Repository navigation
Troubleshooting
Stas Pisockij edited this page Sep 17, 2026
·
1 revision
| Symptom | Fix |
|---|---|
| "ffmpeg and ffprobe are required" |
brew install ffmpeg, then restart the app. |
| No transcript after a call | Settings -> Models: confirm a Whisper model is downloaded. |
| Transcript saved, speakers unnamed | Speaker detection is not set up or failed. Open Review Speakers and retry; the audio is kept until speakers are reviewed. |
| Speaker detection keeps failing | Review Speakers -> Speaker setup -> Check Speaker Setup, and re-select the Python environment. |
| Nothing recorded from the other side | System Settings -> Privacy & Security -> Screen & System Audio Recording: enable Call Recorder, then restart the app. |
| A row says "One side only" | System audio was not captured for that call. Check the permission above. |
codex mcp list shows no server |
Re-run codex mcp add with the exact quoted paths, then restart Codex. |
| Tools say "database is locked" | Close the app's Settings window and retry; the app and the server share one local database. |
| Semantic search returns nothing | The embedding model has not been downloaded yet. Let Settings -> Models finish, or use lexical search. |
| Database looks damaged | Settings -> Recovery: check the database, take a backup, or restore working files. |
Every error surface in the app has Copy Error Details, and Settings -> Recovery can export a redacted diagnostics bundle. Include one of them in an issue; both exclude audio, transcript text, and voiceprints.