What's New in Ayfie Saga 5.5.0

Created by Zbigniew Marszolik
Jul 24, 2023
2 min read

Ayfie Saga 5.5.0 has been released. Here are the highlights:

Content Handler SSL certificate

On initialization the Search UI checks if the Content Handler is installed by sending a request to the Content Handler. This request is using SSL. The SSL certificate in version 5.0.0 of the Content Handler expires September 12th 2023. After the expiry this check will stop working and the Search UI will assume the Content Handler is not installed. The Content Handler needs to be upgraded on the clients before the expiry of the previous SSL certificate. Version 5.0.1 of the Content Handler contains an SSL certificate that expires July 29th 2024. See the Content Handler section of the https://ayfie-dev.atlassian.net/wiki/spaces/SAGA/pages/2400714758 on how to upgrade the Content Handler.

Processing Pipeline

  • Several performance improvements have been done to the Document Service in regard to the management of the Processing Pipeline.

  • Database queries responsible for creating enrichment requests have been optimized.

  • A new boolean configuration option AYFIE_PIPELINE_AUTO_ENRICHMENT_QUEUE has been added. The default value is true (enabled). Setting this option to false will prevent automatic enrichment request migration for all documents when migrating from Saga 4.x to Saga 5.x. If disabled, one should manually add enrichment requests for all documents added to the system before migrating to Saga 5.x.

Lingo Extraction

Lingo could earlier not extract more than 1MB of data from a single document due to Kafka queue message limitations. That limit has now been increased to 10 MB.

Smart Classifier Changes

  • The Smart Classifier is now using the main product database (SQL Server or Postgres depending on the configuration) instead of a local SQLite database.

  • The Smart Classifier throughput is increased by the introduction of parallel document processing. The number of parallel threads is set by AYFIE_SMART_REFINER_CONSUMER_COUNT with 4 as the default value. (NOTE: The number of Smart Classifier ML workers set by AYFIE_SMART_REFINER_ML_WORKER_COUNT should not be set lower than AYFIE_SMART_REFINER_CONSUMER_COUNT).