Website Audit Tool
An AI-powered tool that scrapes any URL and returns structured SEO, content, and UX insights, with every finding grounded in a real page metric. Live on Vercel.
Problem
Most SEO tools hand you vague advice with no evidence. I wanted audits where every recommendation points back to a concrete metric on the page.
What I built
An AI audit pipeline that scrapes a URL with Cheerio, extracts page metrics, and orchestrates Google Gemini (via OpenRouter) against a constrained JSON schema so results are structured and grounded, a responsible-AI pattern where the model can't fabricate findings. Added exponential-backoff retries on rate limits and a GitHub Actions CI pipeline; deployed live on Vercel.
My role
Sole developer: architecture, scraping, AI orchestration, prompt design, and deployment.
Stack
- Next.js 14
- TypeScript
- Cheerio
- Google Gemini
- OpenRouter
- Tailwind CSS
- Vercel
Impact
- Deployed live on Vercel.
- Every insight cites concrete page data (metric-grounded prompting).
- Exponential-backoff retry handling on LLM rate limits.