AutoReviewBot: AI-Powered Pull Request Reviewer
AutoReviewBot is an automated GitHub bot that reviews Pull Requests (PRs) using static code analyzers and Machine Learning (ML). It posts GitHub comments highlighting potential code issues, prioritizes them using AI.
- Runs static code analyzers: Checkstyle, PMD, SpotBugs
- Uses ML to prioritize violations based on developer feedback
- Posts inline comments on GitHub PRs using GitHub API
- Optional LLM summarization for grouped feedback
- Tracks metrics (Precision, Recall) and logs feedback
- Supports adaptive rule weights with
weights.yaml
| Component | Tool/Library |
|---|---|
| Language | Python 3 |
| Static Analysis | Checkstyle, PMD, SpotBugs |
| ML Ranking | LightGBM + TF-IDF + Scikit-learn |
| LLM Summarization | OpenAI GPT (optional) |
| GitHub Integration | REST API + Webhooks |
| Hosting | Ngrok |
| CI/CD Support | GitHub Actions |
AI & ML Integration Model: LightGBM classifier with TF-IDF features from violation messages.
Inputs:
Violation message (text)
Severity (low, medium, high, critical)
Tool name (checkstyle, pmd, spotbugs)
Output: Ranking score → used to prioritize comment importance.
Feedback Learning: Developer responses update feedback_log.csv, improving future predictions.
How It Works (Step-by-Step) GitHub Webhook Triggered: When a PR is opened or updated.
Bot Clones Repo: Uses GitHub token to clone PR branch.
Run Analyzers: Executes Checkstyle, PMD, SpotBugs on the codebase.
Parse Results: Extracts file, line, message, severity, and tool used.
Rank Violations: ML model scores importance.
Comment on PR:
Inline comments
Summary comment with total/critical issues
LLM Summary : Uses GPT to group and summarize violations.
Merge Blocking: Blocks if any critical issue is found.
Feedback Logging: Stores whether developer accepted/ignored suggestions.
Model Retraining: Uses feedback log to improve future rankings.