distil-commit-bot-ts-Qwen3-0.6B

<div align="center"> <img src="https://github.com/distil-labs/badges/blob/main/distillabs-logo.svg?raw=true" width="40%" alt="distil labs" /> </div>

---

<div align="center"> <table> <tr> <td align="center"> <a href="https://www.distillabs.ai/?utm_source=hugging-face&utm_medium=referral&utm_campaign=distil-commit-bot"> <img src="https://github.com/distil-labs/badges/blob/main/badge-distillabs-home.svg?raw=true" alt="Homepage"/> </a> </td> <td align="center"> <a href="https://github.com/distil-labs"> <img src="https://github.com/distil-labs/badges/blob/main/badge-github.svg?raw=true" alt="GitHub"/> </a> </td> <td align="center"> <a href="https://huggingface.co/distil-labs"> <img src="https://github.com/distil-labs/badges/blob/main/badge-huggingface.svg?raw=true" alt="Hugging Face"/> </a> </td> </tr> <tr> <td align="center"> <a href="https://www.linkedin.com/company/distil-labs/"> <img src="https://github.com/distil-labs/badges/blob/main/badge-linkedin.svg?raw=true" alt="LinkedIn"/> </a> </td> <td align="center"> <a href="https://distil-labs-community.slack.com/join/shared_invite/zt-36zqj87le-i3quWUn2bjErRq22xoE58g"> <img src="https://github.com/distil-labs/badges/blob/main/badge-slack.svg?raw=true" alt="Slack"/> </a> </td> <td align="center"> <a href="https://x.com/distil_labs"> <img src="https://github.com/distil-labs/badges/blob/main/badge-twitter.svg?raw=true" alt="Twitter"/> </a> </td> </tr> </table> </div>

distil-commit-bot TS

We trained an SLM assistant for help with commit messages on TypeScript codebases - Qwen 3 model (0.6B parameters) that you can run locally!

Installation

First, install Ollama, following the instructions on their website.

Then set up the virtual environment:

python -m venv .venv
. .venv/bin/activate
pip install huggingface_hub openai watchdog

or using uv:

uv sync

The model is hosted on huggingface:

• distil-labs/distil-commit-bot-ts-Qwen3-0.6B

Finally, download the models from huggingface and build them locally:

hf download distil-labs/distil-commit-bot-ts-Qwen3-0.6B --local-dir distil-model

cd distil-model
ollama create distil-commit-bot-ts-Qwen3-0.6B -f Modelfile

Run the assistant

The commit bot with diff the git repository provided via --repository option and suggest a commit message. Use the --watch option to re-run the assistant whenever the repository changes.

python bot.py --repository <absolute_or_relative_git_repository_path>
# or
uv run bot.py --repository <absolute_or_relative_git_repository_path>

# Watch for file changes in the repository path:
python bot.py --repository <absolute_or_relative_git_repository_path> --watch
# or
uv run bot.py --repository <absolute_or_relative_git_repository_path> --watch

Examples

See examples

Training & Evaluation

The tuned models were trained using knowledge distillation, leveraging the teacher model GPT-OSS-120B. The data+config+script used for finetuning can be found in data. We used 20 typescript git diff examples (created using distillabs' vibe tuning) as seed data and supplemented them with 10,000 synthetic examples across various typescript use cases (frontend, backend, react etc.).

We compare the teacher model and the student model on 10 held-out test examples using LLM-as-a-judge evaluation:

Evaluation Criteria: LLM-as-a-judge: The training config file and train/test data splits are available under /data.

FAQ

Q: Why don't we just use Llama3.X yB for this??

We focus on small models (< 8B parameters), and these make errors when used out of the box (see 5.)

Q: I want to train a small language model for my use-case

A: Visit our website and reach out to us, we offer custom solutions.