WhatsApp Voice Transcription

Cloudflare Worker that transcribes WhatsApp voice messages using Workers AI (Whisper) and replies with the transcript. Optional AI summary via GENERATE_SUMMARY.

Setup

1. Meta / WhatsApp

   • Create a Meta app and add the WhatsApp product.
   • Get your Phone Number ID and Access Token.
   • Configure a webhook URL: https://<your-worker>.<account>.workers.dev/webhook.
   • Set a verify token (any string you choose).

2. Secrets (required):

   npx wrangler secret put WEBHOOK_SECRET      # same as verify token in Meta
   npx wrangler secret put WHATSAPP_PHONE_NUMBER_ID
   npx wrangler secret put WHATSAPP_ACCESS_TOKEN

3. Optional

   • WHITELISTED_PHONE_NUMBERS — comma-separated numbers; if set, only these receive transcript replies. Omit to allow all.
   • In wrangler.jsonc vars: GENERATE_SUMMARY (true/false), WORKERS_AI_MODEL, WORKERS_AI_SUMMARY_MODEL.

Deploy

npm install
npm run deploy

Use the worker URL as your WhatsApp webhook and for verification.

Endpoints

• POST /webhook — WhatsApp webhook (receives events).
• GET /webhook — Verification (Meta sends hub.mode, hub.verify_token, hub.challenge).
• GET /health — Returns OK.