Skip to content
This repository was archived by the owner on Jan 20, 2026. It is now read-only.

v0.7.0 dbt_lever_source

Choose a tag to compare

@fivetran-data-model-bot fivetran-data-model-bot released this 28 Oct 16:57
· 5 commits to main since this release
72cbe9a

PR #26 includes the following updates:

Features

  • For Fivetran Lever connectors created on or after July 27, 2024, the USER and INTERVIEWER_USER source tables have been renamed to USERS and INTERVIEW_USER, respectively. This package now prioritizes the USERS and INTERVIEW_USER tables if available, falling back to USER and INTERVIEWER_USER if not.
    • If you have both tables in your schema and would like to specify this package to leverage the USER and/or INTERVIEWER_USER tables, you can set the variables lever__using_users and/or lever__using_interview_user to false in your dbt_project.yml.
    • For more information, refer to the July 2024 connector release notes and the related README section.
  • Introduced the ability to union source data from multiple Lever connectors. For more details, see the related README section.

Bug fixes

  • Fixed an issue where the dbt package would error due to a missing CONTACT_LINK source table for users without the titular source data, even if it was enabled in the Fivetran Connector. A null-filled table will now be generated in such cases.

Under the hood

  • Turned off freshness tests for USER/USERS and INTERVIEW_USER/INTERVIEWER_USER to avoid possible conflicts.
  • Updated temporary models to union source data using the fivetran_utils.union_data macro.
  • Added the source_relation column in each staging model to identify the origin of each field, utilizing the fivetran_utils.source_relation macro.
  • Updated tests to include the new source_relation column.
  • Added not_null tests for the relevant fields within each staging model.

Full Changelog: v0.6.0...v0.7.0