Morning Briefs
"Know what matters today in 10 minutes." A personalized morning news briefing app — one finite, ranked briefing per day instead of an infinite feed.
Layout
web/ — Next.js (TypeScript, Tailwind, App Router) frontend, deployed to Vercelapi/ — FastAPI backend, deployed to Render. Serves pre-built briefings; never calls LLM/TTS at request timepipeline/ — daily batch jobs (ingestion → clustering → ranking → story generation → audio), run on a schedule via GitHub Actions, writing results into Postgres
Architecture
Expensive work (news ingestion, dedup/clustering, LLM summarization, TTS) runs once per day, shared across all users. Personalization happens in a cheap per-user "assembly" step (selecting/ordering shared stories) with no LLM/TTS calls — so cost scales with story volume, not user count.
Stack: Next.js/TS/Tailwind, FastAPI, Postgres (Neon, pgvector), Clerk (auth), Cloudflare R2 (audio storage), Claude (LLM), Google Cloud TTS.
Local development
Frontend
cd web
npm install
npm run dev
Backend
cd api
.venv/Scripts/activate # Windows
pip install -r requirements.txt
uvicorn app.main:app --reload
Copy .env.example → .env (api) and .env.local.example → .env.local (web) and fill in real values as each phase requires them. Never commit real secrets.
