An intelligent Chrome extension that provides AI-powered chat and page summaries using flexible LLM providers.
NEW: Automatic retry with exponential backoff, intelligent fallback provider switching, and real-time network monitoring for uninterrupted AI assistance!
ENHANCED: Robust error handling, network connectivity detection, and seamless provider failover ensure reliable performance even during outages.
Now supporting 7 major AI providers, including Ollama with 30+ models for every use case:
- ⚡ Groq - Ultra-fast inference
- 🤖 OpenAI - GPT-4o, GPT-4 Turbo
- 🧠 Anthropic - Claude 3.5 Sonnet, Opus
- 🔍 Google Gemini - 2.5 Pro/Flash, Enhanced with structured output
- 😄 Grok (X.AI) - Witty personality, real-time info
- 🦙 SambaNova - Llama 3.1/3.2 (1B-405B)
- 💻 Ollama - Access local models via Ollama (local)
- Multi-Provider AI Support: Choose between 8 major AI providers - OpenAI, Anthropic, Google Gemini, Grok (X.AI), Groq, SambaNova, and GPT-OSS (Online/Offline)
- 🏆 GPT-OSS Integration: OpenAI's open-weight reasoning models with dual hosting options (Groq hosted + Ollama local)
- 🤖 Iterative Agent System: Memory-driven multi-agent framework inspired by Vercel AI SDK patterns for reliable web automation
- 🧠 Agent Memory & Messaging: Persistent context and structured communication between agents for better decision making
- AI Tools System: 10 specialized tools for explaining, rewriting (paraphrase, improve, expand, shorten), and tone changes (academic, professional, persuasive, casual, funny)
- Interactive Sample Prompts: One-click prompt insertion for common tasks like page summarization
- Intelligent Chat: Interactive AI conversations with streaming responses
- Smart Page Content Extraction: Sophisticated system that can extract and analyze content from any webpage structure
- Enhanced Page Attachment: Twitter-card style page previews with persistent context throughout conversations
- Automatic Retry: Exponential backoff retry system for failed requests with configurable settings
- Fallback Provider Switching: Intelligent switching to alternative providers when primary fails
- Network Connectivity Monitoring: Real-time network status detection and handling
- Smart Error Classification: Distinguishes between temporary and permanent failures
- Provider Health Monitoring: Real-time connectivity testing and status reporting
- Configurable Recovery Settings: Customizable retry limits, delays, and backoff behavior
- Lazy Loading System: Intelligent loading of conversation messages with configurable thresholds
- Message Compression: Automatic compression for old messages with Web Worker processing
- Automatic Cleanup: Smart cleanup policies with age-based deletion and space optimization
- Performance Monitoring: Real-time metrics tracking with automatic optimization triggers
- Memory Management: Intelligent cache management with LRU eviction and memory limits
- Context Optimization: Advanced processing for large conversation histories
- Persistent Chat History: Conversations persist across sidepanel ↔ fullscreen mode switches
- Seamless Mode Switching: Continue conversations without interruption when changing modes
- Enhanced Streaming: Real-time content display with graceful streaming visualization
- Smart Button Timing: Copy/retry buttons only appear after AI responses are complete
- Conversation Continuity: Automatic conversation loading and restoration on startup
- Lightning-Fast Loading: Conversation history loads in ~100ms (vs 2-3 seconds before)
- Stop/Cancel Button: Dynamic button to stop AI responses with clean cancellation
- Optimized Storage: Lightweight message storage for instant access
- Non-blocking Operations: All saves happen asynchronously without UI freezing
- Smart History Management: Only loads recent messages for optimal performance
- Universal Page Support: Works on ANY page including chrome://, edge://, settings, new tabs
- Tool-Based Planning: Planner creates simple tool calls instead of complex upfront plans
- Memory-Driven Context: Agents share state through persistent memory for better coordination
- Iterative Execution: Like Vercel AI SDK's stopWhen pattern - plan → execute → store → repeat
- Visual Highlighting: Red borders and numbered labels that appear consistently
- Smart Re-planning: Planner sees actual page elements before planning interactions
- Agent Messaging: Structured communication between agents with conversation history
- Natural Language Processing: Describe web tasks in plain English
- Real-time Execution: Clear progress monitoring with data extraction logging
- Specialized Agents Available: FormHandler, DataExtractor, SearchAgent, ErrorRecovery kept as utilities
- Sub-toolbar Navigation: "New", "Chats", and "Agent" buttons in sidepanel mode for easy access
- Sidepanel Conversation List: Full conversation history with rename/delete in compact sidepanel view
- Smart Scroll Controls: Scroll-to-bottom button with intelligent visibility and auto-scroll behavior
- Multi-line Input: Textarea support with Ctrl+Enter for new lines and auto-resize functionality
- Enhanced Copy Options: Smart text copying with markdown stripping for clean, readable text
- Seamless Navigation: SPA-like transitions between chat, conversation, and agent views
- New Conversation Button: Contextual button that only appears when there's existing conversation history
- Fullscreen Conversation Sidebar: Complete conversation history with sidebar navigation in fullscreen mode
- Conversation Management: Rename, delete, and switch between conversations with three-dot menu
- Session-Based Storage: Persistent conversation sessions with automatic saving and loading
- One-Click Reset: Instantly start fresh conversations with clean state management
- Intelligent UI: Smart visibility controls for better user experience
- Event-Driven Architecture: Clean state management across all components
- Smart Conversation Management: Contextual "New Conversation" button that appears only when needed
- Fullscreen Conversation Sidebar: Complete conversation history navigation with rename/delete capabilities (Chrome only)
- Robust Sidepanel Management: Enhanced tab validation, smart selection, and error recovery
- Network Status Monitoring: Real-time connectivity indicators and provider health dashboard
- Seamless Error Recovery: Transparent fallback handling with user notifications
- Fullscreen Mode: Seamless switching between sidepanel and fullscreen experience (Chrome only)
- Browser Compatibility: Automatic Edge detection with graceful feature degradation
- Comprehensive Documentation: Interactive user guide with troubleshooting and FAQ
- Enhanced Onboarding: Welcome flow opens directly in sidepanel mode for better UX
- Flexible Configuration: Easy provider switching and API key management with connectivity testing
- Modern UI: Clean, responsive interface with dark mode support and improved layouts
- Privacy-Focused: Secure local storage of API keys with encryption - Privacy Policy
- Local Processing: All performance optimization happens locally
- Streamlined Interface: Side panel and fullscreen modes with keyboard shortcuts
- Chrome browser (version 88+) - Recommended for full functionality
- Microsoft Edge (version 88+) - Supported with limited features (sidepanel mode only)
- No API key required initially - Start with 5 free requests!
- API key from supported AI providers for continued use (Google Gemini recommended - free tier available)
-
Clone the repository
git clone https://github.com/kwesinavilot/delight.git cd delight -
Install dependencies
npm install
-
Build the extension
npm run build
-
Load in Chrome
- Open Chrome and navigate to
chrome://extensions/ - Enable "Developer mode"
- Click "Load unpacked" and select the
distfolder
- Open Chrome and navigate to
- Open the extension (click the icon or use
Ctrl+Shift+Q) - Go to Settings and configure your preferred AI provider(s):
- OpenAI: Add your API key from OpenAI Platform
- Anthropic: Add your API key from Anthropic Console
- Google Gemini: Add your API key from Google AI Studio
- Grok (X.AI): Add your API key from X.AI Console
- Groq: Add your API key from Groq Console
- SambaNova: Add your API key from SambaNova Cloud
- GPT-OSS Online: Add your Groq API key from Groq Console
- GPT-OSS Offline: Install Ollama and run
ollama pull gpt-oss-20b
- Try it immediately with 5 free requests, then add your API key for continued use
delight/
├── src/
│ ├── components/ # React components
│ │ ├── Chat/ # Chat interface components
│ │ ├── Summary/ # Summary panel components
│ │ └── ui/ # Reusable UI components
│ ├── services/ # Core business logic
│ │ ├── ai/ # AI service layer
│ │ │ ├── providers/ # AI provider implementations
│ │ │ └── AIService.ts # Main AI orchestration
│ │ └── config/ # Configuration management
│ ├── types/ # TypeScript type definitions
│ ├── utils/ # Utility functions
│ ├── pages/ # Extension pages (popup, sidepanel)
│ └── background/ # Background scripts
├── public/ # Static assets and manifest
└── tests/ # Test files
npm run dev- Start development server with hot reloadnpm run build- Build for productionnpm run preview- Preview production buildnpm run clean- Clean build artifactsnpm test- Run test suite
The extension uses a modular architecture with clear separation of concerns:
- AIService: Main orchestration layer managing provider interactions
- Providers: Individual implementations for OpenAI, Anthropic, Gemini, Grok, Groq, and SambaNova
- Configuration: Secure management of API keys and settings
- React + TypeScript: Modern component-based architecture
- Tailwind CSS: Utility-first styling with Radix UI components
- Responsive Design: Works seamlessly in popup and side panel modes
- Manifest V3: Modern extension architecture with service workers
- Storage API: Secure, encrypted storage of user preferences
- Content Scripts: Page interaction for summary generation
- Models: GPT-4o, GPT-4o Mini, GPT-4 Turbo, GPT-4, GPT-3.5 Turbo
- Features: Streaming responses, function calling, reliable performance
- Use Cases: General conversation, creative writing, code assistance
- Context: Up to 128K tokens
- Models: Claude 3.5 Sonnet, Claude 3 Opus, Claude 3 Sonnet, Claude 3 Haiku
- Features: Large context windows, ethical AI responses, safety-focused
- Use Cases: Research, analysis, long-form content, complex reasoning
- Context: Up to 200K tokens
- Models: Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.0 Flash, Gemma 3 series
- Features: Ultra-fast responses, structured output, agent automation, vision support
- Use Cases: Real-time chat, agent automation, task planning, structured data extraction
- Context: Up to 2M tokens (Gemini 2.5 Pro)
- 🤖 Agent Features: Structured JSON output for task planning and automation workflows
- Models: Grok Beta
- Features: Witty personality, real-time information access, humor
- Use Cases: Conversational AI, creative writing, engaging content, current events
- Context: Up to 131K tokens
- Models: Llama 3.1/3.3/4, Mixtral, Gemma, Qwen, DeepSeek
- Features: Ultra-fast inference, high-performance computing, low latency
- Use Cases: Real-time applications, high-volume processing, speed-critical tasks
- Context: Up to 131K tokens
- Performance: Industry-leading inference speed
- Models: Llama 3.1 (8B, 70B, 405B), Llama 3.2 (1B, 3B, 11B, 90B Vision)
- Features: High-performance computing, open-source models, vision support
- Use Cases: Performance-focused tasks, research, scalable deployment
- Context: Up to 131K tokens
- Models: gpt-oss-120b, gpt-oss-20b
- Features: OpenAI's open-weight reasoning models, ultra-fast inference via Groq
- Use Cases: Hackathon projects, open model research, reasoning tasks
- Context: Up to 131K tokens
- Setup: Requires Groq API key
- Models: gpt-oss-120b, gpt-oss-20b
- Features: Fully local inference, no API key required, complete privacy
- Use Cases: Offline development, privacy-focused applications, local reasoning
- Context: Up to 131K tokens
- Setup: Requires Ollama installation and model download
interface AIConfiguration {
provider: 'openai' | 'anthropic' | 'gemini' | 'grok' | 'groq' | 'sambanova';
apiKey: string;
model: string;
maxTokens: number;
temperature: number;
}The project includes comprehensive test coverage:
# Run all tests
npm test
# Run tests with coverage
npm run test:coverage
# Run tests in watch mode
npm run test:watch- Unit Tests: Individual component and service testing
- Integration Tests: End-to-end workflow testing
- Mocked APIs: Isolated testing without external dependencies
- Local Storage: All data stored locally in Chrome's secure storage
- API Key Encryption: Basic encryption for stored credentials
- No Data Collection: Extension doesn't collect or transmit user data
- Secure Communication: Direct HTTPS communication with AI providers
- ✅ 🔄 Enhanced Error Recovery: Automatic retry with exponential backoff for failed requests
- ✅ 🌐 Intelligent Fallback: Smart provider switching when primary provider fails
- ✅ 📶 Network Monitoring: Real-time connectivity detection and handling
- ✅ 🛠️ Provider Health: Connectivity testing and status dashboard
- ✅ ⚙️ Configurable Recovery: Customizable retry settings and error handling behavior
- ✅ 🎯 Context Menu Integration: Right-click menus for instant AI assistance
- ✅ 🚀 Smart Auto-Send: Automatic message sending for all context actions
- ✅ 📎 Auto Page Context: Automatic page attachment for page-level actions
- ✅ 🛡️ AI Accuracy Education: Comprehensive AI safety and limitation guidance
- ✅ 🔄 New Chat Creation: Fresh conversations for each context menu action
- ✅ 🧠 Agent Memory System: Persistent context and conversation history across automation steps
- ✅ 📨 Agent Messaging: Structured communication between agents with typed message schemas
- ✅ 🔄 Iterative Execution: Vercel AI SDK-inspired pattern - plan → execute → store → repeat
- ✅ 🛠️ Tool-Based Planning: Planner creates simple tool calls instead of complex upfront plans
- ✅ 🎯 Context-Aware Planning: Planner sees actual page elements before planning interactions
- ✅ 🌐 Universal Page Support: Works on ANY page including chrome://, edge://, settings, new tabs
- ✅ 🚀 Simplified Automation: Reliable navigate, click, fill, extract, wait, analyze actions
- ✅ 🎨 Visual Highlighting: Consistent red borders and numbered labels on all pages
- ✅ 🏆 GPT-OSS Integration: OpenAI's open-weight reasoning models (gpt-oss-120b, gpt-oss-20b)
- ✅ 🔄 Dual GPT-OSS Hosting: Groq hosted + Ollama local options
- ✅ 🤖 Multi-Agent System: Enhanced Planner, Navigator, Monitor with specialized utilities
- ✅ 🗣️ Natural Language Tasks: Describe web automation in plain English
- ✅ ⚡ Real-time Execution: Clear progress monitoring and error recovery
- ✅ 🛠️ AI Tools System: 10 specialized tools (explain, rewrite, tone changes)
- ✅ 📎 Smart Page Attachment: Twitter-card style previews with persistent context
- ✅ 🌐 Multi-Provider Support: 8 major AI providers with 27+ models
- ✅ 🔒 Secure & Private: Local storage with encryption, no data collection
- 🔲 Enhanced Tool System: More sophisticated browser automation tools
- 🔲 Agent Learning: Memory-based improvement from user feedback
- 🔲 Visual Validation: Screenshot-based task verification
- 🔲 Specialized Agent Integration: Bring FormHandler, DataExtractor back into main flow
- 🔲 Agent Templates: Pre-built automation workflows for common tasks
- 🔲 Export/Import Conversations: Backup and restore chat history
- 🔲 Custom Prompt Templates: User-defined prompt libraries
- 🔲 Advanced summary customization
- 🔲 Batch processing for multiple pages
- 🔲 Additional AI providers (Cohere, Mistral)
- 🔲 Multi-language support
We welcome contributions! Please see our Contributing Guide for details.
- Fork the repository
- Create a feature branch:
git checkout -b feature/amazing-feature - Make your changes and add tests
- Ensure tests pass:
npm test - Commit your changes:
git commit -m 'Add amazing feature' - Push to the branch:
git push origin feature/amazing-feature - Open a Pull Request
This project is licensed under the MIT License - see the LICENSE file for details.
- Vercel AI SDK for the excellent AI integration library
- Radix UI for accessible UI components
- Tailwind CSS for utility-first styling
- OpenAI and Anthropic for their powerful AI models
- ✅ Full feature support
- ✅ Sidepanel and fullscreen modes
- ✅ All AI tools and page attachment
- ✅ Seamless mode switching
- ✅ Sidepanel mode fully supported
⚠️ Fullscreen mode supported but may have issues- ✅ All AI tools and page attachment
⚠️ Some Chrome-specific behaviors may differ
Note: For the best experience, we recommend using Google Chrome. Edge users will receive helpful notifications about feature limitations.
- Issues: GitHub Issues
- Discussions: GitHub Discussions
- Email: support@delight.com
Made with ❤️ by the Delight team