8000 GitHub - rsrohan99/tig: Terminal-based AI Coding Agent, similar to Claude Code, OpenAI Codex etc. but works with many more LLMs e.g. Gemini, Groq, Deepseek
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content
/ tig Public

Terminal-based AI Coding Agent, similar to Claude Code, OpenAI Codex etc. but works with many more LLMs e.g. Gemini, Groq, Deepseek

Notifications You must be signed in to change notification settings

rsrohan99/tig

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

45 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🐯 Tig

An automonous AI coding agent that runs in the terminal. Similar to Claude Code, OpenAI Codex etc. but works with many more LLMs.

It can:

  • Write code 📝
  • Fix bugs 🐛
  • Execute shell commands 💻
  • Write tests 🧪
  • Analyze codebase 🔍 and many more...

All within the terminal! 🚀

Full tutorial series on how to build it from scratch 👇

tig

Modes

It has different modes for different tasks:

  • 📝 Architect: Designs the system, brainstorms ideas with the user, saves the design in a markdown file.
  • 💻 Code: Implements the architect's plan step by step

✅ LLMs supported so far:

  • Google Gemini
  • OpenAI
  • Claude
  • OpenRouter
  • Deepseek
  • Groq
  • Local LLMs using Ollama

Tools Used

  • LlamaIndex Workflow: For orchestrating and multiple LLMs support
  • Tree-sitter: For searching code definitions (functions, classes, etc.) accross the codebase and checking for syntax errors
  • Ripgrep: For regex search accross the codebase
  • Google's diff-match-patch: For comparing and displaying diffs

Requirements

Tig depends on some external tools:

  • Ripgrep: For regex search accross the codebase

Make sure to install it for your operating system.

For macOS, you can use Homebrew:

brew install ripgrep

For Linux, you can use the package manager for your distro:

For Arch:

sudo pacman -S ripgrep

For Fedora:

sudo dnf install ripgrep

Installation (using pip)

Install Tig using pip:

pip install tig-code

Note: If tig is not in your PATH, you need to add it first (depends on your python installation)

Usage

Tig is configured using .env file.

Specify the LLM and model to use alongside with appropriate API keys in the .env file.

GOOGLE_API_KEY="..."
TIG_MODEL="gemini-2.0-flash"
TIG_PROVIDER="google" # available providers: [google, openai, anthropic, deepseek, groq, ollama, openrouter]

Provide the right API key variable for the right model e.g. GOOGLE_API_KEY for Gemini, OPENAI_API_KEY for OpenAI, ANTHROPIC_API_KEY for Claude, etc.

Now run tig (optionally specify the mode --mode <code| architect>):

tig

If you want to auto-approve all the actions taken by Tig (read, write, update files), run tig with the --auto-approve flag:

tig --auto-approve

Finally when prompted, provide tig with a task to get started:

...
New task: Create a screen recorder website for chrome
...

About

Terminal-based AI Coding Agent, similar to Claude Code, OpenAI Codex etc. but works with many more LLMs e.g. Gemini, Groq, Deepseek

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

0