Export Contract Fields

Exhaustive field-by-field export contract catalog generated from the checked-in inventory.

Inventory schema: dagzoo-export-contract-inventory-v1
Machine-readable source of truth: reference/export_contract_inventory.yaml
Path patterns use * for dynamic map keys and [] for list item shapes.
audit_status is a field-review classification only; it does not change the live export surface.

dataset_catalog.parquet record payload

Canonical per-dataset JSON payload stored in each generated shard catalog row.

Path	Type	Presence	Stability	Producer	Audit	Known Consumer / Rationale
`dataset_id`	`string`	always	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	stable canonical dataset id
`dataset_index`	`int`	always	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	record index within the generated run
`feature_types`	`list[string]`	always	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	emitted feature typing
`feature_types[]`	`string`	when feature count > 0	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	per-column emitted feature type
`group_ids`	`object`	when run grouping ids are available	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	stable downstream grouping keys
`group_ids.cohort`	`string`	when heterogeneous cohort grouping ids are available	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	shared heterogeneous raw-generation cohort key
`group_ids.layout_plan`	`string`	when fixed-layout grouping ids are available	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	shared layout-plan grouping key
`group_ids.request_run`	`string`	when group_ids present	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	request-run grouping key
`intervention`	`object`	when hard-intervention metadata is available	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	summary-only intervention contract for downstream consumers
`intervention.mode`	`string`	when intervention present	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	emitted intervention regime
`intervention.signature`	`string`	when intervention present	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	stable intervention identity summary
`n_classes`	`int	null`	always	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`
`n_features`	`int`	always	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	persisted feature count
`n_test`	`int`	always	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	persisted test row count
`n_train`	`int`	always	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	persisted train row count
`target_derivation`	`string`	when target derivation metadata is available	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	current target-derivation contract marker
`target_relevance`	`object`	when target relevance audit metadata is available	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	observed feature relevance summary for the latent target
`target_relevance.feature_count`	`int`	when target_relevance present	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	count of emitted features with a path into the target node
`target_relevance.feature_fraction`	`float`	when target_relevance present	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	feature_count normalized by emitted feature count
`task`	`string`	always	`stable`	`dagzoo.io.shard_contract.build_dataset_catalog_record`	`keep`	dataset task label

Packed parquet split row

Row schema shared by shard train.parquet and test.parquet.

Path	Type	Presence	Stability	Producer	Audit	Known Consumer / Rationale
`dataset_index`	`int64`	always	`stable`	`dagzoo.io.parquet_writer._build_split_table`	`keep`	row-to-dataset join key
`row_index`	`int64`	always	`stable`	`dagzoo.io.parquet_writer._build_split_table`	`keep`	row position within one split
`x`	`list[float32	float64]`	always	`stable`	`dagzoo.io.parquet_writer._build_split_table`	`keep`
`x[]`	`float32	float64`	when feature count > 0	`stable`	`dagzoo.io.parquet_writer._build_split_table`	`keep`
`y`	`int64	float`	always	`stable`	`dagzoo.io.parquet_writer._build_split_table`	`keep`

Generate handoff manifest

Minimal downstream handoff manifest written by dagzoo generate –handoff-root.

Path	Type	Presence	Stability	Producer	Audit	Known Consumer / Rationale
`artifacts_relative`	`object`	always	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	manifest-relative artifact paths
`artifacts_relative.curated_dir`	`string`	when curated accepted-only shards exist	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	relative curated-dir path
`artifacts_relative.generated_dir`	`string`	always	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	relative generated-dir path
`identity`	`object`	always	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	stable generated corpus identity payload
`identity.generate_run_id`	`string`	always	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	stable generate-run id
`identity.generated_corpus_id`	`string`	always	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	stable generated corpus id
`identity.source_family`	`string`	always	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	handoff source family tag
`provenance`	`object`	when generated catalogs expose corpus-level provenance	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	summarized latent-target provenance for downstream corpus consumers
`provenance.intervention`	`object`	when hard-intervention provenance is present	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	summary-only intervention provenance for downstream corpora
`provenance.intervention.mode`	`string`	when provenance.intervention present	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	generated corpus intervention regime
`provenance.intervention.signature`	`string`	when provenance.intervention present	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	stable intervention identity summary for the generated corpus
`provenance.target_derivation`	`string`	when provenance present	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	current target-derivation contract marker
`provenance.target_relevant_feature_count_range`	`object`	when provenance present	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	min/max relevant feature counts across the generated corpus
`provenance.target_relevant_feature_count_range.max`	`int`	when provenance.target_relevant_feature_count_range present	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	upper bound of observed relevant feature counts
`provenance.target_relevant_feature_count_range.min`	`int`	when provenance.target_relevant_feature_count_range present	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	lower bound of observed relevant feature counts
`provenance.target_relevant_feature_fraction_range`	`object`	when provenance present	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	min/max relevant feature fractions across the generated corpus
`provenance.target_relevant_feature_fraction_range.max`	`float`	when provenance.target_relevant_feature_fraction_range present	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	upper bound of observed relevant feature fractions
`provenance.target_relevant_feature_fraction_range.min`	`float`	when provenance.target_relevant_feature_fraction_range present	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	lower bound of observed relevant feature fractions
`schema_name`	`string`	always	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	handoff manifest schema identifier
`schema_version`	`int`	always	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	handoff manifest schema version
`summary`	`object`	always	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	generated dataset count summary
`summary.generated_datasets`	`int`	always	`stable`	`dagzoo.core.generate_handoff.write_generate_handoff_manifest`	`keep`	number of generated datasets

Coverage summary JSON

Stable top-level diagnostics coverage summary fields.

Path	Type	Presence	Stability	Producer	Audit	Known Consumer / Rationale
`generated_at`	`string`	always	`stable`	`dagzoo.diagnostics.coverage.write_coverage_summary_json`	`keep`	diagnostics summary timestamp
`histogram_bins`	`int`	always	`stable`	`dagzoo.diagnostics.coverage.write_coverage_summary_json`	`keep`	histogram bin count used for diagnostics
`num_datasets`	`int`	always	`stable`	`dagzoo.diagnostics.coverage.write_coverage_summary_json`	`keep`	number of aggregated datasets
`quantiles`	`list[float]`	always	`stable`	`dagzoo.diagnostics.coverage.write_coverage_summary_json`	`keep`	requested quantile levels
`quantiles[]`	`float`	when quantiles present	`stable`	`dagzoo.diagnostics.coverage.write_coverage_summary_json`	`keep`	one requested quantile level
`task_counts`	`object`	always	`stable`	`dagzoo.diagnostics.coverage.write_coverage_summary_json`	`keep`	aggregated task counts
`task_counts.*`	`int`	when task_counts present	`stable`	`dagzoo.diagnostics.coverage.write_coverage_summary_json`	`keep`	one aggregated task count