🐛 Bug: “Enable Whisper STT” button does nothing — local whisper-cli never activated
Summary
Clicking the “Enable Whisper STT” button in Glass v0.2.4 (Mac OS) appears to do nothing. No local Whisper transcription is triggered, and logs confirm that STT continues using OpenAI’s model (gpt-4o-mini-transcribe-glass) even after clicking the button.
⸻
🔁 Reproduction Steps
1. Install and run latest glass from main branch (v0.2.4)
2. Ensure whisper-cli is installed and accessible (via Homebrew)
3. Confirm it works manually:
whisper-cli ~/whisper_test.wav --language en --model ~/whisper-models/ggml-base.en.bin
4. Launch Glass via npm start
5. Navigate to Settings → Models
6. Click Enable Whisper STT
⸻
🧪 Expected Behavior
Whisper should be selected as the active STT engine, and a log line like this should appear:
[WhisperService] Running whisper-cli with args...
⸻
❌ Actual Behavior
• No visual feedback
• No console or terminal logs referencing Whisper usage
• Whisper STT is never selected or activated
• Glass continues using openai-glass STT
⸻
✅ Environment
• Glass version: v0.2.4 (latest main)
• macOS: Apple M3 Pro (macOS 14.5)
• Whisper CLI: Installed via Homebrew and working
• Path: /opt/homebrew/bin/whisper-cli (in $PATH)
• Whisper model: ggml-base.en.bin, downloaded and verified
• Terminal logs confirm:
[WhisperService] Found whisper-cli at: /opt/homebrew/bin/whisper-cli
[WhisperService] Initialized successfully
⸻
💡 Notes
• All audio transcription features work if OpenAI STT is left as the default
• A working whisper-cli + model is confirmed in the environment (output available)
• This may be a UI or model selection logic bug that silently fails or ignores the switch
⸻
🐛 Bug: “Enable Whisper STT” button does nothing — local whisper-cli never activated
Summary
Clicking the “Enable Whisper STT” button in Glass v0.2.4 (Mac OS) appears to do nothing. No local Whisper transcription is triggered, and logs confirm that STT continues using OpenAI’s model (gpt-4o-mini-transcribe-glass) even after clicking the button.
⸻
🔁 Reproduction Steps
1. Install and run latest glass from main branch (v0.2.4)
2. Ensure whisper-cli is installed and accessible (via Homebrew)
3. Confirm it works manually:
whisper-cli ~/whisper_test.wav --language en --model ~/whisper-models/ggml-base.en.bin
⸻
🧪 Expected Behavior
Whisper should be selected as the active STT engine, and a log line like this should appear:
[WhisperService] Running whisper-cli with args...
⸻
❌ Actual Behavior
• No visual feedback
• No console or terminal logs referencing Whisper usage
• Whisper STT is never selected or activated
• Glass continues using openai-glass STT
⸻
✅ Environment
• Glass version: v0.2.4 (latest main)
• macOS: Apple M3 Pro (macOS 14.5)
• Whisper CLI: Installed via Homebrew and working
• Path: /opt/homebrew/bin/whisper-cli (in $PATH)
• Whisper model: ggml-base.en.bin, downloaded and verified
• Terminal logs confirm:
[WhisperService] Found whisper-cli at: /opt/homebrew/bin/whisper-cli
[WhisperService] Initialized successfully
⸻
💡 Notes
• All audio transcription features work if OpenAI STT is left as the default
• A working whisper-cli + model is confirmed in the environment (output available)
• This may be a UI or model selection logic bug that silently fails or ignores the switch
⸻