Bug Description

The gateway ignores stt.enabled: false in config.yaml and always attempts to call OpenAI Whisper API for voice messages, causing 401 errors when no valid OpenAI key is configured.

Expected Behavior
When stt.enabled: false is set, the gateway should skip STT transcription entirely and pass voice messages through without processing, or use a configured alternative provider.

Actual Behavior

1. Gateway receives voice message
2. Attempts OpenAI Whisper API call regardless of stt.enabled setting
3. Returns "401 Incorrect API key" error to user
4. Message flow is blocked

Environment

• Hermes Gateway (latest)
• Config:

stt:
  enabled: false
  # provider: openai  # Tried with/without this

Logs

Processing audio...
Error transcribing audio: 401 Incorrect API key provided...

Workaround
Currently using manual Deepgram transcription via Python script on cached .ogg files from /root/.hermes/audio_cache/.

Suggested Fixes

1. Immediate: Honor stt.enabled: false - skip transcription entirely when disabled
2. Better: Support pluggable STT providers (Deepgram, Whisper local, etc.) configured via YAML
3. Best: Default to disabled if no valid API key is present, don't hardcode OpenAI

Happy to provide more details or test fixes!