🛸 GitHub AI Radar

How much of the code on GitHub is actually written by AI?

GitHub AI Radar scans any GitHub repository and tells you exactly how much of its development is AI-generated. It analyzes Commits and Pull Requests through a three-layer detection pipeline — bot identity filtering, known AI bot matching, and LLM-powered text style auditing — producing an AI Involvement Index (AII) score from 0% to 100%.

English | 中文

Why?

AI coding assistants like Copilot, Cursor, and Codex are reshaping open-source development at an unprecedented pace. But how deep does the AI involvement really go?

GitHub AI Radar gives you the answer — with data, not guesswork.

• Track AI adoption across popular open-source projects
• Daily automated reports published to GitHub Pages
• Share rankings as beautiful images with QR codes
• Zero-config deployment — one GitHub Action, fully automated

---

How It Works

GitHub Events ──→ L1: System Bot Filter ──→ L2: AI Bot Match ──→ L3: LLM Audit ──→ AII Score
                  (dependabot, etc.)        (copilot[bot], etc.)   (text analysis)

1. L1 — Filters out system bots (CI/CD, dependabot) by username
2. L2 — Identifies known AI coding assistants (Copilot, Codex) by username
3. L3 — Explicit pattern detection (PR AI collaboration mentions, commit Git trailers like Assisted-by) + LLM text style audit
4. AII — Dynamically weighted: only dimensions with actual data receive weight (e.g. a repo with no PRs assigns 100% weight to commits)

Features

• 🔍 Three-layer detection — Static rules + explicit AI pattern matching + LLM text audit
• 📊 Beautiful report site — Podium-style rankings, trend charts, sparklines, GitHub avatars, and shareable images
• ⚡ Batch LLM scoring — 10 events per API call, with retry and concurrency for large repos
• 🔌 Multiple LLM backends — OpenAI, GitHub Models, or any OpenAI-compatible endpoint
• 📦 Event-level cache — Skips LLM calls for unchanged events across runs
• 🛠️ Flexible CLI — Single-item analysis, batch reports, and --force full re-scoring
• 📱 Mobile responsive — Adaptive layout for desktop and mobile devices
• 📸 Share as image — One-click ranking snapshot with QR code, perfect for social media

Quick Start

1. Install

pip install -r requirements.txt

2. Configure

Copy and edit the environment file:

cp .env.example .env

Variable	Required	Description
GITHUB_TOKEN	Recommended	GitHub PAT — without it, API is limited to 60 req/h
LLM_PROVIDER	Optional	LLM backend: none, openai, or github (default: config.toml)
LLM_MODEL	Optional	Model name, e.g. gpt-4.1 (default: config.toml)
OPENAI_API_KEY	Optional	Required when using the OpenAI provider
OPENAI_BASE_URL	Optional	OpenAI-compatible endpoint (Azure, GitHub Models, etc.)

You can also configure repos, LLM settings, and bot lists in config.toml.

3. Run

Analyze a single PR / Commit:

python analyze.py https://github.com/owner/repo/pull/42
python analyze.py owner/repo#123
python analyze.py --no-llm <URL>    # skip LLM, use static rules only

Generate a batch report:

# Analyze repos from config.toml → JSON (with event cache)
python -m report.cli --out reports

# Force re-score all events (ignore cache)
python -m report.cli --force

# Render JSON → static HTML site
python -m report.html --input reports --out site

# Preview locally (open http://localhost:8000)
python -m http.server 8000 -d site

Generate mock data for local development:

python scripts/mock_reports.py       # 7 days of mock reports → reports/
python -m report.html --input reports --out site
python -m http.server 8000 -d site   # preview at http://localhost:8000

Deploy with GitHub Actions

Automatically analyze your repos daily and publish to GitHub Pages — zero maintenance:

1. Fork this repo
2. Add secrets in Settings → Secrets: GH_PAT and optionally OPENAI_API_KEY
3. Set Settings → Pages → Source to Deploy from a branch → gh-pages / / (root)
4. Edit config.toml to add the repositories you want to track
5. Trigger manually or wait for the daily schedule

Reports will be available at https://<user>.github.io/<repo>/

License

MIT