--model¶
Why we added this flag: every dataset run needs to know which Pydantic model registry to validate against; making it explicit avoids accidentally targeting the wrong schema bundle.
What it does¶
- Accepts a Python import string (
package.module:ATTRIBUTE) that resolves to a mapping of model names toModelSpecobjects. - Drives schema validation, field listing, and schema signature hashing for manifest tracking.
- Required for all
idxr prepare_datasetsinvocations.
Typical usage¶
idxr prepare_datasets \
--model kb.std.ecc_6_0_ehp_7.registry:MODEL_REGISTRY \
--config configs/hotfix.json \
--output-root build/partitions
Tips¶
- Reference the same registry for prepare and vectorize so schema signatures stay aligned.
- Use environment variables if you switch registries frequently:
export IDXR_MODEL=kb.std...and pass--model "$IDXR_MODEL".