System Log: JARVIS Voice Assistant Rebuild — May 17, 2025

📅 System Log — May 17, 2025
🖥️ Machine: ASUS (Linux Mint)
🎙️ Mic: Logitech C930e
🔊 Output: Logitech analog speakers
🧠 AI Model: phi (via Ollama + Open WebUI)
💬 Voice: PicoTTS (en-GB “Bill” voice)

✅ Completed Today

🧠 Whisper.cpp Rebuild

  • Recompiled whisper.cpp from source
  • Installed ggml-base.en.bin model
  • Recorded and transcribed a 5-second test voice sample
  • Verified output: “This is a test. This is only a test.”

🔄 Google Cloud STT (Active)

  • Enabled USE_GOOGLE_STT=true in talk-to-ai.sh
  • Successfully used Google Cloud Speech-to-Text with Logitech mic
  • Confirmed environment variable for API key is working

🤖 JARVIS Voice Loop Fully Operational

  • Script: talk-to-ai.sh
  • Captures mic input
  • Transcribes via Google or Whisper
  • Sends to Open WebUI via local API
  • Speaks reply using pico2wave and British “Bill” voice

🐳 Docker + WebUI Status

  • Open WebUI container open-webui-open-webui running and healthy
  • Model phi:latest successfully registered in Open WebUI backend
  • Manual test confirmed: JARVIS replies are now vocalized

📁 Logs

  • All interactions stored in: ~/jarvis/logs/YYYY-MM-DD_HH-MM-SS.log
  • System now supports both offline (Whisper) and cloud (Google) transcription

🧾 Next Steps

  • Enhance JARVIS voice with effects or Coqui TTS
  • Enable hotword trigger (e.g., “Jarvis”)
  • Auto-start at login
  • SSL setup to enable remote logging via REST API