Labqoat

labqoat is the public Python SDK and the lab command-line client for the shared Labqoat cloud platform. Application repositories contain only authored resources—models, jobs, datasets, evaluations, and transforms. Terraform, Kubernetes, persistence, and control-plane services stay in the platform repository and are deployed once.

Install

uv tool install labqoat
# or
pipx install labqoat

Python projects that import the SDK should also declare labqoat as a normal dependency.

Start a project

lab init my-lab-project
cd my-lab-project
uv sync

lab model new audio/my-model
lab train new audio/my-training-job
lab dataset new audio/my-dataset

Authenticate once, then open the AWS dashboard with the same stored session:

lab auth
lab auth ui

The browser handoff is automatic; the bearer token is not printed or placed in the HTTP request URL.

The generated repository has this shape and no infrastructure code:

my-lab-project/
├── datasets/
├── evals/
├── jobs/
├── models/
├── transforms/
├── pyproject.toml
└── README.md

Run lab <group> --help for the command surface. Local check, test, and build commands do not require cloud authentication. submit and deploy require a clean pushed Git commit and send a source snapshot to the one shared AWS builder; user repositories never contain or apply Terraform.

Development

uv sync --all-groups --locked
uv run ruff format --check .
uv run ruff check .
uv run pyright
uv run pytest
uv build

Releases

Every push to main that passes CI is published to PyPI automatically. The workflow increments the latest successful release tag's patch version, builds the distributions, publishes them, and pushes the matching vX.Y.Z Git tag.

The repository must have a GitHub Actions secret named PYPI_API_TOKEN whose value is a PyPI token allowed to publish labqoat.