Skip to main content

git-sage

Local AI code review for your git workflow no cloud, no subscriptions, no data leaving your machine.

git-sage is a Python CLI tool that hooks into git and runs an AI-powered code review using a locally hosted model via Ollama every time you push. It intercepts the push, analyzes your staged diff, and either approves it or asks you to revise; all in seconds, entirely on your machine.

What it does​

Most AI code review tools sit at the pull request stage, they analyse your code after it has already been pushed to a remote repository. By that point, a hardcoded secret has already touched a cloud server. A vulnerable dependency has already been committed to a branch that other developers may have pulled. The damage window opens the moment you push, not the moment the PR is reviewed.

git-sage moves the review earlier. It runs as a pre-push git hook, which means it catches issues on your local machine before any code is uploaded to the cloud. No remote repository involved. No branch created. If the model finds a problem, the push is aborted and you fix it right there in your editor.

$ git push

  Staged: 1 file(s)  +6 / -0

  ╭─ Summary ──────────────────────────────────────────────────╮
  │ Adds a login endpoint with bcrypt password hashing.        │
  ╰────────────────────────────────────────────────────────────╯

  Issues  (2 found)

  ●  1. SECRET_KEY is hardcoded as a string literal on line 14.
  ●  2. No rate limiting on the /login route.

  Suggestions  (1)

  ◆  1. Load SECRET_KEY from os.getenv('SECRET_KEY') instead.

  ╭────────────────────────────────────────────────────────────╮
  │  ✗  REVISE                                                 │
  │  Address the issues above before pushing.                  │
  ╰────────────────────────────────────────────────────────────╯

  Push aborted by git-sage. Fix the issues above, or run:
    git push --no-verify   to bypass the hook.

Why local AI?​

Since most code review tools send your code to a remote server. git-sage keeps everything on your machine, ensuring the following:

  • Privacy: your code never leaves your laptop
  • No API costs: runs on your own hardware
  • Works offline: no internet connection required
  • Speed: no network latency, runs in 10–30 seconds depending on your machine

The model: qwen2.5-coder:7b​

git-sage uses qwen2.5-coder:7b by default, a code-specialized model that:

  • Understands unified diffs natively
  • Outperforms general-purpose 7B models on code tasks
  • Runs in ~4.5 GB RAM at Q4 quantization
  • Works on any modern laptop without a GPU

You can swap it for any Ollama-compatible model with the --model flag.

Quick start​

# 1. Install Ollama and pull the model
brew install ollama
ollama serve
ollama pull qwen2.5-coder:7b

# 2. Install git-sage
pip install git-sage

# 3. Install the hook in your repo
cd your-project
git-sage install

# 4. Push normally; git-sage runs automatically
git push

That's it. Continue to Installation for a detailed setup guide.