Lightdash metrics in the generated dbt project, and release 1.14.0

[jarichb]

Adds a Lightdash exporter beside the Ossie one, so a metrics file reaches Lightdash too.

What changes

• generate --lightdash writes the model's metrics, dimensions and joins into the generated dbt project:
  • Where: under config.meta of each staging model's own models/staging/stg_<table>.yml. The meta is merged in next to the tests, because dbt reads a model's properties from one file only.
  • Which metrics: the --metrics file's, or without one the metrics the model implies.
• model2data metrics export --to lightdash writes the same meta as one properties file, without generating.
• Python API: model2data.metrics.to_lightdash and write_lightdash, beside to_ossie.

Mapping. Format chosen: the dbt YAML meta that Lightdash documents as its default for dbt projects. Its MetricFlow support is beta, and Lightdash YAML is for projects without dbt.

• Joins: each table is joined many-to-one to every table it reaches along exactly one path.
• Metrics:
  • simple metrics sit on their column;
  • row counts are count_distinct of a one-column primary key;
  • ratios and expressions are type: number metrics over ${input}.
• Text fields: ai_context becomes ai_hint. percent and currency become '0.00%' and '#,##0.00'.
• Filters:
  • they become filters wherever Lightdash's filter grammar means exactly the same, checked against its source, not only the docs;
  • a not-equal gets '!null' beside it, since Lightdash's lets nulls through;
  • an any group or a date comparison becomes a CASE WHEN in sql, so the metric counts what its known value counts.
• Lossiness: what Lightdash has no place for is listed in the run's summary, the same way as Ossie. That covers an enum's members, a table's role and grain, a currency, a metric dated through another table, and a ratio whose inputs sit on two tables, which is left out.

Docs: https://docs.lightdash.com/references/metrics, /references/tables, /references/dimensions, /references/writing-descriptions, /integrations/dbt/metricflow.

Checked

• The full suite passes: 2175 tests. ruff, ruff format and ty are clean, and lightdash.py has 100% coverage.
• Every example's output validates against Lightdash's published JSON Schema for dbt YAML (lightdash-dbt-2.0.json, MIT, vendored with its licence).
• With dbt-core 1.12.3 and dbt-duckdb 1.11.0, dbt parse is clean and dbt build of the coffee example passes (79/79).
• @lightdash/cli 2.177.0 lightdash compile --no-partial-compilation succeeds for every explore. The test runs only when lightdash is on PATH, so CI skips it.
• Byte-identical without --lightdash: I diffed origin/main against this branch on three runs: with metrics, with --days 2 --defects messy, and with --unit-tests.

Not done

• No live deploy to a Lightdash instance. Lightdash can't open a local DuckDB file, so that needs Postgres or MotherDuck.
• dbt 1.11, the CI floor, wasn't run locally; CI covers it.

Release

• CHANGELOG 1.14.0, pyproject.toml 1.14.0 and the lock.
• After merge, draft the v1.14.0 release; publishing it uploads to PyPI.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!