stem-splitter
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| SHADOWROOM_HOME | No | Base directory for the virtualenv and output directories. Default: ~/Documents/ShadowRoom | |
| STEM_SPLIT_PYTHON | No | The interpreter that has Demucs. Default: <ShadowRoom>/_venvs/stem-splitter/bin/python | |
| STEM_SPLIT_COMMAND | No | Remote command template for separation, e.g. `ssh gpu 'split {input} {output_dir}'` with `{input}`, `{output_dir}`, `{model}` placeholders. Leave unset to use local Demucs. | |
| STEM_SPLIT_JOB_DIR | No | Directory for job records. Default: <ShadowRoom>/stem-splitter-jobs | |
| STEM_SPLIT_OUT_DIR | No | Where the stems are written. Default: <ShadowRoom>/stems |
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 | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| separate_stemsA | Start separating one audio file into vocals / drums / bass / other. Returns a job id immediately. |
| stem_job_statusA | Read one separation job (or the recent ones): status, stage, stems, errors. |
| stem_job_cancelA | Stop a running separation and remove its half-written stems. |
| stem_backendsA | Which separation backends this machine has (local demucs, or STEM_SPLIT_COMMAND) and how to enable them. |
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 4 tools
The four tools map to distinct actions: starting a job, checking status, cancelling, and listing backends. There is no overlap in purpose, so an agent should be able to select the correct one without ambiguity.
Naming is mixed: separate_stems follows verb_noun, while stem_job_status and stem_backends are noun phrases and stem_job_cancel reverses the object/verb order. A consistent pattern such as get_stem_job_status, cancel_stem_job, and list_stem_backends would make the set more predictable.
Four tools is well-scoped for a job-based audio stem separation server. Every tool covers a necessary lifecycle action without redundancy.
The core job lifecycle is covered: start, inspect status, and cancel, plus backend discovery. A minor gap is the lack of an explicit result-retrieval or cleanup tool, though status may expose stem outputs indirectly.