 Login |  Chat |  Streaming |
 Skills |  MCP Config |  Share |
 Roles |  Settings |  Mobile |
🤖 Agent System
- deepagents Architecture — Compiled graph with fine-grained state management
- Multi-Agent Types — Core / Fast / Search agents
- Plugin System —
@register_agent("id")decorator for custom agents - Streaming Output — Native SSE support
- Sub-agents — Multi-level nesting
- Thinking Mode — Extended thinking for Anthropic models
- Human-in-the-Loop — Approval system for sensitive operations
🔌 MCP Integration
- System + User Level — Global and per-user MCP configs
- Encrypted Storage — API keys encrypted at rest
- Dynamic Caching — Tool caching with manual refresh
- Multiple Transports — stdio / SSE / HTTP
- Permission Control — Transport-level access control
🛠️ Skills System
- Dual Storage — File system + MongoDB backup
- Access Control — User-level permissions
- GitHub Sync — Import skills from GitHub repos
- Skill Creator — Built-in creation toolkit with evaluation tools
💬 Feedback · 📁 Files · 🔄 Realtime · 🔐 Auth · ⚙️ Tasks · 📊 Observability
- Feedback — Thumbs rating, text comments, session-linked, run-level stats
- Documents — PDF / Word / Excel / PPT / Markdown / Mermaid preview + image viewer
- Cloud Storage — S3 / OSS / MinIO integration, drag & drop upload
- Realtime — Dual-write (Redis + MongoDB), WebSocket, auto-reconnect, session sharing
- Security — JWT, RBAC (Admin/User/Guest), bcrypt, OAuth, email verification, sandbox
- Tasks — Concurrency control, cancellation, heartbeat, pub/sub notifications
- Observability — LangSmith tracing, structured logging, health checks
- Channels — Feishu (Lark) native integration, extensible multi-channel system
🎨 Frontend
- React 19 + Vite + TailwindCSS
- ChatGPT-style interface with dark/light theme
- i18n — English, Chinese, Japanese, Korean
- Responsive — Mobile, tablet, desktop
14 setting categories configurable via UI or environment variables:
| Category | Description |
|---|---|
| Frontend | Default agent, welcome suggestions, UI preferences |
| Agent | Debug mode, logging level |
| LLM | Model, temperature, max tokens, API key & base URL |
| Session | Session management |
| Database | MongoDB connection |
| Storage | Persistent storage, S3/OSS/MinIO |
| Security | Encryption & security policies |
| Sandbox | Code sandbox settings |
| Skills | Skill system config |
| Tools | Tool system settings |
| Tracing | LangSmith & tracing |
| User | User management |
| Memory | Memory system (hindsight) |
- Python 3.12+ · Node.js 18+ · MongoDB · Redis
git clone https://github.com/Yanyutin753/LambChat.git cd LambChat # Docker (recommended) cd deploy && cp .env.example .env # Edit with your config docker compose up -d # Or local development cp .env.example .env # Edit with your config make install && make dev→ Open http://localhost:8000
ruff format src/ # Format ruff check src/ # Lint mypy src/ # Type checksrc/ ├── agents/ # Agent implementations (core, fast, search) ├── api/ # FastAPI routes & middleware ├── infra/ # Infrastructure services │ ├── agent/ # Agent config & events │ ├── auth/ # JWT authentication │ ├── channel/ # Multi-channel (Feishu, etc.) │ ├── feedback/ # Feedback system │ ├── llm/ # LLM integration │ ├── mcp/ # MCP protocol │ ├── memory/ # Memory & hindsight │ ├── role/ # RBAC │ ├── sandbox/ # Sandbox execution │ ├── session/ # Session management (dual-write) │ ├── skill/ # Skills system │ ├── storage/ # S3/OSS/MinIO │ ├── task/ # Task management │ ├── tool/ # Tool registry & MCP client │ └── ... ├── kernel/ # Core schemas, config, types └── skills/ # Built-in skills MIT — Project name "LambChat" and its logo may not be changed or removed.
Made with ❤️ by Clivia