git-contribution-analyzer

A high-performance tool to analyze GitHub Pull Requests by author, generating comprehensive summaries with LLM-powered insights.

Features

• Parallel batch processing for faster PR analysis
• Safe git operations with mutex-based locking
• Full PR diffs with automatic fallback for large PRs
• Auto-numbered output files
• Robust error handling
• LLM-powered analysis using Google Gemini
• Retry mechanism for failed analyses

Prerequisites

• Bun runtime
• GitHub CLI (authenticated)
• Git
• Google Gemini API key

Installation

bun install
gh auth login
cp .env.example .env
# Add your GEMINI_API_KEY to .env

Usage

Step 1: Fetch PRs and Generate Diffs

bun run src/analyzePR.ts --repo OWNER/REPO --author USERNAME --limit 50

Options:

• --repo (required): Repository in OWNER/REPO format
• --author (required): GitHub username
• --limit (optional): Number of PRs to analyze (default: 10)
• --batch-size (optional): Parallel processing batch size (default: 20)

Example:

bun run src/analyzePR.ts --repo OWNER/REPO --author USERNAME --limit 100

Step 2: Run LLM Analysis

bun run src/llmRunner.ts ./cache/report.json

Analyzes PRs from cache and generates comprehensive reports.

Step 3: Retry Failed Analyses (Optional)

bun run src/llmRunner.ts retry ./reports/1_pr-analysis-report.json ./cache/report.json

Output

PR Reports: cache/{N}_{author}_report.json

• Contains PR metadata and diff file references

Diff Files: cache/diff/{N}_{repo}_pr{number}_diff.patch

• Complete PR changes in git patch format

Analysis Reports: reports/{N}_pr-analysis-report.json

• LLM-generated insights and summaries

Customization

Edit src/prompt.ts to customize LLM analysis prompts.