GitHub PR & Commit Analyzer

中文文档 | English

A powerful command-line tool for intelligently collecting, analyzing, and summarizing GitHub Pull Requests and commit records.

🚀 Installation & Setup

1. Prerequisites

• Python 3.8+
• Git
• GitHub CLI (gh): Must be logged in (gh auth login)

2. Install

pip install github-pr-analyzer

3. Configuration (Environment Variables)

# Example
export CURSOR_AGENT_PATH=/path/to/cursor-agent
export DEFAULT_MONTHS=6

📖 Quick Start

# 1. Interactive Mode (Best for starting)
ghpa interactive

# 2. Search with AI Analysis (English output)
ghpa search "authentication bug" --analyze

# 3. Search with AI Analysis (Chinese output)
ghpa search "authentication bug" --analyze -cn

# 4. Collect Data
ghpa collect --save-json

# 5. Generate Daily/Weekly Report + Export Datasets
ghpa traverse --days 7 --save-json
ghpa traverse -r pytorch/pytorch --days 7 --save-json -cn

All CLI workflows expose matching --save-json / --no-save-json toggles so you can enable exports when needed and opt out (for example, disable the default view-pr export with --no-save-json).

Language Options

Use -cn or --chinese flag to get AI analysis output in Chinese:

# Chinese output
ghpa search "quantization" -a -cn
ghpa view-pr 588 -a -cn
ghpa traverse --days 7 -cn

✨ Features

• 🔍 Smart Search: AI-powered keyword extraction
• 📊 Data Collection: PRs and merge commits statistics
• 🔄 Diff Viewing: Syntax-highlighted code changes
• 🤖 AI Analysis: Summarization via cursor-agent with English/Chinese output
• 📅 Traverse Mode: Batch analysis for reporting
• 🗂 JSON Export: Persist PR, commit, and review conversations as structured JSON
• 🌐 Multi-language: Support for English and Chinese AI analysis output
• 💾 Instant Save: JSON files saved immediately after each PR analysis
• 🎨 Enhanced Display: Beautiful terminal output with colors and formatting

For detailed command usage, see USAGE.md.

🗂 JSON Export Format

All major workflows (collect, search, traverse, view-pr) share the same --save-json / --no-save-json flags (with view-pr defaulting to export unless --no-save-json is specified). Files land in gh_pr_exports/ unless --output-dir is provided.

File Naming Convention

Files follow the pattern:

{repo}_{merged_pr|open_pr}_{pr_number}_{title}_{timestamp}.json

Examples:

• NVIDIA_TensorRT_merged_pr_588_Support_AutoQuantize_20251125_1423.json
• pytorch_pytorch_open_pr_123_Add_feature_20251126_0930.json

The timestamp uses the merge time for merged PRs, or the last update time for open PRs.

JSON Structure

Each JSON document contains:

• repo: the owner/repo slug used for collection
• pr: metadata including:
  • number, title, url, state
  • author, author_name, author_email
  • created_at, updated_at, merged_at
  • base_ref, head_ref
  • body: PR description as line array (for readability)
• commits: ordered list of commits with:
  • id: full commit SHA
  • title: first line of the commit message
  • message: full commit body
  • committer_name, committer_email
  • files: array of { "path": "<file>", "diff": ["line1", "line2", ...] }
• conversation: threaded review data
  • issue_comments: top-level PR discussion
  • review_threads: code review threads (note: limited availability via GitHub API)
  • reviews: review summaries (approve/comment/request changes)

Example snippet:

{
  "repo": "octo-org/octo-repo",
  "pr": {
    "number": 42,
    "title": "Fix login",
    "author": "octocat",
    "author_name": "The Octocat",
    "author_email": "octocat@github.com",
    "state": "MERGED",
    "body": [
      "## Summary",
      "This PR fixes the login flow.",
      "",
      "## Changes",
      "- Fixed authentication"
    ]
  },
  "commits": [
    {
      "id": "abc123...",
      "title": "Adjust auth flow",
      "message": "Adjust auth flow\n\n- add checks...\n",
      "committer_name": "The Octocat",
      "committer_email": "octocat@github.com",
      "files": [
        {
          "path": "auth/login.py",
          "diff": [
            "@@ -1,3 +1,4 @@",
            " import os",
            "-old_line",
            "+new_line"
          ]
        }
      ]
    }
  ],
  "conversation": {
    "issue_comments": [],
    "review_threads": [],
    "reviews": []
  }
}

Use these exports to feed downstream tooling, audits, or offline review workflows.

🎨 AI Analysis Display

The AI analysis output features:

• Metadata Panel: Shows author, email, state, URL with color-coded status
• Analysis Panel: Dark background with white text for readability
• Left-aligned Content: All text left-aligned for better reading experience
• Adaptive Width: Automatically adjusts to terminal window size

State indicators:

• ✅ MERGED (magenta)
• 🔄 OPEN (green)
• ❌ CLOSED (red)