mtchat

A Deno-based chatbot powered by RAG (Retrieval-Augmented Generation) for querying knowledge bases stored in local SQLite databases. Designed to work seamlessly with monkeytab for document and data management.

Features

• Local-first RAG system using SQLite for document storage
• Natural language queries over your knowledge base
• Integration with monkeytab for editing and managing data as tables
• Fast document retrieval and semantic search
• Privacy-focused: all data stays on your machine

Requirements

• Deno v2.0 or higher
• SQLite database with your knowledge base

Installation

git clone https://github.com/yourusername/mtchat.git
cd mtchat

Usage

deno task start

Configuration

Create a .env file in the project root:

SQLITE_DB_PATH=./data/knowledge.db

How It Works

1. Document Storage: Your documents and knowledge base are stored in a local SQLite database, managed through monkeytab
2. Retrieval: When you ask a question, mtchat searches the database for relevant documents using semantic similarity
3. Generation: Retrieved context is combined with your query to generate accurate, grounded responses

Project Structure

mtchat/
├── src/
│   ├── chat/       # Chat interface and conversation handling
│   ├── rag/        # RAG pipeline components
│   ├── db/         # SQLite database operations
│   └── search/     # Document retrieval and search
├── data/           # Local SQLite databases
├── deno.json       # Deno configuration
└── README.md

Development

# Run in development mode
deno task dev

# Run tests
deno task test

# Type check
deno check src/main.ts

License

MIT License - see LICENSE for details.

Author

Juan Pablo Marin Diaz