Creating datasets

Create a dataset

1. Open your workspace root or the dataset that should contain the new dataset.
2. Select New dataset.
3. Enter one name for the new child:
   New dataset

   Create a dataset to keep related tables and views together.

   Dataset name

   Create dataset
4. Select Create dataset. It appears beneath the selected parent in the sidebar.

You need Admin access on the selected parent. Editors can add and change tables and views, but cannot create, rename, move, or delete datasets.

Create deeper organization

Repeat the same action inside any dataset. For example, create commerce, open it and create stripe, then create production inside stripe. Loam creates only the child you requested, so each parent must already exist.

CREATE NAMESPACE commerce;
CREATE NAMESPACE commerce.stripe;
CREATE NAMESPACE commerce.stripe.production;

The UI calls these containers datasets. SQL calls each dataset in the path a namespace, which is why the command is CREATE NAMESPACE.

Organization and SQL use the same path

The dataset tree is the table's SQL path. If you add charges to commerce.stripe.production, its absolute SQL name is:

commerce.stripe.production.charges

Paths always start at the workspace root, without the workspace name. There is no current dataset, search path, automatic default, or fallback path.

Organizing tips

• Group by purpose. Datasets such as finance, marketing, and commerce make broad areas easy to find.
• Add depth only when useful. For example, commerce.stripe.production can separate one provider and environment without requiring special integration folders.
• Remember permission inheritance. Grants on a dataset apply through its descendants. A deny on the authenticated user or one of their current teams vetoes authenticated access; Everyone denies affect only anonymous public access.

Choose a dataset for connected data

When you connect a provider, select an existing dataset as its destination. If you need another destination, create it with the same New dataset action first. Integrations do not create a special dataset type or an automatic provider hierarchy. Synced tables appear as ordinary tables directly inside the selected dataset.

Naming rules

• Each component must be a valid SQL identifier.
• Quote a component containing spaces, punctuation, uppercase letters, or a reserved word.
• Names are unique across datasets, tables, and views with the same parent.
• A relation path can contain up to 32 dataset components (called namespaces in SQL) plus the final table or view name.

What's next

Open the destination dataset and select Add table to create a Table from CSV. Once a Table exists, the same dialog can load matching CSV, TSV, JSON, XLSX, or Parquet data into it. Loam does not choose an automatic destination dataset during upload.