Whisper Valet
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| process_audioA | Queue one audio/video file for the full pipeline (voice isolation -> transcription -> speaker diarization). Returns immediately; results land in // minutes later — poll status(). |
| process_folderA | Queue every audio/video file in a folder (non-recursive). |
| statusA | Pipeline status: inbox queue, watcher activity, recent clips, log tail. |
| list_clipsB | List all processed clip folders with their status. |
| get_transcriptB | Speaker-labeled transcript for a processed clip (see list_clips()). |
| get_reportB | Full per-segment speaker-confidence report for a processed clip. |
| configureA | Adjust diarization for FUTURE clips. num_speakers: "auto" (detect per clip — works for any recording), "N" to force exactly N, or "MIN-MAX" to bound the range (e.g. "2-5"). labels: comma-separated names in order of first appearance, e.g. "Interviewer,Responder" — applied only when the detected speaker count matches the label count, otherwise speakers get generic "Speaker N" names. Pass "" to leave a setting unchanged. Returns the active configuration. |
| transcribe_quickA | Plain synchronous transcription of a short clip via the MacWhisper CLI — no enhancement, no diarization, just text now. For noisy audio or speaker labels use process_audio() instead. Optional model override in engine:model-id format. |
| list_modelsA | List MacWhisper's downloaded transcription models (active one marked). |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 9 tools
Each tool has a clearly distinct purpose: folder queueing, single-file pipeline, status, clip listing, transcript retrieval, report retrieval, configuration, quick transcription, and model listing. While process_audio and transcribe_quick both handle audio, their descriptions draw a clear line between full pipeline and plain synchronous transcription.
Most tools follow a verb_noun pattern (process_folder, list_clips, get_transcript, list_models), but a few deviate: 'status' is a bare noun, 'configure' is a bare verb, and 'transcribe_quick' uses an adjective modifier instead of a noun. This is readable and close to consistent, but not perfectly uniform.
With 9 tools, the set is well-scoped for a transcription pipeline. Each tool fills a necessary role in the workflow—from intake (process_*) to monitoring (status) to output retrieval (get_*)—without redundancy or bloat.
The core workflow is fully covered: queueing, processing, status, results, and configuration. Missing are management operations like deleting/canceling clips or controlling the watcher, but these are optional extras rather than gaps that would break the primary use case.