelte output to extract individual files""" files = { 'src/App.svelte': '', 'src/app.css': '' } import re # First try to extract using code block patterns svelte_pattern = r'```svelte\s*\n([\s\S]+?)\n```' css_pattern = r'```css\s*\n([\s\S]+?)\n```' # Extract svelte block for App.svelte svelte_match = re.search(svelte_pattern, text, re.IGNORECASE) css_match = re.search(css_pattern, text, re.IGNORECASE) if svelte_match: files['src/App.svelte'] = svelte_match.group(1).strip() if css_match: files['src/app.css'] = css_match.group(1).strip() # Fallback: support === filename === format if any file is missing if not (files['src/App.svelte'] and files['src/app.css']): # Use regex to extract sections app_svelte_fallback = re.search(r'===\s*src/App\.svelte\s*===\n([\s\S]+?)(?=\n===|$)', 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" return gr.update(visible=is_ernie_vl or is_glm_vl) 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 extract_image_prompts_from_text(text: str, num_images_needed: int = 1) -> list: """Extract image generation prompts from the full text based on number of images needed""" # Use the entire text as the base prompt for image generation # Clean up the text and create variations for the required number of images # Clean the text cleaned_text = text.strip() if not cleaned_text: return [] # Create variations of the prompt for the required number of images prompts = [] # Generate exactly the number of images needed for i in range(num_images_needed): if i == 0: # First image: Use the full prompt as-is prompts.append(cleaned_text) elif i == 1: # Second image: Add "visual representation" to make it more image-focused prompts.append(f"Visual representation of {cleaned_text}") elif i == 2: # Third image: Add "illustration" to create a different style prompts.append(f"Illustration of {cleaned_text}") else: # For additional images, use different variations variations = [ f"Digital art of {cleaned_text}", f"Modern design of {cleaned_text}", f"Professional illustration of {cleaned_text}", f"Clean design of {cleaned_text}", f"Beautiful visualization of {cleaned_text}", f"Stylish representation of {cleaned_text}", f"Contemporary design of {cleaned_text}", f"Elegant illustration of {cleaned_text}" ] variation_index = (i - 3) % len(variations) prompts.append(variations[variation_index]) return prompts def create_image_replacement_blocks(html_content: str, user_prompt: str) -> str: """Create search/replace blocks to replace placeholder images with generated Qwen images""" if not user_prompt: return "" # Find existing image placeholders in the HTML first import re # Common patterns for placeholder images placeholder_patterns = [ r']*src=["\'](?:placeholder|dummy|sample|example)[^"\']*["\'][^>]*>', r']*src=["\']https?://via\.placeholder\.com[^"\']*["\'][^>]*>', r']*src=["\']https?://picsum\.photos[^"\']*["\'][^>]*>', r']*src=["\']https?://dummyimage\.com[^"\']*["\'][^>]*>', r']*alt=["\'][^"\']*placeholder[^"\']*["\'][^>]*>', r']*class=["\'][^"\']*placeholder[^"\']*["\'][^>]*>', r']*id=["\'][^"\']*placeholder[^"\']*["\'][^>]*>', r']*src=["\']data:image[^"\']*["\'][^>]*>', # Base64 images r']*src=["\']#["\'][^>]*>', # Empty src r']*src=["\']about:blank["\'][^>]*>', # About blank ] # Find all placeholder images placeholder_images = [] for pattern in placeholder_patterns: matches = re.findall(pattern, html_content, re.IGNORECASE) placeholder_images.extend(matches) # If no placeholder images found, look for any img tags if not placeholder_images: img_pattern = r']*>' placeholder_images = re.findall(img_pattern, html_content) # Also look for div elements that might be image placeholders div_placeholder_patterns = [ r']*class=["\'][^"\']*(?:image|img|photo|picture)[^"\']*["\'][^>]*>.*?', r']*id=["\'][^"\']*(?:image|img|photo|picture)[^"\']*["\'][^>]*>.*?', ] for pattern in div_placeholder_patterns: matches = re.findall(pattern, html_content, re.IGNORECASE | re.DOTALL) placeholder_images.extend(matches) # Count how many images we need to generate num_images_needed = len(placeholder_images) if num_images_needed == 0: return "" # Generate image prompts based on the number of images found image_prompts = extract_image_prompts_from_text(user_prompt, num_images_needed) # Generate images for each prompt generated_images = [] for i, prompt in enumerate(image_prompts): image_html = generate_image_with_qwen(prompt, i) if not image_html.startswith("Error"): generated_images.append((i, image_html)) if not generated_images: return "" # Create search/replace blocks replacement_blocks = [] for i, (prompt_index, generated_image) in enumerate(generated_images): if i < len(placeholder_images): # Replace existing placeholder placeholder = placeholder_images[i] # Clean up the placeholder for better matching placeholder_clean = re.sub(r'\s+', ' ', placeholder.strip()) # Try multiple variations of the placeholder for better matching placeholder_variations = [ placeholder_clean, placeholder_clean.replace('"', "'"), placeholder_clean.replace("'", '"'), re.sub(r'\s+', ' ', placeholder_clean), placeholder_clean.replace(' ', ' '), ] # Create a replacement block for each variation for variation in placeholder_variations: replacement_blocks.append(f"""{SEARCH_START} {variation} {DIVIDER} {generated_image} {REPLACE_END}""") else: # Add new image if we have more generated images than placeholders # Find a good insertion point (after body tag or main content) if '', html_content.find(' str: """Apply search/replace changes to content (HTML, Python, etc.)""" if not changes_text.strip(): return original_content # Split the changes text into individual search/replace blocks blocks = [] current_block = "" lines = changes_text.split('\n') for line in lines: if line.strip() == SEARCH_START: if current_block.strip(): blocks.append(current_block.strip()) current_block = line + '\n' elif line.strip() == REPLACE_END: current_block += line + '\n' blocks.append(current_block.strip()) current_block = "" else: current_block += line + '\n' if current_block.strip(): blocks.append(current_block.strip()) modified_content = original_content for block in blocks: if not block.strip(): continue # Parse the search/replace block lines = block.split('\n') search_lines = [] replace_lines = [] in_search = False in_replace = False for line in lines: if line.strip() == SEARCH_START: in_search = True in_replace = False elif line.strip() == DIVIDER: in_search = False in_replace = True elif line.strip() == REPLACE_END: in_replace = False elif in_search: search_lines.append(line) elif in_replace: replace_lines.append(line) # Apply the search/replace if search_lines: search_text = '\n'.join(search_lines).strip() replace_text = '\n'.join(replace_lines).strip() if search_text in modified_content: modified_content = modified_content.replace(search_text, replace_text) else: print(f"Warning: Search text not found in content: {search_text[:100]}...") return modified_content def apply_transformers_js_search_replace_changes(original_formatted_content: str, changes_text: str) -> str: """Apply search/replace changes to transformers.js formatted content (three files)""" if not changes_text.strip(): return original_formatted_content # Parse the original formatted content to get the three files files = parse_transformers_js_output(original_formatted_content) # Split the changes text into individual search/replace blocks blocks = [] current_block = "" lines = changes_text.split('\n') for line in lines: if line.strip() == SEARCH_START: if current_block.strip(): blocks.append(current_block.strip()) current_block = line + '\n' elif line.strip() == REPLACE_END: current_block += line + '\n' blocks.append(current_block.strip()) current_block = "" else: current_block += line + '\n' if current_block.strip(): blocks.append(current_block.strip()) # Process each block and apply changes to the appropriate file for block in blocks: if not block.strip(): continue # Parse the search/replace block lines = block.split('\n') search_lines = [] replace_lines = [] in_search = False in_replace = False target_file = None for line in lines: if line.strip() == SEARCH_START: in_search = True in_replace = False elif line.strip() == DIVIDER: in_search = False in_replace = True elif line.strip() == REPLACE_END: in_replace = False elif in_search: search_lines.append(line) elif in_replace: replace_lines.append(line) # Determine which file this change targets based on the search content if search_lines: search_text = '\n'.join(search_lines).strip() replace_text = '\n'.join(replace_lines).strip() # Check which file contains the search text if search_text in files['index.html']: target_file = 'index.html' elif search_text in files['index.js']: target_file = 'index.js' elif search_text in files['style.css']: target_file = 'style.css' # Apply the change to the target file if target_file and search_text in files[target_file]: files[target_file] = files[target_file].replace(search_text, replace_text) else: print(f"Warning: Search text not found in any transformers.js file: {search_text[:100]}...") # Reformat the modified files return format_transformers_js_output(files) # Updated for faster Tavily search and closer prompt usage # Uses 'advanced' search_depth and auto_parameters=True for speed and relevance def perform_web_search(query: str, max_results: int = 5, include_domains=None, exclude_domains=None) -> str: """Perform web search using Tavily with default parameters""" if not tavily_client: return "Web search is not available. Please set the TAVILY_API_KEY environment variable." try: # Use Tavily defaults with advanced search depth for better results search_params = { "search_depth": "advanced", "max_results": min(max(1, max_results), 20) } if include_domains is not None: search_params["include_domains"] = include_domains if exclude_domains is not None: search_params["exclude_domains"] = exclude_domains response = tavily_client.search(query, **search_params) search_results = [] for result in response.get('results', []): title = result.get('title', 'No title') url = result.get('url', 'No URL') content = result.get('content', 'No content') search_results.append(f"Title: {title}\nURL: {url}\nContent: {content}\n") if search_results: return "Web Search Results:\n\n" + "\n---\n".join(search_results) else: return "No search results found." except Exception as e: return f"Search error: {str(e)}" def enhance_query_with_search(query: str, enable_search: bool) -> str: """Enhance the query with web search results if search is enabled""" if not enable_search or not tavily_client: return query # Perform search to get relevant information search_results = perform_web_search(query) # Combine original query with search results enhanced_query = f"""Original Query: {query} {search_results} Please use the search results above to help create the requested application with the most up-to-date information and best practices.""" return enhanced_query def send_to_sandbox(code): # Add a wrapper to inject necessary permissions and ensure full HTML wrapped_code = f""" {code} """ encoded_html = base64.b64encode(wrapped_code.encode('utf-8')).decode('utf-8') data_uri = f"data:text/html;charset=utf-8;base64,{encoded_html}" iframe = f'' return iframe def demo_card_click(e: gr.EventData): try: # Get the index from the event data if hasattr(e, '_data') and e._data: # Try different ways to get the index if 'index' in e._data: index = e._data['index'] elif 'component' in e._data and 'index' in e._data['component']: index = e._data['component']['index'] elif 'target' in e._data and 'index' in e._data['target']: index = e._data['target']['index'] else: # If we can't get the index, try to extract it from the card data index = 0 else: index = 0 # Ensure index is within bounds if index >= len(DEMO_LIST): index = 0 return DEMO_LIST[index]['description'] except (KeyError, IndexError, AttributeError) as e: # Return the first demo description as fallback return DEMO_LIST[0]['description'] def extract_text_from_image(image_path): """Extract text from image using OCR""" try: # Check if tesseract is available try: pytesseract.get_tesseract_version() except Exception: return "Error: Tesseract OCR is not installed. Please install Tesseract to extract text from images. See install_tesseract.md for instructions." # Read image using OpenCV image = cv2.imread(image_path) if image is None: return "Error: Could not read image file" # Convert to RGB (OpenCV uses BGR) image_rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB) # Preprocess image for better OCR results # Convert to grayscale gray = cv2.cvtColor(image_rgb, cv2.COLOR_RGB2GRAY) # Apply thresholding to get binary image _, binary = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU) # Extract text using pytesseract text = pytesseract.image_to_string(binary, config='--psm 6') return text.strip() if text.strip() else "No text found in image" except Exception as e: return f"Error extracting text from image: {e}" def extract_text_from_file(file_path): if not file_path: return "" mime, _ = mimetypes.guess_type(file_path) ext = os.path.splitext(file_path)[1].lower() try: if ext == ".pdf": with open(file_path, "rb") as f: reader = PyPDF2.PdfReader(f) return "\n".join(page.extract_text() or "" for page in reader.pages) elif ext in [".txt", ".md"]: with open(file_path, "r", encoding="utf-8") as f: return f.read() elif ext == ".csv": with open(file_path, "r", encoding="utf-8") as f: return f.read() elif ext == ".docx": doc = docx.Document(file_path) return "\n".join([para.text for para in doc.paragraphs]) elif ext.lower() in [".jpg", ".jpeg", ".png", ".bmp", ".tiff", ".tif", ".gif", ".webp"]: return extract_text_from_image(file_path) else: return "" except Exception as e: return f"Error extracting text: {e}" def extract_website_content(url: str) -> str: """Extract HTML code and content from a website URL""" try: # Validate URL parsed_url = urlparse(url) if not parsed_url.scheme: url = "https://" + url parsed_url = urlparse(url) if not parsed_url.netloc: return "Error: Invalid URL provided" # Set comprehensive headers to mimic a real browser request headers = { 'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36', 'Accept': 'text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,image/apng,*/*;q=0.8', 'Accept-Language': 'en-US,en;q=0.9', 'Accept-Encoding': 'gzip, deflate, br', 'DNT': '1', 'Connection': 'keep-alive', 'Upgrade-Insecure-Requests': '1', 'Sec-Fetch-Dest': 'document', 'Sec-Fetch-Mode': 'navigate', 'Sec-Fetch-Site': 'none', 'Sec-Fetch-User': '?1', 'Cache-Control': 'max-age=0' } # Create a session to maintain cookies and handle redirects session = requests.Session() session.headers.update(headers) # Make the request with retry logic max_retries = 3 for attempt in range(max_retries): try: response = session.get(url, timeout=15, allow_redirects=True) response.raise_for_status() break except requests.exceptions.HTTPError as e: if e.response.status_code == 403 and attempt < max_retries - 1: # Try with different User-Agent on 403 session.headers['User-Agent'] = 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36' continue else: raise # Get the raw HTML content with proper encoding try: # Try to get the content with automatic encoding detection response.encoding = response.apparent_encoding raw_html = response.text except: # Fallback to UTF-8 if encoding detection fails raw_html = response.content.decode('utf-8', errors='ignore') # Debug: Check if we got valid HTML if not raw_html.strip().startswith(' 10: print(f"Warning: This site has {len(script_tags)} script tags - it may be a JavaScript-heavy site") print("The content might be loaded dynamically and not available in the initial HTML") # Extract title title = soup.find('title') title_text = title.get_text().strip() if title else "No title found" # Extract meta description meta_desc = soup.find('meta', attrs={'name': 'description'}) description = meta_desc.get('content', '') if meta_desc else "" # Extract main content areas for analysis content_sections = [] main_selectors = [ 'main', 'article', '.content', '.main-content', '.post-content', '#content', '#main', '.entry-content', '.post-body' ] for selector in main_selectors: elements = soup.select(selector) for element in elements: text = element.get_text().strip() if len(text) > 100: # Only include substantial content content_sections.append(text) # Extract navigation links for analysis nav_links = [] nav_elements = soup.find_all(['nav', 'header']) for nav in nav_elements: links = nav.find_all('a') for link in links: link_text = link.get_text().strip() link_href = link.get('href', '') if link_text and link_href: nav_links.append(f"{link_text}: {link_href}") # Extract and fix image URLs in the HTML img_elements = soup.find_all('img') for img in img_elements: src = img.get('src', '') if src: # Handle different URL formats if src.startswith('//'): # Protocol-relative URL absolute_src = 'https:' + src img['src'] = absolute_src elif src.startswith('/'): # Root-relative URL absolute_src = urljoin(url, src) img['src'] = absolute_src elif not src.startswith(('http://', 'https://')): # Relative URL absolute_src = urljoin(url, src) img['src'] = absolute_src # If it's already absolute, keep it as is # Also check for data-src (lazy loading) and other common attributes data_src = img.get('data-src', '') if data_src and not src: # Use data-src if src is empty if data_src.startswith('//'): absolute_data_src = 'https:' + data_src img['src'] = absolute_data_src elif data_src.startswith('/'): absolute_data_src = urljoin(url, data_src) img['src'] = absolute_data_src elif not data_src.startswith(('http://', 'https://')): absolute_data_src = urljoin(url, data_src) img['src'] = absolute_data_src else: img['src'] = data_src # Also fix background image URLs in style attributes elements_with_style = soup.find_all(attrs={'style': True}) for element in elements_with_style: style_attr = element.get('style', '') # Find and replace relative URLs in background-image import re bg_pattern = r'background-image:\s*url\(["\']?([^"\']+)["\']?\)' matches = re.findall(bg_pattern, style_attr, re.IGNORECASE) for match in matches: if match: if match.startswith('//'): absolute_bg = 'https:' + match style_attr = style_attr.replace(match, absolute_bg) elif match.startswith('/'): absolute_bg = urljoin(url, match) style_attr = style_attr.replace(match, absolute_bg) elif not match.startswith(('http://', 'https://')): absolute_bg = urljoin(url, match) style_attr = style_attr.replace(match, absolute_bg) element['style'] = style_attr # Fix background images in