Kitsu — AI Pull Request Summaries via GitHub Actions
Project Snapshot
- Role: Lead Developer
- Duration: ~4 months
- Type: Open-source GitHub Action
- Repository: <https://github.com/Souvikns/kitsu>
Background
Code review velocity often drops when pull requests are large or cross multiple modules. Reviewers spend significant time understanding change context before they can evaluate implementation quality.
Problem Statement
How can teams automatically generate high-quality PR summaries at pull request open time, while letting each team bring its own LLM provider/API key?
Goals
- Trigger automatically when a pull request is opened.
- Generate a readable summary of the code changes for reviewers.
- Use a user-provided LLM API key so teams control provider choice and usage.
- Keep setup simple through standard GitHub Actions workflow configuration.
Constraints
- PR diffs can vary from very small to very large.
- Teams may use different LLM providers and key management practices.
- Summaries must be useful without exposing sensitive repository information.
Solution
Kitsu was built as a GitHub Action that executes on PR open events, inspects the changes, and sends relevant diff/context to an LLM using credentials supplied by the repository owner. It then generates a structured summary to help reviewers quickly understand what changed and why.
Implementation Highlights
- Event-driven execution on pull request open.
- LLM API integration using user-supplied API credentials.
- Automated change summarization to reduce reviewer ramp-up time.
- Workflow-first design that fits into existing CI/review pipelines.
Outcome
- Reduced time required to understand incoming pull requests.
- Improved review consistency by standardizing summary quality across PRs.
- Enabled teams to adopt AI-assisted review workflows without changing their core development process.
Key Learnings
- PR summaries are most effective when they emphasize intent, scope, and risk areas.
- “Bring your own API key” improves flexibility and enterprise adoption.
- Tight event scoping (run on PR open) keeps automation predictable and cost-aware.