@openwhispr/cli

The unified openwhispr command-line tool. Works against either the local desktop app (via a loopback HTTP bridge) or the cloud REST API, with auto-detection.

Install

npm i -g @openwhispr/cli

Requires Node.js >= 20.

Quick start

# If the desktop app is running, this just works:
openwhispr notes list

# Otherwise, configure remote access:
openwhispr auth login            # paste an owk_live_... API key
openwhispr notes list

# Diagnose connectivity to both backends:
openwhispr doctor

Backend selection

The CLI picks a backend in this order:

1. Explicit flag: --local or --remote
2. Env var: OPENWHISPR_BACKEND=auto|local|remote
3. Config file: ~/.openwhispr/cli-config.json → backend
4. Auto (default): try local first, fall back to remote, fail if neither is configured

Files

• ~/.openwhispr/cli-config.json (mode 0600) — CLI config plus API key
• ~/.openwhispr/cli-bridge.json (mode 0600, written by the desktop app) — port + bearer token for the local bridge

Compatibility matrix

CLI version	Min desktop version	Min API version
0.1.x	TBD	TBD

The desktop bridge contract is versioned at /v1/. Breaking changes ship under a new major version.

Exit codes

Code	Meaning
0	Success
1	User error (bad args, missing flag)
2	Backend unreachable
3	Auth failure
4	Not found
5	Safety gate refused

Command surface

openwhispr doctor
openwhispr version

openwhispr auth login | logout | status
openwhispr config get | set <key> <value>

openwhispr notes list | get | create | update | delete | search
openwhispr folders list | create
openwhispr transcriptions list | get | delete
openwhispr audio delete <transcription-id>
openwhispr dictionary list | add <word...> | remove <word...>
openwhispr snippets list | add <trigger> <replacement> | remove <trigger...>
openwhispr transcribe <file> [--model <id>] [--language <code>] [--prompt <text>]
                             [--format text|json] [--note] [--title <t>] [--folder <name>]

Run openwhispr <command> --help for full flags.

Dictionary and snippets

The dictionary is a list of words (names, jargon, acronyms) that biases transcription toward the spelling you want. Snippets expand a spoken trigger phrase into saved text. Both commands work against the running desktop app or the cloud API; remote API keys need the dictionary:read/dictionary:write and snippets:read/snippets:write scopes.

Transcribing audio files

openwhispr transcribe <file> turns an audio file into text. With the desktop app running it uses the app's local models: free, on-device, no size limit. --model picks a downloaded model by name (e.g. base); a rejected model prints the available ones. Add --remote to use OpenWhispr Cloud instead, which requires a Pro or Business plan and an API key with the transcriptions:write scope. --prompt is cloud only; --model is local only.

openwhispr transcribe meeting.m4a                       # local, prints the transcript
openwhispr transcribe meeting.m4a --format json         # text plus provider, model, duration
openwhispr transcribe meeting.m4a --note --folder Work  # save as a note (folder by name)
openwhispr transcribe meeting.m4a --remote --prompt "Names: Ada, Linus"

Cloud transcription is in beta and its limits may change. Each request is capped at 4 MB, so larger files are split into 4-minute chunks with ffmpeg (must be on your PATH), uploaded one at a time, and the transcripts joined. Cloud minutes are limited to 600 per month per account. The audio is uploaded; the file's path on your machine is not.

Enhanced notes

For notes with an AI-enhanced version (e.g. meeting notes), notes get's markdown output shows the enhanced note when one exists, matching the desktop app. The raw fields are always available via --format json (content, enhanced_content).

Fetching just the transcript

openwhispr notes get <id> returns the whole note payload. To pull only the transcript, pass --transcript:

# Transcript as JSON: { "transcript": "..." }
openwhispr notes get <id> --transcript --format json

# Transcript as a clean, speaker-labeled markdown blob
openwhispr notes get <id> --transcript --format markdown

The markdown output parses the transcript into a readable, copy-paste-ready document — a ## Transcript section with one line per segment, each timestamped and labeled with the speaker: Me for your own mic, otherwise the segment's speakerName when available, falling back to the diarized id (speaker_0 → Speaker 1, matching the desktop app):

# Meeting name

## Transcript

**jordan.lee@example.com** _(15:32:18)_: Hi, pleasure to meet you...

**Me** _(15:33:23)_: I'm Sam. I lead the product team...