This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady
Writing code is exhilarating. Writing Pull Request descriptions, technical changelogs, testing instructions, and release notes is… exhausting.
During Hacktoberfest, developers are eager to ship contributions, but writing thorough PR documentation often turns into an afterthought—or a rushed, one-line summary ("fixed stuff").
CommitSloth 🦥 is an offline-first, privacy-guaranteed CLI assistant that automates developer paperwork using open-weight AI. It analyzes your local git repository (working directory, staged changes, branches, or specific commits), automatically strips noisy lockfiles and binary assets, and prompts a local Google Gemma model via Ollama to generate:
feat(parser): add lockfile filtering and token guardrails)
Every developer who writes code! Instead of spending 15–20 minutes context-switching to write PR descriptions and markdown paperwork, you simply run commitsloth and get coffee. CommitSloth does the slow and steady documentation for you.
CommitSloth runs seamlessly in your terminal with live token streaming and formatted terminal tables:
python commitsloth.py --stream
python commitsloth.py -o PULL_REQUEST.md
## 📋 Changelog & Key Changes
- **`commitsloth/git_parser.py`**: Added safe subprocess invocation with automatic lockfile and artifact filtering.
- **`commitsloth/agent_core.py`**: Implemented local Ollama loopback client with zero-cloud-leakage guardrails.
- **`commitsloth/cli.py`**: Created terminal interface with Rich styling and Markdown file export.
## 🧪 Testing Instructions
1. Run automated test suite:
``` bash
pytest tests/ -v
- Verify diff extraction on local changes:
python commitsloth.py --diff-only - Test local Ollama endpoint connection:
python commitsloth.py --mock
🚀 Release Summary #
CommitSloth provides a 100% offline developer productivity CLI that automates Pull Request and release notes generation using local Google Gemma models.
**"Slow and Steady" developer paperwork using local, open-weight AI.**
*Built for the Hacktoberfest 2026 Open-Source AI Challenge.*
Developers love writing code, but writing Pull Request descriptions, changelogs, and release paperwork is tedious and easy to rush. During Hacktoberfest's **"Slow and Steady"** week, CommitSloth encourages developers to take a breath, slow down, and let a thoughtful local AI assistant craft professional documentation—**without sending proprietary code to the cloud.**
**CommitSloth** runs entirely on your machine. It analyzes your local git repository, filters out irrelevant build noise and lockfiles, and prompts Google's open-weight **Gemma** model via **Ollama** (`http://localhost:11434`) to produce:
`feat(auth): implement offline token verification`)
flowchart TD subgraph Local Repository
…
CommitSloth is written in Python (3.10+) and architected into modular components:
CommitSloth Architecture: [Git Working Tree] ──> [Safe Subprocess] ──> [GitParser & Noise Filter] │ ▼ [Rich Terminal / Output] <── [AgentCore] <── [Gemma Prompt Constructor] │ (Local Loopback HTTP) ▼ [Ollama Daemon: localhost:11434] │ [Google Gemma (2B / 7B)]
`gemma:2b` (ultra-fast for quick local loops) and `gemma:7b` (high-depth analysis) running locally via `AgentCore` validates that every outbound request resolves strictly to a loopback interface (`localhost`, `127.0.0.1`, `::1`). If an external URL is passed, it halts execution immediately with a `SecurityViolationError`.` package-lock.json` or `Cargo.lock`) choke local LLM context windows. CommitSloth's `GitParser` automatically strips lockfiles, minified assets (`*.min.js`), and binaries, compressing the git diff down to meaningful code logic within a 14,000-character budget.`rich` for live progress spinners, diff statistics tables (`+` additions / `-` deletions), and real-time token streaming.
Open innovation is what made CommitSloth possible.
With closed, proprietary APIs:
Open-weight AI turns the developer workstation into an autonomous, private documentation engine.
Developing CommitSloth involved iterative prompt engineering, subprocess security testing, test-driven validation with `pytest`, and lint formatting. You can review the development session transcript and workflow via the [CommitSloth GitHub repository](https://github.com/Babin123456/CommitSloth).
`python commitsloth.py -o PULL_REQUEST.md`).
Reviewers get clean, Conventional-Commits-compliant PRs with actionable reproduction steps, and authors never have to type `"fixes and updates"` as a PR title ever again.
Developer documentation shouldn't be an afterthought or a rushed chore. The spirit of **"Slow and Steady"** isn't about sluggish progress—it's about taking pride in thoughtful engineering while letting local, privacy-first AI take care of the repetitive paperwork.
If you are participating in Hacktoberfest or simply want cleaner, more descriptive Pull Requests without sending proprietary diffs to third-party clouds, give **CommitSloth** a spin:
git clone https://github.com/Babin123456/CommitSloth.git cd CommitSloth pip install -r requirements.txt python commitsloth.py
Slow down, write great code, and let the sloth handle the paperwork. 🦥✨
*Feedback, ideas, and contributions are warmly welcome on [GitHub](https://github.com/Babin123456/CommitSloth). Happy Hacktoberfest!*