Don

Don is an AI agent that connects Large Language Models (LLMs) directly to command-line tools, enabling autonomous task execution without requiring a separate MCP client.

Features

• Direct LLM connectivity: Connect to OpenAI, Anthropic, Ollama, and other LLM providers
• Multi-agent architecture: Uses orchestrator and tool-runner agents for complex task execution
• RAG support: Retrieval-Augmented Generation for document-based Q&A
• Flexible configuration: YAML-based configuration with environment variable substitution
• Multiple retrieval strategies: Chunked embeddings, BM25, and hybrid search

Quick Start

1. Create an agent configuration file at ~/.don/agent.yaml:

   agent:
     models:
       - model: "gpt-4o"
         class: "openai"
         name: "gpt-4o"
         default: true
         api-key: "${OPENAI_API_KEY}"
         api-url: "https://api.openai.com/v1"

2. Set your API key:

   export OPENAI_API_KEY="sk-..."

3. Run the agent with a tools configuration:

   don --tools=examples/tools.yaml "Help me debug this issue"

Installation

go install github.com/inercia/don@latest

Or build from source:

git clone https://github.com/inercia/don
cd don
make build

Usage

Basic Usage

# Run with a specific model
don --tools=tools.yaml --model gpt-4o "Your question here"

# One-shot mode (exit after response)
don --tools=tools.yaml --once "What's the disk usage?"

# Interactive mode
don --tools=tools.yaml

With RAG

# Enable RAG sources from config
don --tools=tools.yaml --rag=docs "What does the documentation say about X?"

Subcommands

# Show agent configuration
don info

# Create default configuration
don config create

# Show current configuration
don config show

Configuration

See the Configuration Guide for detailed configuration options.

Documentation

• Usage Guide - Getting started and basic usage
• Configuration - Agent and model configuration
• RAG Guide - Using Retrieval-Augmented Generation
• Architecture - Technical architecture and design

License

This project is licensed under the MIT License - see the LICENSE file for details.