[Suggestion] WhatsApp Business voice message transcription with Audio Intelligence (Node.js)

[deepgram-robot]

What to build

A Node.js integration that receives WhatsApp Business voice messages via webhook, transcribes them with Deepgram Nova-3 (with smart formatting and language detection), applies Audio Intelligence (sentiment, entities), and replies with the transcript and a brief summary. Demonstrates the complete webhook → transcribe → respond loop.

Why this matters

WhatsApp has 2B+ users and voice messages are a primary communication mode in many markets. Customer support teams, healthcare providers, and field service organizations receive voice messages they cannot search, archive, or analyze. Developers building WhatsApp integrations need a reference showing how to handle voice message media, transcode OGG/Opus to a Deepgram-compatible format, and return enriched responses — currently no working example exists.

Suggested scope

• Language/framework: Node.js with Express
• Deepgram APIs: Pre-recorded STT (Nova-3), Audio Intelligence (sentiment, entity detection, summarization)
• Integration: WhatsApp Business API (Cloud API via webhooks)
• Key patterns: Webhook verification, media URL download with auth token, audio format handling (OGG Opus), transcription with language detection, reply message formatting
• Complexity: Medium — webhook server + media handling + API calls

Acceptance criteria

• Runnable with minimal setup (clone, add API key + WhatsApp token, run)
• README explains the pattern clearly with WhatsApp Business API setup steps
• Uses current SDK version
• Handles OGG Opus voice messages correctly
• Returns transcript and summary as WhatsApp text reply
• Includes ngrok/tunnel instructions for local development

---

Raised by the DX intelligence system.