DocuChat

DocuChat is a simple Terminal RAG (Retrieval-Augmented Generation) application that allows you to talk to your files in complete privacy. Using Ollama, both the embedding and inference happens locally.

Features

• Document Processing: Automatically processes PDF, TXT, and DOCX files from multiple directories.
• Source Switching: Quickly and easily switch source folders, or access all your sources together.
• Intelligent Indexing: Creates and maintains an efficient index of document content for quick retrieval.
• Natural Language Queries: Allows users to ask questions in natural language about the content of their documents.
• Citation Support: Provides citations for information sources, linking responses directly to document pages.
• Conversation Tracking: Saves transcripts of conversations for future reference.
• User-Friendly Interface: Offers a clean, color-coded command-line interface for easy interaction.

Prerequisites

• Python 3.8+
• Ollama (for running the LLM locally)

Installation

1. Clone the repository:

   git clone https://github.com/rb81/docuchat.git
   cd docuchat
   

2. Install the required dependencies:

   pip install -r requirements.txt
   

3. Configure the application by modifying the config.py file with your settings.

Usage

1. Ensure Ollama is running with your chosen model.

2. Run the DocuChat application:

   python main.py
   

3. Start chatting! Ask questions about your documents, and DocuChat will provide answers with relevant citations.

4. Type /source to switch the source folder anytime (must be configured in the config.py file).

5. Type /quit to exit the application and save your conversation transcript.

License

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

Transparency Disclaimer

ai.collaboratedwith.me in creating this project.