Skip to content
Vibedata

What's new · v0.1.79

Ask the semantic model in your platform's SQL

The semantic query compiles level, time-series and dimension-rule metrics in the dialect of the Domain's platform: DuckDB, MotherDuck, Microsoft Fabric, Databricks or Snowflake.

A translator's dial on a powder field: a navy disc with a thin cyan rim. The center carries ONE QUESTION in white above the white phrase revenue by region. Three small tabs with cyan outlines sit around the rim, each with a ruled line of a different length. The lower-left tab ends in a powder tick.
One question, answered in the dialect your platform reads.

A compiled query used to come back in one dialect. It now comes back in the dialect of the platform the Domain runs on.

  1. Ask for a governed metric

    A conversation card on a powder field. A user bubble reads revenue by region. A reply bubble below it reads compiled for your platform.
    Ask for a governed metric in the conversation.

    Ask in the conversation. The semantic query reads the Domain's canonical semantic model and compiles the question against the metrics, entities and join keys the model declares.

  2. Let the model compile it

    A flow on a powder field: a question box reads revenue by region, an arrow leads to a semantic model box listing level, time-series and dimension-rule metrics, and a second arrow leads to an SQL box reading your dialect.
    The semantic model compiles the question in your platform's dialect.

    The compiler handles level, time-series and dimension-rule metrics and writes the SQL in the platform's dialect: DuckDB, MotherDuck, Microsoft Fabric, Databricks and Snowflake. A platform without a mapped dialect is refused with a reason.

  3. Read the SQL before you run it

    A result card on a powder field shows the SQL compiled in the platform's dialect, covering level, time-series and dimension-rule metrics, with undeclared relations refused. A footer line reads: run mode executes on DuckDB; elsewhere, a run call is refused and points to SQL mode.
    Read the SQL before you run it.

    Use SQL mode to get the statement in the platform's dialect. Run mode executes on DuckDB. On any other platform a run call is refused up front and points you to SQL mode.

Before Write the query for one engine, then rewrite its functions and identifier rules for another.After Ask for the metric in SQL mode and get the statement in the target platform's dialect.

The SQL is readable before it reaches a dashboard. Start with one governed metric your team argues about.