Commit & PR Assistant

Generate high-quality commit messages and pull request drafts directly inside VS Code using local Ollama models.

❓ Why Commit & PR Assistant?

⚡ Faster reviews: Turn staged diffs into clean commit messages and PR descriptions in seconds.

🧩 Built-in PR templates: Pick focused formats like Bug Fix, Feature, Security, Performance, and more.

🏠 Local-first AI: Runs entirely on your Ollama setup with models you control.

🔁 GitHub & GitLab ready: Open PR or MR pages with generated titles and paste-ready bodies.

🧠 Native VS Code experience: Works seamlessly with Source Control and the Command Palette.

🔐 No OpenAI API Key Required

This extension does not call OpenAI and does not require an OPENAI_API_KEY.

All prompts are sent to your configured Ollama endpoint (default: http://localhost:11434). When using a local Ollama instance, your code and diffs stay on your machine and never leave your environment 🛡️

🚀 Getting Started with Ollama

1️⃣ Install Ollama

Download from: https://ollama.com/download

2️⃣ Start Ollama

ollama serve

If Ollama is already running as a background service, you can skip this step.

3️⃣ Pull a model

Default model used by this extension:

ollama pull qwen2.5-coder:3b

You can use any model available in your Ollama setup.

4️⃣ Verify Ollama is reachable

curl http://localhost:11434/api/tags

You should see a JSON list of installed models.

5️⃣ Configure the extension in VS Code

Open Settings, search for Commit Assistant, and set:

• ⚙️ commitAssistant.ollamaEndpoint: http://localhost:11434
• 🧠 commitAssistant.model: qwen2.5-coder:3b
• 🌡️ commitAssistant.temperature: 0.3
• 📏 commitAssistant.maxDiffChars: 8000

6️⃣ Available commands

• ✍️ Commit Assist: Suggest Commit Message
• 📝 Commit Assist: Draft Pull Request Description
• 🔗 Commit Assist: Open Pull Request
• 📋 Commit Assist: Select PR Template

🌟 Key Features

✍️ AI Commit Message Generation

• Generates a single, concise commit message from the staged diff.
• Writes directly into the Source Control commit input.
• Offers to stage all changes if nothing is staged.

📝 AI Pull Request Drafting

• Generates a PR title and Markdown body from staged changes.
• Always prompts you to choose a PR template before generation.
• Opens the result in a Markdown document for easy review and edits.

🔗 Open PR / MR Flow

• Detects GitHub or GitLab from your git remote.
• Opens the compare or new MR page with the generated title.
• Copies the generated PR body to your clipboard for quick pasting.

📦 Built-in PR Templates

• General
• Bug Fix 🐛
• Feature ✨
• Refactor 🔧
• Performance ⚡
• Security 🔐
• Tests 🧪
• Chore 🧹
• Documentation 📚

⚡ Quick Start

Step 1: Configure

• Install the extension from the VS Code Marketplace.
• Make sure Ollama is running.
• Set your Ollama endpoint and model in settings.

Step 2: Generate a Commit Message

• Stage your changes.
• Run Commit Assist: Suggest Commit Message.
• The message appears in the SCM commit input.

Step 3: Draft or Open a Pull Request

• Stage your changes.
• Run Commit Assist: Draft Pull Request Description or Commit Assist: Open Pull Request.
• Select a PR template when prompted.

⚙️ Configuration Options

Search for Commit Assistant in VS Code settings.

• commitAssistant.ollamaEndpoint Default: http://localhost:11434
• commitAssistant.model Default: qwen2.5-coder:3b
• commitAssistant.temperature Default: 0.3
• commitAssistant.maxDiffChars Default: 8000

📌 Requirements

• VS Code ^1.109.0
• A Git repository opened in VS Code
• Reachable Ollama endpoint
• At least one pulled Ollama model

🛠️ How It Works

1. Reads the staged diff from the active git repository.
2. Builds a prompt based on the selected template or style.
3. Calls the Ollama /api/generate endpoint.
4. Applies the output to:

• the commit message box
• a Markdown document
• the clipboard
• or the PR / MR URL flow

🧯 Troubleshooting

❌ Ollama request failed (404)

Check commitAssistant.ollamaEndpoint and confirm it points to a running Ollama server.

❌ Ollama request failed (500)

The selected model may not be available. Pull it first:

ollama pull <model>

🤔 Empty or weak output

Try a stronger model or reduce diff size using commitAssistant.maxDiffChars.

⏱️ Connection refused or timeout

Ensure Ollama is running and reachable from VS Code.

🔒 Privacy & Security

• Uses only your configured Ollama endpoint for generation.
• With a local endpoint, code is processed entirely on your machine 🖥️
• Switching to a remote endpoint will send data to that endpoint.
• No telemetry is collected.
• Generated PR bodies are copied only to your local clipboard.

Q&A

What does Commit & PR Assistant do?

It analyzes your staged Git changes and generates clear, well-structured commit messages and pull request descriptions. The goal is to save time and keep your history and PRs easy to review.

Does this extension use OpenAI or any paid API?

No. Everything runs locally using Ollama. There’s no OpenAI, no cloud calls, and no API key required.

What models are supported?

Any model available in your local Ollama setup. By default, it uses qwen2.5-coder:3b, but you can switch to another model in the settings.

Is my code sent anywhere?

No. Your diffs are sent only to your local Ollama instance. Nothing leaves your machine.

How does it generate commit messages?

It reads the staged Git diff, summarizes the intent of the changes, and formats a concise commit message based on what actually changed.

Can it generate pull request descriptions too?

Yes. It can draft a full PR description, including summaries and structured sections, making it easier to open review-ready PRs.

Does it support PR templates?

Yes. You can select from available PR templates, and the assistant will adapt the generated description to match the chosen template.

What Git platforms are supported?

The extension works with Git repositories in VS Code and is compatible with workflows for GitHub, GitLab, and similar platforms.

What if my diff is very large?

You can control how much of the diff is sent to the model using the commitAssistant.maxDiffChars setting. This helps keep responses fast and focused.

Can I customize the AI behavior?

Yes. You can adjust the model name, temperature, and Ollama endpoint directly from the extension settings.

Is this extension free?

Yes. It’s open-source and licensed under MIT.

Who is this extension for?

Anyone who wants cleaner commits and better PRs without relying on cloud AI. It’s especially useful for teams that care about privacy and consistency.

📄 License

MIT