Skip to content

dbt

Depositori reads a dbt project from two files dbt writes on every run: manifest.json and catalog.json. It reads the models, columns, descriptions, tests and lineage.

  • The Owner role in Depositori.
  • The project’s pipeline publishes manifest.json (and catalog.json, for column types) to a folder in Azure Storage. Ask whoever runs the pipeline for the folder’s address. It looks like https://<account>.blob.core.windows.net/<container>/<folder>.
  • A Microsoft account with the role Storage Blob Data Reader on that container.
  1. Open Connections, choose Add connection, then dbt.
  2. Paste the Artifacts folder, give it a Name. With only the container, you can pick from the projects in it.
  3. Test, then Save connection.
Connect dbt with the artifacts folderConnect dbt with the artifacts folder

Right after saving, choose Add as a model under Read it now. Later, use Add as a model on the connection’s card, or Read dbt models in Add a model. Give the model a name and choose Import.

The card of the dbt connectionThe card of the dbt connection

When nothing changed since the last read, Depositori says “Nothing new since the last read.” and opens the model it has.

The card has no Taki may query: dbt’s artifacts hold definitions, not rows. Taki reads the project from the model and runs no queries on it.

A dbt source is loaded into the warehouse by something outside dbt (dlt, Fivetran, …), so its lineage stops there. The Catalog and the lineage show it as loaded from outside, naming the loader when dbt knows it.

When the warehouse is read for lineage and a dbt source isn’t in it, the model notes “dbt source not found in this warehouse; if it should be here, check the read account’s rights”. That is a note, not a problem: the rest of the read is complete.

Choose Upload files on the guided start, or the + beside Models, then dbt, and upload the files from your computer. No account is needed. To update the model later, upload them again.