Tabby, my open Copilot
summary
Tabby (TabbyML/tabby) is an open-source alternative to GitHub Copilot that you deploy locally (requires Node.js and model downloads) and can run offline; its core is written in Rust and distributed via Docker/docker‑compose or Heroku with an optional web admin. The demo shows using StarCoder 1B for inline code completions (examples: Celsius→Fahrenheit, prime check) and WizardCoder 3B for chat; a single instance can serve about 50 developers, the admin UI is free for up to 5 users, and scaling to more users or larger models requires paid team plans and NVIDIA GPUs.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:02Generative AI and coding copilots
- 3:10Tabby ML: open-source alternative
- 5:20Installing and configuring Tabby
- 7:07Demo with VS Code
- 8:30Admin portal and language models
- 10:17Chat model and customization
- 12:49Key limitations and concerns
- 14:35Summary and conclusion
related talks
24mTabby, my free 'Copilot'
42mTabby, My Free 'Copilot'
31mTabby, my free "Copilot"
01h 20mGithub Copilot - The World's Most Widely Adopted AI Developer Tool
55mCopilot is my Copilot: Pair programming with an AI
59mBeyond The Hype: Every Day, Real World Uses for Copilot
40mUnlocking Productivity with GitHub Copilot Best Practices and Tips
40mLevel Up with Copilot