schemancer

schemancer scans a flat folder of CSV, TSV, and Parquet files, infers a registry, adds conservative relationship hints, and generates a standalone Python package with SQLAlchemy models and load helpers.

CLI

schemancer scan <source-dir> --registry-out <path>
schemancer infer-relationships <registry-path>
schemancer generate <registry-path> --output-dir <dir> --package-name <name>
schemancer build <source-dir> --output-dir <dir> --package-name <name>

Example With test_data

From the repo root:

schemancer scan --registry-out new_schema ~/CODE/schemancer/test_data

If new_schema is an existing directory, this writes:

new_schema/registry.json

Then infer relationships into the same file:

schemancer infer-relationships new_schema/registry.json

Generate a standalone package:

schemancer generate new_schema/registry.json --output-dir generated --package-name my_schema

Or do the whole flow in one command:

schemancer build /Users/z3061723/Documents/CODE/schemancer/test_data --output-dir generated --package-name my_schema

After generation, the package will be at:

generated/my_schema

Development

uv run --extra dev pytest