Skip to content

CLAUDE.md

Latest commit

 

History

History
42 lines (35 loc) · 1.63 KB

CLAUDE.md

File metadata and controls

42 lines (35 loc) · 1.63 KB

brain-dump

Voice capture + transcription PWA. Record on phone, read anywhere.

Stack

  • Frontend: React 19 + Vite + TailwindCSS v4 + Zustand + vite-plugin-pwa
  • Backend: Cloudflare Worker (static assets + API in single deploy)
  • Transcription: CF Workers AI Whisper v3 Turbo (with VAD filter)
  • Storage: R2 (audio files) + D1 (transcripts/metadata)

Dev Commands

npm install              # install dependencies
npm run dev              # Vite dev server on :5190
npx wrangler dev         # Worker + D1/R2/AI locally on :8787
npm run build            # tsc + vite build
npm run deploy           # build + wrangler deploy
npm run db:migrate       # apply D1 migrations to remote

Architecture

  • src/ - React frontend (components, pages, lib)
  • src/worker/ - Cloudflare Worker backend (API routes)
  • migrations/ - D1 SQL migrations
  • wrangler.toml - Worker config (D1, R2, AI bindings)

API Endpoints

Method Path Purpose
POST /api/recordings Upload audio + transcribe
GET /api/recordings List recordings
GET /api/recordings/:id Get single recording
PATCH /api/recordings/:id Update transcript/title

Key Details

  • Whisper expects base64-encoded audio (not raw bytes)
  • VAD filter (vad_filter: true) prevents hallucination loops on trailing silence
  • Audio meter uses RMS time-domain analysis for responsive level display
  • MediaRecorder format auto-detection (webm for Chrome, mp4 for Safari)
  • Bottom nav disabled during active recording to prevent state corruption
  • raw_transcript is immutable; edited_transcript is the user-editable copy