English | 简体中文

A production-ready AI Agent system built with FastAPI + deepagents

Login Page	Chat Interface
Streaming Response	Share Dialog

Skills Management	MCP Configuration
System Settings	Feedback System
Shared Session	Role Management

Mobile View	Tablet View

• deepagents Architecture - Compiled graph with fine-grained state management
• Multi-Agent Types - Core Agent (default), Fast Agent (optimized speed), Search Agent (web search)
• Plugin System - Register custom agents with @register_agent("id") decorator
• Streaming Output - Native SSE (Server-Sent Events) support
• Sub-agents - Multi-level agent nesting support
• Thinking Mode - Extended thinking mode for Anthropic models
• Code Interpreter - Built-in code execution with sandbox support
• Human-in-the-Loop - Approval system for sensitive operations

• Web Search Agent - Search the web with rich results (title, URL, summary, site icon)
• Domain Filtering - Limit search results to specific domains
• Time Range Filter - Filter by day/week/month/year
• Region Support - Optimize for CN or US regions
• Content Size Control - Balanced (400-600 words) or comprehensive (2500 words) summaries

• System + User Level MCP - Global and personal MCP server configs
• Encrypted Storage - Sensitive data like API keys are encrypted
• Dynamic Caching - Tool caching with manual refresh support
• Multiple Transports - Support for stdio, SSE, and HTTP transports
• Permission Control - Transport-level access permissions

• Dual Storage - File system + MongoDB backup
• Access Control - User-level skill permissions
• GitHub Sync - Sync custom skills from GitHub repositories
• Skill Creator - Built-in skill creation toolkit with evaluation tools
• 5 Skill Slots - Enable up to 5 skills per session

• Thumbs Rating - Simple positive/negative feedback
• Text Comments - Detailed user feedback
• Session Linking - Feedback tied to specific sessions/messages
• Run-Level Stats - Aggregate feedback statistics per run

• Multi-format Preview - PDF / Word / Excel / PPT / Markdown / Mermaid
• Image Viewer - Built-in image preview with zoom support
• File Upload - Drag & drop or click to upload multiple files
• Cloud Storage - S3 / OSS / MinIO integration
• Project Management - Organize conversations into projects
• Session Search - Full-text search across all conversations

• Dual-write Mechanism - Redis for real-time, MongoDB for persistence
• WebSocket Support - Real-time bidirectional communication
• Auto Reconnect - Resume conversation after disconnection
• Session Sharing - Share conversations with public or authenticated links

• JWT Authentication - Complete auth flow with token refresh
• RBAC Roles - Admin / User / Guest levels
• Password Encryption - bcrypt hashing
• OAuth Support - Login with Google, GitHub, etc.
• Email Verification - Secure email-based account verification
• Sandbox Execution - Isolated code execution environment

• Concurrency Control - Task execution queue with concurrency limits
• Cancellation - Cancel running tasks
• Heartbeat - Task health monitoring
• Pub/Sub - Event-driven task notifications
• Status Tracking - Real-time task status updates

• Feishu (Lark) - Native integration with Lark/Feishu platform
• Multi-Channel - Extensible channel system for messaging platforms
• Email Service - Built-in email notification support
• Project Management - Organize chats by projects

• LangSmith Tracing - Optional LangSmith integration for agent tracing
• Structured Logging - Context-aware structured logging
• Health Check - API health and readiness endpoints
• User Management - View and manage users
• Role Assignment - Configure agent access per role

• Modern Stack - React 19 + Vite + TailwindCSS
• ChatGPT Style - Familiar chat interface
• Theme Support - Dark/Light mode with smooth transitions
• i18n - Multi-language support (English, Chinese, Japanese, Korean)
• Responsive Design - Mobile, tablet, and desktop optimized
• Agent Switcher - Toggle between Core/Fast/Search agents

LambChat supports 14 setting categories, configurable via the Settings page or environment variables:

• Python 3.12+
• Node.js 18+
• MongoDB
• Redis

# Clone repository git clone https://github.com/Yanyutin753/LambChat.git cd LambChat # Copy environment file cp .env.example .env # Edit .env with your configuration # Start with Docker (recommended) docker-compose up -d # Or run locally make install # Install dependencies make dev # Start development server

Access the app at http://localhost:8000

# Format code ruff format src/ # Check style ruff check src/ # Type check mypy src/

src/ ├── agents/ # Agent implementations (core, fast, search) ├── api/ # FastAPI routes and middleware ├── infra/ # Infrastructure services │ ├── auth/ # JWT authentication │ ├── backend/ # Backend management │ ├── channel/ # Multi-channel (Feishu, etc.) │ ├── email/ # Email service │ ├── feedback/ # Feedback system │ ├── folder/ # Project management │ ├── llm/ # LLM integration │ ├── memory/ # Memory & hindsight │ ├── mcp/ # MCP protocol │ ├── patches/ # Monkey patches and compat fixes │ ├── role/ # RBAC role management │ ├── sandbox/ # Sandbox execution │ ├── session/ # Session management (dual-write) │ ├── settings/ # Settings service │ ├── share/ # Session sharing │ ├── skill/ # Skills system │ ├── storage/ # File storage │ ├── task/ # Task management │ ├── tool/ # Tool registry & MCP client │ ├── tracing/ # LangSmith tracing │ ├── user/ # User management │ └── websocket/ # WebSocket support ├── kernel/ # Core schemas, config, types └── skills/ # Built-in skills (skill-creator) 

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

MIT

Made with ❤️ by Clivia