Tabby

Tabby is an open-source, self-hosted AI coding assistant that lets teams easily set up their own LLM-powered code completion servers. With Tabby, you can configure deployments with simple TOML configurations and host your own deployment for flexibility and security.

Tabby is built with Rust, which offers best-in-class speed and safety, giving you a fast and reliable coding experience. The project is open source, so you can audit the entire codebase on GitHub for security or compliance.

Some of the key features of Tabby include:

• Easy Configuration: Control deployment with simple TOML configurations.
• Open Source: The entire codebase is available for auditing and hosting on GitHub.
• Rust Powered: Speed and safety for your coding experience.
• IDE/Editor Extensions: Native support for popular IDEs and editors.
• Models Directory: Curated list of models recommended for use with Tabby.
• API References: Detailed documentation for the Tabby API.
• Community Support: Engage with the community through GitHub, Slack, Twitter, LinkedIn, and a newsletter.

Tabby supports a variety of use cases:

• IDE Extensions: Streaming and cancellation with an adaptive caching strategy for fast completion.
• Model Serving: Code parsing into Tree Sitter tags for great prompts.
• User and Developer Experience: Optimized experience for AI experts, EngOps teams, and developers.

You can install Tabby using Docker, Homebrew, Hugging Face Space, or other deployment methods. Tabby supports both CUDA and CPU configurations and can be extended with repositories from Git, GitHub, GitLab, and more through a user-friendly interface or configuration files.

The project is actively maintained with new features and bug fixes added regularly. Upgrading Tabby is as easy as backing up the database and restarting the application.

Tabby is designed to help teams get more out of their productivity by offering a self-hosted AI coding assistant that is flexible and secure. Check out the Tabby website for more information and to explore its full range of features and documentation.

Published on June 9, 2024