Select and refresh catalog metadata

← Sources

Catalog contains the source names, tables, and columns used to build models and check queries. Refresh it after creating a connector or changing a warehouse schema.

You need an existing connector. Explorers can browse and refresh Catalog; changing the selection needs the Editor role or higher.

Choose the tables you need

Open Sources → Catalog and pick the connector in the command bar. Connectors with a catalog level, such as BigQuery projects or Snowflake databases, also have a catalog picker and Manage catalogs.
Select Choose tables to open the Choose tables to load panel. Check whole schemas or specific tables. A checked schema also includes tables added to it later.
Select Save and refresh to save the selection and load its tables and columns. Save only saves the selection without loading.

Limit selection to useful scopes so that you can review what Monomer will read.

Discover the Workspace, read workspace, and follow availableOperations through the connector to its Catalog paths. Read the current selection, then send its exact EDIT entry to monomer_describe before constructing the patch. EDIT has no inputKind; the resource path determines the schema. Selection uses one root replacement: the patch has exactly one replace operation with path: "". Omit revision entirely. Include every existing schema and table selection you intend to keep; omitted selections are removed.

For a PostgreSQL connector where analytics.orders is already selected, this monomer_edit example keeps orders and adds customers. Replace the IDs and names, and retain any other selections from the current document.

monomer_edit input

{
  "workspaceId": "0198ef28-2000-7000-8000-000000000001",
  "path": "workspace/connectors/0198ef28-2000-7000-8000-000000000002/catalog/selection",
  "patch": [{
    "op": "replace",
    "path": "",
    "value": {"selectedSchemas": [{"schemaName": "analytics", "relations": {"kind": "SELECTED_RELATIONS", "relationNames": ["orders", "customers"]}}]}
  }]
}

For a source with a catalog level, use the exact catalog selection path returned by discovery. Its schema/table selection has the same selectedSchemas shape. The connector’s separate catalog/catalog-selection resource uses selectedCatalogNames; read and preserve that full list when choosing catalogs. Empty selection arrays clear those selections. To retain all relations in a schema, preserve its {"kind":"ALL_RELATIONS"} value instead of converting it to a partial table list.

Read result.path after editing to check the saved scope and discover the next operations. Invoke the advertised ingestion action to collect metadata; selection-options refresh only retrieves choices. Omit input on advertised input-less actions. Read the returned job path until completion. Saving selection alone does not collect metadata. See the Agent request schema for the exact selection resource shapes.

Refresh and inspect

Select Refresh in the command bar to load the selected schemas again. A selected schema that has never been loaded shows Not loaded yet.
Wait for the outcome. The command bar shows progress and Cancel; you can leave and return to the page.
Open Recent refreshes to see past refreshes and their errors. A schema or table that failed shows the reason in its row. A refresh with errors does not confirm that all selected tables were read.
Select a schema and a table, or use Find table (/) to search loaded table and column names across connectors. The table detail lists columns, keys, and the semantic models that use it.
From the table detail, continue with Query in Explore (opens a SELECT for the table without running it), Create semantic model (opens the editor with this table as its source), or View lineage. The Catalog URL keeps the schema and table, so you can share the location.

If you cancel a refresh, wait for the final outcome; a cancellation request alone does not mean the work has stopped. When a refresh is already active, inspect it instead of repeatedly starting new work.

If tables are missing

Check the selected connector and catalog, saved selection, source grants, and Recent refreshes. Find table only finds tables that have been loaded. If the source changed, refresh again after resolving the cause. If Monomer reports that collection is unsupported for the selected source, keep the error and source type for your administrator rather than pointing models at unrelated source tables.

Catalog is metadata, not a live data sample. A listed table can still fail a query if source credentials or data access have changed. Once the table and columns are correct, create a semantic model.

BigQuery partition details

A BigQuery table can show its partition key, clustering keys in warehouse order, and Filter required. Partition columns are marked in the field list. Ingestion-time keys such as _PARTITIONTIME are shown as partition metadata rather than ordinary fields. These details come from the latest successful refresh; refresh after changing the warehouse option. Other source types do not report these details yet.