[Suggestion] Phoenix LiveView real-time streaming transcription (Elixir)

[deepgram-robot]

What to build

An Elixir/Phoenix LiveView example that streams microphone audio to Deepgram's real-time STT API and displays live transcription results, leveraging LiveView's built-in WebSocket infrastructure for a seamless real-time experience with zero custom JavaScript.

Why this matters

Phoenix LiveView is purpose-built for real-time web applications and has a rapidly growing developer community. Its server-rendered model with bidirectional WebSocket communication is an ideal architectural fit for streaming transcription — the LiveView socket can proxy audio to Deepgram and push transcript updates back to the client with minimal JavaScript. No speech AI provider has a Phoenix/Elixir example, making this a first-mover opportunity that would attract attention in the Elixir community.

Suggested scope

• Language/framework: Elixir, Phoenix 1.7+, LiveView
• Deepgram APIs: Streaming STT (WebSocket)
• Features: Client-side mic capture (JS hook), audio streaming through LiveView socket to server, server-side Deepgram WebSocket connection via GenServer, live transcript display with LiveView updates
• Structure: LiveView page with a small JS hook for getUserMedia, GenServer managing the Deepgram WebSocket, PubSub for transcript distribution
• Complexity: Moderate — demonstrates Elixir's concurrency model for audio pipeline management

Acceptance criteria

• Runnable with minimal setup (clone, add API key, mix phx.server)
• README explains the pattern clearly
• Uses direct WebSocket connection to Deepgram API (no SDK required for Elixir)
• Leverages OTP patterns (GenServer/Task for WebSocket management)
• Handles connection lifecycle and cleanup properly
• Minimal JavaScript — only what's needed for microphone capture

---

Raised by the DX intelligence system.