Skip to main content

Command Reference

This page covers the public CLI surface in one place.

Use slmcortex or python -m slmcortex.

On macOS and Linux, the expected first install path is the Homebrew tap: alvarolorentedev/homebrew-SLMCortex.

Start With These Commands​

If you are new to the project, learn these first:

CommandWhat it is for
slmcortex doctorcheck install health, workspace paths, and backend visibility
slmcortex compose-folderscan one folder, pick packages, compose a runtime, and optionally export it
slmcortex infer --dry-runinspect routing without loading a real model
slmcortex serveexpose the local compatibility server
slmcortex agent runrun bounded repository work against a local checkout

Product Paths​

There are two practical ways to think about the CLI:

  • Composer-first product path: doctor, compose-folder, infer, serve, agent run
  • Factory/package path: generate-dataset, validate-dataset, train-slm, package-slm, compose-slms

The Composer-first path is what most visitors should start with.

Common Conventions​

  • Prefer --dry-run when you only want to inspect routing or control flow.
  • infer requires exactly one of --runtime or --slms-dir.
  • agent run requires exactly one of --runtime or --slms-dir unless project defaults exist.
  • backend: auto resolves to MLX on macOS Apple Silicon and GGUF elsewhere.
  • GGUF requires a .gguf runtime model path.

Composer-First Commands​

doctor​

Check packaged-product readiness and workspace health.

slmcortex doctor
slmcortex doctor --workspace /tmp/slmcortex-app
slmcortex doctor --export-support-bundle

compose-folder​

Compose a runtime from one local folder and the external app workspace.

slmcortex compose-folder \
  --workspace /tmp/slmcortex-app \
  --folder /path/to/repo \
  --task "Create a FastAPI endpoint with request validation" \
  --export-descriptor /tmp/slmcortex-app/exports/repo.json

init​

Create project-local SLMCortex state for the project-owned LoRA flow.

slmcortex init

loras download​

Download a selected project-owned LoRA into the local SLM directory.

slmcortex loras download fastapi
slmcortex loras download hf://owner/repo --as fastapi

infer​

Run local inference or inspect routing with --dry-run.

slmcortex infer \
  --runtime /tmp/slmcortex-demo/runtime \
  --prompt "Fix this Python traceback" \
  --dry-run

Or use an SLM directory directly:

slmcortex infer \
  --slms-dir .slmcortex/prototype-slms \
  --prompt "Fix a FastAPI validation bug" \
  --allow-remote-loras \
  --dry-run

serve​

Expose the local OpenAI-compatible compatibility server.

slmcortex serve --runtime /tmp/slmcortex-demo/runtime

Or use the adaptive path:

slmcortex serve \
  --slms-dir .slmcortex/prototype-slms \
  --allow-remote-loras \
  --dry-run

agent run​

Run bounded local repository work.

slmcortex agent run \
  --runtime /tmp/slmcortex-demo/runtime \
  --repo /path/to/local/repo \
  --task "Fix the failing answer implementation." \
  --dry-run

Factory And Package Commands​

generate-dataset​

Generate deterministic train and eval JSONL datasets.

slmcortex generate-dataset \
  --slm-id fastapi_contract \
  --domain fastapi \
  --report-output /tmp/fastapi_contract-report.json

validate-dataset​

Validate one dataset and optionally check leakage against eval data.

slmcortex validate-dataset datasets/fastapi_contract/train.jsonl \
  --eval-dataset datasets/fastapi_contract/eval.jsonl

train-slm​

Train a LoRA SLM from datasets and package it as a Slm Cortex artifact.

slmcortex train-slm \
  --slm-id fastapi_contract \
  --name "FastAPI Contract Slm" \
  --train-dataset datasets/fastapi_contract/train.jsonl \
  --eval-dataset datasets/fastapi_contract/eval.jsonl \
  --output slms/fastapi_contract

package-slm​

Wrap an existing adapter as an SLM package.

slmcortex package-slm \
  --slm-id python_slm \
  --name "Python Slm" \
  --adapter-dir artifacts/adapters/python_slm \
  --train-dataset tests/fixtures/slmcortex_demo/train.jsonl \
  --eval-dataset tests/fixtures/slmcortex_demo/eval.jsonl \
  --eval-summary tests/fixtures/slmcortex_demo/eval-summary.json \
  --output /tmp/slmcortex-demo/python_slm

validate-slm-package​

Verify package structure, checksums, and protected inputs.

slmcortex validate-slm-package --path slms/python_slm

compose-slms​

Compose validated packages into one deterministic runtime bundle.

slmcortex compose-slms \
  --slms /tmp/slmcortex-demo/python_slm,/tmp/slmcortex-demo/debugging_slm \
  --output /tmp/slmcortex-demo/runtime

validate-runtime​

Check a runtime bundle before inference or serving.

slmcortex validate-runtime --runtime /tmp/slmcortex-demo/runtime

route​

Inspect which packages match a task without loading adapters.

slmcortex route \
  --slms-dir slms \
  --repo . \
  --task "Create a FastAPI endpoint" \
  --explain