Scripts2026年3月31日·1 分钟阅读

Tabby — Self-Hosted AI Coding Assistant

Self-hosted AI code completion and chat assistant. Privacy-first alternative to GitHub Copilot. Supports 20+ models, repo-aware context, and IDE integrations. 33K+ stars.

TO
TokRepo精选 · Community
快速使用

先拿来用,再决定要不要深挖

这里应该同时让用户和 Agent 知道第一步该复制什么、安装什么、落到哪里。

# Run with Docker (GPU)
docker run -it --gpus all -p 8080:8080 \
  tabbyml/tabby serve --model StarCoder-1B --chat-model Qwen2.5-Coder-1.5B-Instruct

# Or CPU-only
docker run -it -p 8080:8080 \
  tabbyml/tabby serve --model StarCoder-1B --device cpu

Open http://localhost:8080 for the admin dashboard. Install the VS Code/JetBrains extension and point it to your Tabby server.


介绍

Tabby is a self-hosted AI coding assistant — a privacy-first, open-source alternative to GitHub Copilot. It provides code completion and chat powered by your choice of 20+ open-source models, with repo-aware context that understands your codebase. Supports VS Code, JetBrains, Vim/Neovim, and Emacs. No data leaves your infrastructure. 33,000+ GitHub stars.

Best for: Teams needing AI code assistance with full data privacy and self-hosting control Works with: VS Code, JetBrains (IntelliJ, PyCharm, WebStorm), Vim, Neovim, Emacs


Key Features

Code Completion

Fast, inline code suggestions as you type — like Copilot but self-hosted.

Chat

Ask questions about your code, get explanations, refactoring suggestions, and bug fixes.

Repository Context

Index your codebase for repo-aware completions. Tabby understands your project's patterns, APIs, and conventions.

20+ Models

StarCoder, CodeLlama, DeepSeek-Coder, Qwen-Coder, and more. Switch models via config.

Admin Dashboard

Web UI for model management, user management, analytics, and repository indexing.

Privacy

  • Self-hosted — runs on your own GPU or cloud instance
  • No telemetry — zero data sent externally
  • Air-gapped — works fully offline

FAQ

Q: What is Tabby? A: A self-hosted AI coding assistant with code completion and chat. Open-source alternative to GitHub Copilot with full data privacy. 33K+ GitHub stars.

Q: What hardware do I need? A: A GPU with 4GB+ VRAM for small models (StarCoder-1B), 8GB+ for larger models. CPU-only mode available but slower.


🙏

来源与感谢

Created by TabbyML. Licensed under SSPL (Server Side Public License). TabbyML/tabby — 33,000+ GitHub stars

相关资产