AIDB 7.3.2 release notes v7
Released: 17 Apr 2026
AIDB 7.3.2 includes the following enhancements and bug fixes:
| Type | Description |
|---|---|
| Enhancement | Pipeline DLQ — Phase 3: Metrics and error log API: Expanded the Dead Letter Queue with a metrics and error log API, providing programmatic access to per-pipeline error counts, error details, and historical error logs. |
| Enhancement | Multi-user pipeline PGD trigger wrappers: Added multi-user aware trigger wrappers for PGD pipelines, ensuring that pipeline auto-processing triggers work correctly in multi-user PGD environments. |
| Enhancement | Knowledge base tables in PGD replication setup: Knowledge base tables are now automatically included in PGD replication configuration during setup, ensuring knowledge base data is replicated across PGD nodes. |
| Enhancement | Improved PDF parsing: Updated the lopdf library to a newer version with improved structural PDF parsing, resulting in better text extraction from complex and malformed PDF documents. |
| Enhancement | HTTP 429 retry handling: AIDB now automatically delays and retries requests that receive an HTTP 429 (Too Many Requests) response from model provider APIs, improving resilience when API rate limits are hit. |
| Enhancement | Deduplicate retrieve results by source key: Retrieve operations now deduplicate results based on the source key, preventing duplicate records from appearing in retrieval output when the same source document is indexed multiple times or referenced across multiple pipeline runs. |
| Bug fix | Fix handling of wrong input type errors: Input type validation errors are now caught and surfaced correctly at the input stage rather than propagating as unexpected errors downstream. |
| Bug fix | Fix multi-pipeline ID bug: Corrected an issue where pipeline IDs were incorrectly assigned in multi-pipeline configurations, which could cause pipelines to be misidentified or executed incorrectly. |
| Bug fix | Fix auto-processing state bug: Resolved a state management issue in the pipeline auto-processing system that could cause pipelines to enter or remain in an incorrect processing state. |