🧠 LLM Git Commit Message Generator

A Python-based tool that leverages a local LLM (via Ollama) to generate clean, context-aware commit messages for both git and yadm.

---

💡 What It Does

This tool inspects staged changes and sends either:

• A summary of file names (for large commits), or
• A diff or full content summary (for small changes),

to a locally running LLM model using Ollama, and generates a meaningful, concise commit message.

Structured output is parsed using pydantic models to keep LLM "thinking" out of the final result.

---

🚀 Why It Exists

Conventional commit message generators are either:

• Dumb keyword scanners, or
• Cloud-based LLMs (privacy issue).

This project offers:

• Full local execution with Ollama
• Intelligent message structuring
• Support for both git and yadm
• Experimental file-wise summarization for complex commits

---

🧱 Project Structure

.
├── main.py                # CLI entrypoint
└── helper/
    ├── ollama_client_direct.py
    ├── ollama_client_api.py
    ├── commit_schema.py
    └── __init__.py

---

⚙️ Flags & CLI Usage

python main.py [options]

🧾 Supported Flags

Flag	Description
--git	Use Git repository (default if .git is found)
--yadm	Use YADM (assumes global config)
--path /repo/path	Path to Git repo (optional, used with --git)
--smallchange	Use full diff/content for 1-2 changed files
--bigchange	Use filenames only for large commits
--experimentalSum	Generate summaries of each file then merge via LLM
--use-api	Use requests instead of ollama SDK
--help	Show help and usage

---

📥 Setting Up Ollama

Ensure you have Ollama installed and the desired model pulled.

🛠️ Install Ollama

Follow instructions at: https://ollama.com

⬇️ Pull the gemma Model

ollama pull gemma:latest

---

🧪 Example

Simulated File Change

Modified: config.yaml
+ added GPU acceleration toggle
+ changed log level to DEBUG

Modified: server.py
+ added async support to request handler

Generated Commit Message

Enable GPU toggle and add async request support

Explanation (optional if --debug enabled):

This commit enables GPU support via config.yaml and introduces async request handling in server.py.

---

🔧 How to Use in a Git Workflow

This tool prints a commit message to stdout.

🧼 Basic Usage

msg=$(python main.py --git --smallchange)
git commit -m "$msg"

💾 With YADM

msg=$(python main.py --yadm --bigchange)
yadm commit -m "$msg"

🧪 With Experimental Summary Mode

msg=$(python main.py --git --experimentalSum)
git commit -m "$msg"

You can wrap this into an alias or shell script for convenience.

---

🔮 Future Scope

• ✅ Add debug/verbose flags to print LLM reasoning
• ✅ Add model selection via flag (e.g. --model llama3)
• ✅ Add fallback logic between SDK and API
• 🚧 Create a VS Code extension that:
  • Hooks into the Source Control panel
  • Adds a "Generate Commit Message with LLM" button
  • Uses the local Ollama API for instant commit suggestions
• 🚧 Allow config file (.llmcommitrc) to define default flags, model, temperature
• 🚧 Add support for pre-commit hooks integration

---

🧑‍💻 Author

Built by an engineer tired of writing garbage commit messages. Make machines write them better.

---