semql-introspect 0.2.1

Bootstrap a semql Catalog from a live database — emits Python cube stubs from Information Schema with heuristic measure/dimension inference.

semql-introspect

Bootstrap a semql Catalog from a live database.

Reads Information Schema, emits Python Cube stubs with heuristic measure / dimension / time-dimension inference and foreign-key derived joins. Designed for greenfield adoption — a team with 200 tables can generate the mechanical 80% of a catalog in seconds, then hand-edit the heuristic guesses.

Usage

import duckdb
from semql.model import Backend
from semql_introspect import introspect_to_python

con = duckdb.connect("warehouse.db")
print(introspect_to_python(con, backend=Backend.DUCKDB, schema="main"))

Or via CLI:

semql-introspect --backend duckdb --schema main --conn "warehouse.db"

Heuristics

• Numeric columns named amount / price / revenue / cost / total / value / qty / quantity / count → Measure(agg="sum").
• Columns ending in _id → Measure(agg="count_distinct") (the table's cardinality is usually interesting).
• date / timestamp columns → TimeDimension.
• Foreign keys → Join(relationship="many_to_one") plus the foreign-side Dimension(foreign_key=...).
• Everything else → Dimension typed by the column's SQL type.

Heuristic guesses get a # TODO: review comment so the diff makes the inference choices reviewable.