Skip to content

v0.2.4

Released: 2026-11-07

Headline: safer MotherDuck queries and reliable dbt run history.

This release improves how tycoon works with MotherDuck and makes dbt execution history more reliable.

MotherDuck warehouses are now recognised consistently through both md: and motherduck: prefixes. More importantly, tycoon data query now refuses write statements against MotherDuck unless --allow-writes is explicitly supplied. This provides a safer default when querying remote databases, where an accidental write could affect production data.

The release also improves how tycoon records and displays dbt runs. Successful builds that include no-op results are no longer incorrectly marked as failed, and failed runs are recorded even when dbt does not produce a run_results.json file. Builds triggered through run-all are now included in the metadata store, and data status correctly identifies the last build for each dbt layer.

Alongside these changes, MotherDuck schema inspection now uses estimated row counts by default, reducing the cost of counting large remote tables. The city renderer also receives improvements to district labels and dbt lineage for sources outside the warehouse.

Landed

MotherDuck is safer and better supported

  • Both MotherDuck connection prefixes are recognised (#308). Projects using motherduck:<name> now work consistently across registration, initialisation, dbt profile detection and secret redaction. Existing md: and local database configurations continue to work.
  • Writes through tycoon data query require explicit permission (#309). Queries against MotherDuck are restricted to supported read operations by default. Use --allow-writes when a command or script intentionally needs to execute other statements. This change affects MotherDuck queries; local DuckDB files are unchanged. The restriction is an accidental-write safeguard, not a permission boundary. Use appropriately scoped MotherDuck credentials where stronger restrictions are required.
  • Schema inspection uses estimated row counts (#310). tycoon data schema reads catalog estimates instead of running an exact count against every MotherDuck table. Use --exact when you need exact counts. Estimates may remain higher than the actual row count after deletions. Views and tables without catalog estimates are identified accordingly.

dbt execution history is more reliable

  • Successful builds are no longer marked as failed because of no-op results (#341). Tycoon now identifies dbt failure statuses rather than treating every unrecognised status as an error.
  • Failed runs are recorded even when dbt does not produce a run-results file (#377). History now includes the exit code and a concise error summary. Runs whose selector matches no nodes are also identified as having nothing to run.
  • run-all now records its dbt build (#326). Builds triggered through tycoon data run-all are captured in the metadata store, making them available to history and the city renderer.
  • run-all resolves dbt consistently (#439). It now uses the dbt installation associated with the Python environment running tycoon, rather than whichever executable happens to appear first on PATH.
  • data status correctly reports the last build for each dbt layer (#390). It now matches models using dbt's unique_id and reads build timestamps in local time. Existing metadata ledgers do not require migration.

The city renderer provides better lineage and labels

  • District and civic labels avoid collisions (#385). Overlapping chips are repositioned where possible, and civic labels that still have no room are hidden until the view changes.
  • External dbt sources appear in model lineage (#387). Staging models now identify sources located outside the warehouse, including sources in separate databases, attached DuckDB files and DuckLake. These sources are represented in lineage but do not appear as city buildings because their contents cannot be measured from the warehouse.

Upgrading from 0.2.3

  • Review MotherDuck write commands. If you use tycoon data query to execute write statements against MotherDuck, add --allow-writes to those commands and scripts. Without it, unsupported write statements are refused before connecting.
  • Use --exact when exact MotherDuck row counts are required. Schema inspection now uses catalog estimates by default. Estimates can remain higher than the actual row count after deletions.
  • No metadata migration is required. Existing dbt run ledgers continue to work with the corrected data status behaviour.
  • Local DuckDB projects are unaffected by the MotherDuck query restrictions and row-count changes.