lukawi 0.1.0

A lightweight AI Agent framework with ReAct loop, tool calling, and memory

Lukawi Agent

A lightweight, extensible AI Agent framework with ReAct loop, tool calling, memory, and knowledge base.

Features

• ReAct Loop Agent — reasoning + acting with automatic tool calling
• Streaming Chat — token-by-token real-time output in WebUI
• Tool System — built-in tools: web fetch, file operations, shell commands
• Extensible Skills — drop SKILL.md files to teach the agent new abilities
• MCP Protocol — connect external tool servers (sequential thinking, code context, etc.)
• Memory — session-based short-term + persistent long-term memory with search
• RAG Knowledge Base — upload documents, semantic search via DashScope embeddings
• WebUI — React-based chat interface with model switching, session management
• TUI — terminal-based interactive REPL with Rich formatting

Quick Start

1. Install

pip install lukawi

2. Initialize

lukawi-init

This interactive wizard will ask for:

• DeepSeek API Key (required for LLM)
• DashScope API Key (optional, for RAG / knowledge base)
• MCP Servers (optional, for extra tools)

3. Launch

# Web UI (recommended)
lukawi webui

# Terminal REPL
lukawi

Open http://localhost:50109 in your browser.

4. Add Skills (optional)

Drop SKILL.md files into the skills/ directory:

skills/
├── my-skill/
│   └── SKILL.md

Configuration

All configuration is read from config/default.yaml. Override via environment variables or a .env file:

DEEPSEEK_API_KEY=sk-...
DEEPSEEK_BASE_URL=https://api.deepseek.com/v1
DASHSCOPE_API_KEY=sk-...

MCP servers are configured in ~/.lukawi/mcp-servers.json.

Keyboard Shortcuts

Shortcut	Action
Ctrl+B	Toggle sidebar
Ctrl+L	Clear chat
Enter	Send message
Shift + Enter	New line
/	Command mode

Development

git clone https://github.com/lukawi-team/lukawi.git
cd lukawi
pip install -e ".[dev]"

# Run tests
pytest

# Build frontend
cd web && npm install && npm run build

License

MIT