Python SDK overview
The Labelty Python SDK (labelty-sdk on PyPI, import name labelty_sdk) lets you automate project work from Python: connect with an API key, read schemas, download assets, register models, and ingest predictions.
What you can do with the SDK today
Section titled “What you can do with the SDK today”| Area | Support |
|---|---|
| Connect / current project | Yes — LabeltyClient, client.project |
| Label schema | Yes — list/get/create |
| Assets | Yes — list/get/download/batch download |
| Model registry | Yes — create, upload, activate, YOLO helpers, publish |
| Predictions → annotations | Yes — predictions.create_batch |
| Org model store / marketplace | Yes — client.org_models, client.marketplace |
| Custom inference handlers | Yes — labelty_sdk.inference (runtime helpers, not HTTP) |
| Training runs | No high-level wrappers — use REST or the web UI |
| Inference workflows (deploy/run) | No client wrappers — use REST (run is allowlisted for keys) |
| Full annotation CRUD | Partial — use client.request(...) for unwrapped endpoints |
Package identity
Section titled “Package identity”| Item | Value |
|---|---|
| Install | pip install labelty-sdk |
| Import | from labelty_sdk import LabeltyClient |
| Python | 3.10+ |
| HTTP library | httpx |
When developing against the backend git submodule, the same classes may be imported as from app.sdk import …. Published user code should use labelty_sdk.
Suggested learning order
Section titled “Suggested learning order”Related product features
Section titled “Related product features”- Interactive exploration: Notebooks
- Managed training: Training overview
- Production inference API: Workflows overview