Tabby

Tabby is an open-source, self-hosted web app and server that provides AI-assisted coding capabilities. It functions as a server-side application that integrates with various IDEs to provide code completions and chat-based assistance while keeping data on-premises.

Users deploy Tabby as a server, often via Docker, to maintain full control over their infrastructure and privacy. It is designed to integrate directly into existing developer workflows through dedicated plugins for VSCode, Vim, and IntelliJ, allowing developers to access AI suggestions without leaving their primary editor.

Key features

• Self-contained architecture with no requirement for a DBMS or cloud service
• Support for consumer-grade GPUs and Apple M1/M2 Metal inference
• OpenAPI interface for integration with Cloud IDEs and other infrastructure
• RAG-based code completion using repository-level context
• Answer Engine for internal engineering team knowledge management
• Integration with GitHub and GitLab for SSO and merge request indexing
• Inline code editing and chat side panel for IDEs
• Automated commit message generation
• Support for multiple models including Codestral, CodeGemma, and CodeQwen

Tabby is built for engineering teams and individual developers who require privacy-focused AI tools. The architecture allows for the indexing of GitLab Merge Requests and the addition of custom documentation through REST APIs to improve context awareness. It includes a full feature admin UI for system management and provides team-wise analytics through a reports tab to monitor usage. The system also supports LDAP authentication and the ability to switch between different backend chat models within the Answer Engine.

Tabby is a private, on-premises AI coding assistant for teams that cannot use external SaaS providers.