Notebooks overview

A Labelty notebook is a managed Amazon SageMaker Studio JupyterLab environment tied to one project. It is for interactive Python work against your project via the Python SDK.

Labelty does not embed Jupyter inside the web app. Clicking Notebook opens Studio in a new browser tab.

Diagram: Labelty UI → ensure SageMaker space/app → presigned Studio URL → JupyterLab with lifecycle env injection

Why use a notebook?

• Experiment with dataset downloads and model code next to your labeled data
• Reuse the same project API key the platform creates for the space
• Persist files on the space’s EBS volume across Stop → Open cycles

Prerequisites

Requirement	Detail
Active organization subscription	Required to open
Plan feature notebook_enabled	Must be allowed on your plan
Prepaid wallet	Minimum balance required (commonly $1.00 USD)
Permission	Ability to read the project (PROJECT_READ)
Platform infra	SageMaker domain configured for the environment

If open fails for billing reasons, top up the organization wallet and retry.

How to open (UI)

1. Open your project in the Labelty dashboard.
2. Click Notebook (near Members / Labels on the project header).
3. In the modal, pick an instance size.
4. Click Open Notebook.
5. Wait while status polls (starting …). Cold start is often ~30–60 seconds.
6. When status is running, click Open JupyterLab to open a fresh presigned URL.
7. When finished, click Stop to delete the app (files on the space remain).

Screenshot: Project header → Notebook button
Screenshot: Notebook modal — instance picker and Open JupyterLab

URL timing

• Presigned open links are short-lived (about 5 minutes, single-use). If the tab fails, click Open JupyterLab again.
• Studio sessions last on the order of 12 hours.

Instance sizes (UI)

Exact prices can change; the UI shows approximate hourly rates. Backend safelist includes:

Type	Typical UI label
ml.t3.medium	Small (CPU)
ml.t3.large	Medium (CPU)
ml.m5.large	Large (CPU)
ml.g4dn.xlarge	GPU (T4)

Billing is PAYG from the org wallet while the app runs. Idle timeout is fixed at 60 minutes. Wallet ≤ $0 can auto-stop the notebook.

GPU image caveat

The product picker may offer GPU, but confirm with your team that the environment actually provisions a GPU-capable image. See Open questions.

Backend API (advanced)

Base: /api/v1/projects/{project_id}/notebook

Method	Path	Purpose
POST	/open	Ensure space + app; body { "instance_type": "ml.t3.medium" }
GET	/status	Read state without creating resources
POST	/stop	Delete app and finalize billing

States: stopped | starting | running | stopping | failed.

What Labelty configures for you

On first open, Labelty creates a SageMaker space for the project and a project-scoped API key (name pattern like notebook-…). Space tags store:

• labelty:api_url
• labelty:api_key

On every JupyterLab start, lifecycle.sh injects those values as environment variables (see next page).

Next step

Use the SDK inside JupyterLab: Work in JupyterLab.