Ollama Commit

Ollama Commit cover

Generate Git commit messages in VS Code with Ollama, with Groq, Gemini, and Codex/OpenAI fallback support.

Ollama Commit is a lightweight VS Code extension that reads your Git diff, sends it to your local or remote Ollama server, and writes the generated commit message into the Source Control input box for you. If Ollama is unavailable, it can fall back to local Codex CLI first, then Groq, then Gemini.

Why Use It

Write commit messages faster
Keep commit messages consistent
Use your own Ollama model
Stay local when running Ollama on your machine

Features

Generate commit messages from staged changes
Fall back to Groq, Gemini, and Codex/OpenAI when Ollama is unavailable
Fall back gracefully when nothing is staged
Add a Generate Commit Message button to the Source Control toolbar
Open a dedicated settings page inside VS Code
Load available models directly from Ollama
Optionally copy the generated message to the clipboard
Handle common Windows and WSL networking cases

Quick Start

Install and start Ollama.
Start the model you want to use.
Install this extension in VS Code.
Open a Git repository.
Stage your changes if needed.
Open Source Control and click Generate Commit Message.

Optional fallback setup:

Create a Groq API key and set ollamacommit.groqApiKey.
Create a Gemini API key and set ollamacommit.geminiApiKey.
Optionally keep a local Codex login on the same machine so the extension can use codex exec as a final fallback.
Keep Ollama as the primary provider. The extension will try Codex first, Groq second, and Gemini third if Ollama cannot be reached.

Example:

ollama serve
ollama run minimax-m2.7:cloud

Then set ollamacommit.model to minimax-m2.7:cloud in the extension settings.

Default Ollama URL:

http://127.0.0.1:11434

How It Works

The extension reads your staged diff.
If no files are staged, it lets you choose between:
- Stage All and Generate
- Use Unstaged Changes
The diff is trimmed to the configured max size.
The extension sends the diff to Ollama.
If Ollama is unavailable, the extension tries local Codex CLI, then Groq, then Gemini.
The active provider returns a commit message.
The result is inserted into the VS Code Source Control commit box.

If VS Code cannot insert the message into Source Control directly, the extension copies it to your clipboard instead.

Commands

Ollama Commit: Generate Commit Message
Ollama Commit: Open Settings Page

Settings

Open the built-in settings page with:

Ollama Commit: Open Settings Page

Or search for Ollama Commit in VS Code Settings.

Setting	Description	Default
`ollamacommit.baseUrl`	Ollama server URL	`http://127.0.0.1:11434`
`ollamacommit.model`	Model used to generate commit messages. For this setup, use `minimax-m2.7:cloud`.	`qwen2.5-coder:7b`
`ollamacommit.groqApiKey`	Groq API key used when Ollama is unavailable	`""`
`ollamacommit.groqModel`	Groq model used as the first fallback	`openai/gpt-oss-20b`
`ollamacommit.geminiApiKey`	Gemini API key used when Ollama and Groq are unavailable	`""`
`ollamacommit.geminiModel`	Gemini model used as the second fallback	`gemini-2.0-flash-lite`
`ollamacommit.openaiModel`	Optional model override for the last fallback through local Codex CLI. Leave empty to use Codex's configured default model	`""`
`ollamacommit.codexPath`	Optional absolute path to the Codex CLI binary when the extension host PATH does not include it	`""`
`ollamacommit.systemPrompt`	System prompt sent before the diff	Built-in prompt
`ollamacommit.enableThinking`	Allow model thinking mode before the final answer	`false`
`ollamacommit.ollamaUnavailableCooldownMs`	After an Ollama connection failure, skip retrying Ollama for this many milliseconds and jump straight to fallbacks. Each request still does a quick hidden probe and re-enables Ollama immediately if it comes back	`172800000`
`ollamacommit.maxDiffChars`	Maximum diff characters sent to Ollama	`12000`
`ollamacommit.temperature`	Sampling temperature	`0.2`
`ollamacommit.copyToClipboard`	Also copy the result to the clipboard	`false`

Requirements

VS Code 1.90.0 or newer
Git installed and available in your environment
Ollama installed and running
At least one Ollama model available to run

Install From Source

npm install
npm run build

Run in Extension Development Host:

Open this repository in VS Code.
Press F5.

Build a VSIX package:

npm run package

Install the generated .vsix from VS Code with Extensions: Install from VSIX...

WSL Notes

If you use VS Code with Remote - WSL, 127.0.0.1 may point to WSL instead of the Windows host running Ollama.

Ollama Commit tries your configured baseUrl first and then attempts common WSL fallback host addresses automatically.

Privacy

The extension sends your Git diff to the active provider:

Ollama through ollamacommit.baseUrl
Groq when ollamacommit.groqApiKey is configured
Gemini when ollamacommit.geminiApiKey is configured
Codex CLI when you are logged in locally and codex exec is available in the extension host environment

If you run Ollama locally, your data stays in your local environment. Cloud fallbacks send the diff to the selected provider, and the Codex fallback sends the diff through your local Codex CLI session.

License

MIT. See LICENSE.

Ollama Commit Message

Nice0w0