[Suggestion] Pipecat WhatsApp real-time voice calling with Deepgram STT (Python)

[deepgram-robot]

What to build

A working example showing how to build a real-time voice agent that makes and receives voice calls over WhatsApp using Pipecat's new WhatsApp transport with Deepgram as the STT provider.

Why this matters

WhatsApp has over 2 billion users globally and is the primary business communication channel in India, Brazil, and Southeast Asia. Developers building voice AI for these markets need a clear path from "Deepgram API key" to "working WhatsApp voice agent." Pipecat's new bidirectional WhatsApp voice transport already uses Deepgram STT as the default provider — this example captures that momentum and shows developers how to build on it.

Suggested scope

• Language: Python
• Framework: Pipecat (v1.1.0+) with small-webrtc-transport and WhatsApp Cloud API
• Deepgram APIs: STT (Nova-3 streaming), TTS (Aura)
• What it should include:
  • Inbound WhatsApp voice call handling via webhook
  • Outbound WhatsApp voice call initiation within the 7-day callback window
  • Deepgram STT for real-time transcription
  • Deepgram TTS for voice responses
  • A simple LLM-powered conversational agent
  • WhatsApp Cloud API authentication setup
• Complexity: Medium — requires Meta Business Account and WhatsApp Cloud API credentials

Acceptance criteria

• Runnable with minimal setup (clone, add API keys, run)
• README explains the WhatsApp Cloud API setup clearly
• Uses current Pipecat and Deepgram SDK versions
• Handles both inbound and outbound call flows
• Includes error handling for WhatsApp webhook edge cases

---

Raised by the DX intelligence system.