git-copilot-commit

AI-powered Git commit assistant that generates conventional commit messages using GitHub Copilot or any OpenAI-compatible LLM.

Features

• Generates commit messages based on your staged changes
• Supports GitHub Copilot and OpenAI-compatible /v1/chat/completions, /v1/responses, and /v1/models endpoints
• Supports multiple LLM models: GPT, Claude, Gemini, local models, and more
• Allows editing of generated messages before committing
• Follows the Conventional Commits standard

Installation

Install the tool using uv

Install uv:

# On macOS and Linux.
curl -LsSf https://astral.sh/uv/install.sh | sh

# On Windows.
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

You can run the latest version of tool directly every time by invoking this one command:

# Every invocation installs latest version into temporary environment and runs --help
uvx git-copilot-commit --help

Alternatively, you can install the tool once into a global isolated environment and run git-copilot-commit to invoke it:

# Install into global isolated environment
uv tool install git-copilot-commit

# Run --help to see available commands
git-copilot-commit --help

Prerequisites

• Either an active GitHub Copilot subscription or access to an OpenAI-compatible API endpoint

Quick Start

GitHub Copilot

1. Authenticate with GitHub Copilot:

   uvx git-copilot-commit authenticate
   

   If your cached GitHub token is revoked or expires, refresh it with:

   uvx git-copilot-commit authenticate --force
   

2. Make changes in your repository.

3. Generate and commit:

   uvx git-copilot-commit commit
   # Or, if you want to stage all files and accept the generated commit message, use:
   uvx git-copilot-commit commit --all --yes
   

OpenAI-compatible provider

1. List models by pointing the CLI at your server's /models endpoint.

   uvx git-copilot-commit models \
     --provider openai \
     --base-url http://127.0.0.1:11434/v1/models
   

2. Generate and commit by pointing the CLI at the generation endpoint you want to use.

   uvx git-copilot-commit commit \
     --provider openai \
     --base-url http://127.0.0.1:11434/v1/chat/completions \
     --model your-model-id
   

   If your server requires an API key, also pass --api-key ... or set OPENAI_API_KEY.

   OpenAI-compatible generation URLs must end with /chat/completions or /responses. Model listing URLs must end with /models.

Usage

Commit changes

$ uvx git-copilot-commit commit --help

 Usage: git-copilot-commit commit [ARGS]

 Generate commit message based on changes in the current git repository and commit them.

╭─ Parameters ─────────────────────────────────────────────────────────────────╮
│ ALL --all -a --no-all     Stage all files before committing [default: False] │
│ SPLIT --split --no-split  Split staged hunks into multiple commits           │
│                           automatically. Pass --split=N to express a         │
│                           preference for N commits. [default: False]         │
│ MODEL --model -m          Model to use for generating commit message         │
│ YES --yes -y --no-yes     Automatically accept the generated commit message  │
│                           [default: False]                                   │
│ CONTEXT --context -c      Optional user-provided context to guide commit     │
│                           message [default: ""]                              │
│ DISABLE-THINKING          Disable or minimize reasoning/thinking tokens for  │
│   --disable-thinking      commit-message requests. [default: True]           │
│   --enable-thinking                                                          │
│ MAX-TOKENS --max-tokens   Maximum output tokens for LLM generation.          │
│                           [default: 1024]                                    │
│ PROVIDER --provider       LLM provider to use: copilot or openai.            │
│ BASE-URL --base-url       Endpoint URL for an OpenAI-compatible provider,    │
│                           for example                                        │
│                           http://127.0.0.1:11434/v1/chat/completions.        │
│ API-KEY --api-key         API key for an OpenAI-compatible provider. Omit    │
│                           when the server does not require one.              │
│ CA-BUNDLE --ca-bundle     Path to a custom CA bundle (PEM)                   │
│ INSECURE --insecure       Disable SSL certificate verification. [default:    │
│   --no-insecure           False]                                             │
│ NATIVE-TLS --native-tls   Use the OS's native certificate store via          │
│   --no-native-tls         'truststore' for httpx instead of the Python       │
│                           bundle. Ignored if --ca-bundle or --insecure is    │
│                           used. [default: True]                              │
╰──────────────────────────────────────────────────────────────────────────────╯

Examples

Commit all changes:

uvx git-copilot-commit commit --all

Accept the generated commit message without editing:

uvx git-copilot-commit commit --yes

Use a specific model:

uvx git-copilot-commit commit --model claude-3.5-sonnet

Use a local OpenAI-compatible server:

uvx git-copilot-commit commit \
  --provider openai \
  --base-url http://127.0.0.1:11434/v1/chat/completions \
  --model your-model-id

Example with openai/gpt-oss-120b and Qwen/Qwen3.6-35B-A3B:

uvx git-copilot-commit commit \
  --provider openai \
  --base-url http://example.com:8001/v1/chat/completions \
  --model openai/gpt-oss-120b

uvx git-copilot-commit commit \
  --provider openai \
  --base-url http://example.com:8002/v1/chat/completions \
  --model Qwen/Qwen3.6-35B-A3B

Use the Responses API endpoint:

uvx git-copilot-commit commit \
  --provider openai \
  --base-url http://example.com:8002/v1/responses \
  --model your-model-id

Increase the output token budget:

uvx git-copilot-commit commit --max-tokens 4096

Thinking/reasoning is disabled or minimized by default for commit-message requests. To let the selected model use its default thinking behavior, pass:

uvx git-copilot-commit commit --enable-thinking

TLS uses the operating system's native certificate store by default. To use Python's default certificate bundle instead, pass:

uvx git-copilot-commit commit --no-native-tls

Split staged hunks into separate commits:

uvx git-copilot-commit commit --split

Prefer two commits:

uvx git-copilot-commit commit --split 2

Commit Message Format

Follows Conventional Commits:

<type>[optional scope]: <description>

Types:

• feat: New feature
• fix: Bug fix
• docs: Documentation
• style: Formatting only
• refactor: Code restructure
• perf: Performance
• test: Tests
• chore: Maintenance
• revert: Revert changes

Git Configuration

Add a git alias by adding the following to your ~/.gitconfig:

[alias]
    ai-commit = "!f() { uvx git-copilot-commit commit $@; }; f"

Now you can run to review the message before committing:

git ai-commit

Alternatively, you can stage all files and auto accept the commit message and specify which model should be used to generate the commit in one CLI invocation.

git ai-commit --all --yes --model claude-3.5-sonnet

You can also set provider defaults with environment variables:

export GIT_COPILOT_COMMIT_PROVIDER=openai
export GIT_COPILOT_COMMIT_BASE_URL=http://127.0.0.1:11434/v1/chat/completions
export GIT_COPILOT_COMMIT_API_KEY=...
export OPENAI_API_KEY=...
git ai-commit --provider openai --model your-model-id

For example:

export GIT_COPILOT_COMMIT_PROVIDER=openai
export GIT_COPILOT_COMMIT_BASE_URL=http://example.com:8001/v1/chat/completions
git ai-commit --model openai/gpt-oss-120b