Adrian Wedd - AI-Enhanced CV

A professional CV website driven by real GitHub activity evidence, with AI-proposed content improvements that are verified before they can touch the CV. Runs on GitHub Actions and deploys to GitHub Pages.

Live site: cv.adrianwedd.com

How It Works

1. Activity Tracker (activity-tracker.yml) runs daily, collecting cross-repo GitHub activity evidence via activity-collector.js. It commits only on material change, and then triggers the enhancement pipeline via repository_dispatch
2. CV Enhancement Pipeline (cv-enhancement.yml) runs weekly (Mondays 21:00 UTC), on evidence dispatch, or manually. Stages: ingest → propose (enhance.js writes proposals to data/ai-enhancements.json) → verify (verify-proposals.js applies only evidence-supported proposals to base-cv.json) → validate → render (cv-generator.js) → publish
3. Validation gate blocks deployment if the hallucination detector or content guardian finds issues

Project Structure

cv/
├── index.html                    # Main CV webpage
├── watch-me-work.html            # Redirect stub → adrianwedd.com/activity/ (dashboard migrated)
├── assets/
│   ├── styles.css                # CSS with dark/light themes
│   └── script.js                 # Interactive features
├── data/
│   └── base-cv.json              # Core CV data
├── .github/
│   ├── workflows/
│   │   ├── cv-enhancement.yml    # AI enhancement pipeline
│   │   └── activity-tracker.yml  # GitHub activity collection
│   └── scripts/
│       ├── activity-collector.js # Cross-repo GitHub evidence collection
│       ├── ai/client.js          # Provider-neutral AI client (openrouter/ollama/gemini)
│       ├── enhance.js            # AI proposal stage
│       ├── verify-proposals.js   # Proposal verification + apply stage
│       ├── cv-generator.js       # Website generator
│       ├── ai-hallucination-detector.js  # Content validation
│       └── content-guardian.js   # Prevents fabricated claims
└── package.json

Local Development

# Serve locally
python3 -m http.server 8000
# or
npx serve .

# Validate JSON data
npm run validate:json

# Run pipeline scripts
cd .github/scripts && npm install
node activity-collector.js # requires GITHUB_TOKEN
node enhance.js            # requires OPENROUTER_API_KEY (or GEMINI_API_KEY / OLLAMA_HOST)
node verify-proposals.js
node cv-generator.js

Configuration

GitHub secrets:

• OPENROUTER_API_KEY - default AI provider (optional: without it the enhancement stage is skipped and the site still builds from curated content)
• GEMINI_API_KEY - alternate AI provider
• PERSONAL_ACCESS_TOKEN - lets the activity tracker's repository_dispatch start the enhancement workflow
• GITHUB_TOKEN - provided automatically by GitHub Actions

Content Integrity

The system includes safeguards against AI hallucination:

• Proposal verification (verify-proposals.js) rejects AI proposals with unsupported numbers, new credentials, corporate filler, or drastic length changes before they can reach base-cv.json
• Hallucination detector validates claims against actual GitHub metrics
• Content guardian maintains a registry of verified claims and blocks fabricated ones
• The detector and guardian run as gates in the CI pipeline - failures block deployment

License

MIT