[Suggestion] Genkit (Firebase AI) voice agent with Deepgram STT and TTS (TypeScript)

[deepgram-robot]

What to build

A working example that integrates Google's Genkit (Firebase AI) framework with Deepgram's STT and TTS APIs to build a voice-enabled AI agent. The example should show how to use Genkit's flow system with Deepgram for audio input/output in a conversational agent.

Why this matters

Genkit is Google's open-source AI development framework, growing rapidly in the Firebase/Google Cloud ecosystem. Developers building AI applications with Genkit currently have no reference for adding voice capabilities via Deepgram. A working integration example makes Deepgram discoverable to the large Firebase developer community and provides a production-ready pattern for voice-enabling Genkit flows.

Suggested scope

• Language: TypeScript (Genkit's primary SDK)
• Framework: Genkit with Firebase plugin
• Deepgram APIs: STT (Nova-3 streaming) + TTS (Aura)
• Architecture: Genkit flow that accepts audio input via Deepgram STT, processes through an LLM, and returns audio via Deepgram TTS
• Complexity: Medium — requires Genkit flow definition, audio streaming, and Deepgram SDK integration
• Should include a simple web UI or CLI for testing the voice interaction

Acceptance criteria

• Runnable with minimal setup (clone, add API keys, run)
• README explains the pattern clearly
• Uses current SDK version
• Demonstrates Genkit flow → Deepgram STT → LLM → Deepgram TTS pipeline
• Includes both streaming and batch transcription paths

---

Raised by the DX intelligence system.