Skip to content

WHISPLAY_IDEAS.md

Latest commit

 

History

History
21 lines (14 loc) · 1.91 KB

WHISPLAY_IDEAS.md

File metadata and controls

21 lines (14 loc) · 1.91 KB

Ideas from PiSugar whisplay-ai-chatbot

This project is inspired by whisplay-ai-chatbot. Below is what we’ve adopted and what could be added.

Implemented

  • TTS playback – Optional spoken responses via OpenAI TTS (like whisplay). Set ENABLE_TTS=true in .env so the device speaks the assistant response after it’s streamed. Uses existing tts_openai.py (OpenAI TTS + aplay). Cancel with the button during playback.
  • Battery display – Already present: PiSugar socket + sysfs fallback in display.py (top-right corner).

Possible next steps

  • Conversation reset on idle – Whisplay resets conversation history after 5 minutes with no speech. If OpenClaw gains a session/thread API and a “clear” or “new thread” endpoint, we could add CONVERSATION_RESET_IDLE_SEC=300 and call it when idle that long so the next query starts fresh.
  • Wake word – Whisplay supports optional wake word (e.g. “hey Amy”) via openwakeword and sox for continuous mic. We could run a small subprocess that prints WAKE on detection; main would then start recording (and use sentence-based or timeout stop). Requires wiring a listener process and VAD/timeouts.
  • Image on screen – If we add image generation or OpenClaw returns images, we could show them using the same PIL/display path we use for the response (e.g. full-screen image then back to text).
  • Volume control by assistant – Whisplay lets the AI adjust volume. We could parse responses for volume intent or add a tool that calls amixer (lower priority).

Reference

  • Repo: https://github.com/PiSugar/whisplay-ai-chatbot
  • Wake word: python/wakeword.py (openwakeword + sox, env: WAKE_WORD_ENABLED, WAKE_WORDS, etc.)
  • Their Python UI: python/chatbot-ui.py (socket server + render thread; we use a single process and our own display flow).