A personal AI-model benchmark index: like Artificial Analysis, but the weights are yours. Rank a lineup of models, get benchmark weights fitted to that ranking, then explore the leaderboard, Pareto front and metric gallery. Live at https://mypriors.com. Signed-in users publish a customizable page at mypriors.com/<handle>.
Stack: Next.js, React, TypeScript, Tailwind, ECharts, Convex, Clerk, Vitest, Playwright, and a tsx scraper pipeline.
npm install
npm run dev
It runs without any keys: Clerk and Convex are optional, and the benchmark data is committed in data/snapshots/. Add .env.local (see docs/operations.md) for sign-in, published pages and the mix assistant.
| Command | What it does |
|---|---|
npm run dev |
Dev server |
npm test |
Vitest |
npm run ci |
Route types, tests, tsc, ESLint. The deploy gate |
npm run validate |
ci plus a production build |
BASE=http://localhost:3000 node scripts/verify/smoke.mjs |
Browser check of every route and the onboarding |
npm run scrape -- --date YYYY-MM-DD [--only aa] |
Build a data snapshot |
npx convex dev --once |
Push Convex schema and functions to dev |
Start with AGENTS.md: the map of the code, the rules that must hold, and an index of the topic docs in docs/.