Four tables indexed from the science.alt.dataset.* namespace, plus cursor state for firehose crash recovery. All tables use (did, rkey) as composite primary key. Schema auto-applies on startup.
Stores dataset schema definitions. Record key format: {schema-id}@{semver} (e.g., my.schema@1.0.0).
Fields: name, version, schema_type (default jsonSchema), schema_body (JSONB), description, metadata (JSONB).
Stores dataset metadata records. Includes weighted full-text search (search_tsv) across name (A), description (B), and tags (C).
Fields: name, schema_ref (AT-URI to schema), storage (JSONB — storageHttp, storageS3, or storageBlobs), description, tags (text array), license, size_samples, size_bytes, size_shards, content_metadata (JSONB).
Named version pointers to dataset entries (analogous to git tags).
Fields: name, dataset_uri (AT-URI to entry), version, description.
Bidirectional schema transforms with executable code for migrating data between schema versions.
Fields: name, source_schema (AT-URI), target_schema (AT-URI), getter_code (JSONB), putter_code (JSONB), description, language.