Command line¶
scribe doctor check microphone, GPU, speech model, LLM server, data dir
scribe mics list input devices with their index
scribe record OUT.wav [--seconds N]
record from the microphone until Enter (or N seconds)
scribe transcribe AUDIO [--json FILE]
print a timestamped transcript; optionally save JSON
scribe note TRANSCRIPT [--template NAME] [--instructions TEXT]
draft a note from a .json transcript or a .txt file
scribe run AUDIO [--template NAME] [--label TEXT] [--instructions TEXT]
ingest + transcribe + draft, saved as a session
scribe sessions list sessions, newest first
scribe show ID print meta, transcript and note for one session
scribe purge delete unsigned sessions older than the retention period
scribe serve start the review app on SCRIBE_HOST:SCRIBE_PORT
Exit code is 0 on success, 1 on a reported error (printed as error: ...), 130 on Ctrl-C.
scribe doctor exits 1 if the LLM server is unreachable, so it works as a health check in scripts.
Common sequences¶
Transcribe now, draft later with two templates:
scribe run samples/visit-01.wav --label "room 2" # does both; or:
scribe transcribe room2.wav --json room2.json
scribe note room2.json --template soap > room2-soap.md
scribe note room2.json --template after-visit-summary > room2-avs.md
Nightly cleanup, only needed if you never start the review app (which purges on its own):