What is Labelty?

Labelty is a platform for data labeling, model training, and production inference.

You use it to:

Upload and organize images (and other asset types) in a project
Create label schemas and annotate assets with humans and ML models
Train models on approved labeled data
Run experiments in a project notebook (SageMaker JupyterLab)
Expose inference through an inference workflow REST API

Who these docs are for

These docs assume you have never used Labelty before. You should be comfortable with:

Using a web browser
Basic Python (for the SDK and notebooks)
Optional: HTTP / cURL (for the workflow API)

You do not need prior knowledge of SageMaker, Ultralytics, or Labelty’s internal architecture.

What this documentation covers

Area	What you will learn
Authentication	Project API keys (`lty_…`) and how to store them
Python SDK	Install, connect, assets, schema, models, predictions
Training	Dataset preview → train → monitor → build inference package
Inference workflows	Deploy `workflow.py` and call `POST …/workflows/{id}/run`
Notebooks	Open JupyterLab with `LABELTY_API_URL` / `LABELTY_API_KEY` injected

High-level product flow

Diagram: End-to-end Labelty flow — Label → Train → Build → Auto-label or Workflow API

1. Create a project and label schema
2. Upload assets and annotate (human and/or auto-label)
3. Approve labeled assets for the training dataset
4. Start a training run (SageMaker)
5. Build an inference package and activate the model
6a. Auto-label more assets inside the project
   — or —
6b. Deploy an inference workflow and call the run API from your app

Environments

Examples in these docs often use the development API:

https://api.development.labelty.ai/api/v1

Your organization may use a different base URL for production.

Next step

Continue with Core concepts so the rest of the guides use shared vocabulary.