zylarian/dockyard-tabby repository overview

⁠Tabby for Dockyard

Self-hosted AI coding assistant. An open-source alternative to GitHub Copilot.

⁠Quick Start

make service tabby start

Access the interface at http://localhost:8100⁠

⁠Configuration

⁠Environment Variables

⁠GPU Support

This service is configured to use CPU by default for maximum compatibility.

To enable GPU acceleration:

1. Ensure you have NVIDIA Container Toolkit installed
2. Create a .env file from .env.example
3. Set TABBY_DEVICE=cuda
4. Add the following to docker-compose.yml under the tabby service:

   deploy:
     resources:
       reservations:
         devices:
           - driver: nvidia
             count: 1
             capabilities: [gpu]
   

   Copy

⁠IDE Integration

1. Install the Tabby plugin for VS Code or IntelliJ
2. Set the endpoint to http://localhost:8100 (or your server IP)
3. Set the authentication token (found in the logs or dashboard)