GitHubDiscord

Troubleshooting Guide

If you run into issues launching the dashboard, connecting your microphone, or getting Yumi's voice outputs, follow this comprehensive troubleshooting registry.

🛠️ Windows Issues

Error: KMP_DUPLICATE_LIB_OK

  • Symptom: The backend terminal crashes instantly on start, complaining about multiple OpenMP runtimes being loaded.
  • Cause: faster-whisper and other ML packages (like Intel MKL or PyTorch) import duplicate OpenMP instances.
  • The Fix: Ensure the environment variable is declared in your shell before starting.
    • PowerShell: 
      $env:KMP_DUPLICATE_LIB_OK="TRUE"
    • Command Prompt: 
      set KMP_DUPLICATE_LIB_OK=TRUE

Keyring Connection Blocked

  • Symptom: CLI crashes with a permission error when saving or getting API keys.
  • The Fix: Right-click your terminal emulator (e.g. PowerShell) and run it as a standard user (not Administrator). Windows Credential Manager sometimes locks access to administrative terminals for security.

🎙️ Microphone & Audio Issues

Speech Not Detected ([Voice Activity Detected] is missing)

  • Symptom: You talk to Yumi but nothing happens, and your terminal console does not register speech.
  • The Fix:
    1. Verify that your default system recording device is unmuted and capturing audio.
    2. If running inside a browser overlay (like OBS or Custom Chrome), ensure that browser page has been granted Microphone Access.
    3. Click the WebUI pink button "Connect & Start Audio Context" to ensure the WebSocket pipeline wakes up.

🔑 API & Secret Keys Issues

Authentication Failures (ElevenLabs or Groq)

  • Symptom: Subtitles show error messages or Yumi responds with fallback dialogue.
  • The Fix:
    1. API keys are saved in your OS Keychain, not in .env files.
    2. Re-run the configuration wizard to overwrite corrupt keys: 
      yumi --config
    3. Select Configure Senses and re-paste your active credentials.

Proceed to the Performance Tuning page to reduce latency!