nmdc-lakehouse-schema

The NMDC flat-schema product defines how nested source records become lakehouse tables. This package contains the schema generator, runtime flattener, schema comparison tools, and the canonical generated LinkML target schema.

Projection 1.3.0 retains TextValue strings directly on their containing records. For example, BiosampleFlat.host_diet is a multivalued string slot; there is no BiosampleSetHostDiet class or separate TextValue table. With source schema 11.24.0, the product has 61 table classes, including the two nested substance helpers introduced in 1.3.0.

• Generated schema reference
• TextValue projection and migration
• Mobile-phase substances and occurrence keys
• Supported transformations, helper tables, and known data-loss boundaries
• Schema ownership, versions, Just recipes, and documentation deployment
• Production TextValue and record-type audit

The source model is maintained in nmdc-schema. The nmdc-lakehouse application consumes the projection to write and validate data; adopting a new package and migrating existing data are separate from publishing this documentation.