AIDB 7.3.0 release notes v7

Released: 08 Apr 2026

AIDB 7.3.0 includes the following enhancements and bug fixes:

TypeDescription
EnhancementPipeline DLQ — Phase 2: Nonblocking errors: The Dead Letter Queue now handles nonblocking errors, meaning that transient or non-critical processing failures no longer halt pipeline execution. Failed records are queued for inspection and reprocessing.
EnhancementTransient error classification: Model and database errors are now classified as PipelineTemporary where appropriate, enabling the pipeline to retry transient failures automatically rather than treating them as permanent errors.
EnhancementDLQ user APIs: Introduced user-facing SQL functions for interacting with the Dead Letter Queue, including the ability to list, inspect, and reprocess failed pipeline records.
EnhancementComments in semantic knowledge base search: Semantic knowledge base search now includes support for searching over comments stored alongside document content.

Deprecated interfaces

AIDB 7.3.0 removes the following SQL views and functions as part of the major version interface cleanup. Replacements are noted for each.

Deprecated objectsReplacement
Views
aidb.preparersaidb.pipelines (filter by operation type if needed)
aidb.prepsaidb.pipelines
aidb.retrieversaidb.knowledge_bases or aidb.kbs
aidb.preparer_registryaidb.pipeline_registry (internal)
aidb.preparer_registry_source_tableaidb.pipeline_registry (internal)
aidb.preparer_registry_source_volumeaidb.pipeline_registry (internal)
Functions
aidb.create_table_knowledge_base(...)aidb.create_pipeline(... step_1 => 'KnowledgeBase')
aidb.create_volume_knowledge_base(...)aidb.create_pipeline(... step_1 => 'KnowledgeBase')
aidb.set_auto_knowledge_base(...)aidb.update_pipeline(name, auto_processing => ...)
aidb.assert_knowledge_base_preconditions(...)Removed — no-op; validation is now handled internally
aidb.create_retriever_for_table(...)aidb.create_pipeline(... step_1 => 'KnowledgeBase')
aidb.create_retriever_for_volume(...)aidb.create_pipeline(... step_1 => 'KnowledgeBase')
aidb.delete_retriever(...)aidb.delete_pipeline(...)
aidb.set_retriever_auto_processing(...)aidb.update_pipeline(name, auto_processing => ...)
aidb.enable_auto_embedding_for_table(...)aidb.update_pipeline(name, auto_processing => 'Live')
aidb.disable_auto_embedding_for_table(...)aidb.update_pipeline(name, auto_processing => 'Disabled')
aidb.create_table_preparer(...)aidb.create_pipeline(...)
aidb.create_volume_preparer(...)aidb.create_pipeline(...)
aidb.create_preparer_for_table(...)aidb.create_pipeline(...)
aidb.create_preparer_for_volume(...)aidb.create_pipeline(...)
aidb.delete_preparer(...)aidb.delete_pipeline(...)
aidb.set_preparer_auto_processing(...)aidb.update_pipeline(name, auto_processing => ...)
aidb.get_preparer_data_source(...)Query aidb.pipelines view directly
aidb.assert_preparer_preconditions(...)Removed — validation is now handled internally
aidb.assert_table_preparer_preconditions(...)Removed
aidb.assert_volume_preparer_preconditions(...)Removed
aidb.bulk_embedding(name)aidb.run_pipeline(name)
aidb.bulk_data_preparation(name)aidb.run_pipeline(name)
aidb.single_embedding(name, key, text)aidb.run_pipeline(name) or direct model calls
aidb.single_embedding(name, key, bytea)aidb.run_pipeline(name) or direct model calls
aidb.set_auto_preparer(name, mode)aidb.update_pipeline(name, auto_processing => mode)
aidb.upsert_chunk_text(name, key, value)Use pipeline execution via aidb.run_pipeline()
aidb.upsert_summarize_text(name, key, value)Use pipeline execution via aidb.run_pipeline()
aidb.upsert_parse_html(name, key, value)Use pipeline execution via aidb.run_pipeline()
aidb.upsert_parse_pdf(name, key, value)Use pipeline execution via aidb.run_pipeline()
aidb.upsert_perform_ocr(name, key, value)Use pipeline execution via aidb.run_pipeline()