[Suggestion] FastHTML real-time streaming transcription with WebSocket (Python)

[deepgram-robot]

What to build

A real-time audio transcription web app built with FastHTML that streams microphone audio to Deepgram's STT API via WebSocket and displays live transcription results — all in pure Python with no JavaScript framework.

Why this matters

FastHTML is a rapidly growing Python web framework that lets developers build interactive web apps entirely in Python using HTMX for reactivity. It has a large and enthusiastic community of Python developers who prefer to avoid JavaScript toolchains. A Deepgram + FastHTML example gives these developers a native way to build voice-powered apps in their preferred stack, and demonstrates that real-time streaming transcription is achievable without a JavaScript frontend framework.

Suggested scope

• Language: Python
• Framework: FastHTML + HTMX
• Deepgram APIs: STT (Nova-3 streaming WebSocket)
• Features:
  • Browser microphone capture via minimal inline JavaScript (getUserMedia)
  • WebSocket bridge: browser → FastHTML server → Deepgram streaming API
  • Live transcript display updated via HTMX WebSocket extension
  • Interim and final result handling with visual distinction
  • Smart formatting and punctuation enabled
• Includes: Single Python file app, minimal dependencies, no build step
• Complexity: Beginner-Intermediate

Acceptance criteria

• Runnable with minimal setup (pip install -r requirements.txt, add API key, python main.py)
• README explains the FastHTML + Deepgram WebSocket pattern
• Uses current SDK version
• Single-file or minimal-file implementation (FastHTML's strength is simplicity)
• Live transcription updates appear in the browser without page refresh
• Distinguishes between interim and final transcription results visually
• No Node.js, npm, or JavaScript build tools required

---

Raised by the DX intelligence system.