Tech Stack

NoSQL DB provider supported

• Elasticsearch versions

• 8.

• 12.

• 2
• Opensearch v. 1.1, 2.8

Java supported

• OpenJDK 17

Python supported

• Python version 3.11.

• 9

Node.js (for development)

• Node.js LTS v. 18.17.1.

Improvements

Framework

• Saga logger: Ability to log into NoSQL DB (in addition to filesystem)

Docker/Kubernetes

• Several improvements to docker file/image to make it easier to deploy and work Saga in Kubernetes.
  • Dynamic load configuration (env file, path, dynamic environment variables for Saga and Python Bridge)

Recognizers/Processors

• Entity Recognizer - New flags (TOTAL_MATCH and PARTIAL_MATCH). 
• Python Model Recognizer - Added "Match Threshold" option.
• Python Model Recognizer - Inform hostname error properly.
• Python Classification Watcher/Python Model Recognizers - Improved authentication process with Python Bridge. 
• Spellchecker stage using now the abstract indexer (it was connecting directly to Elasticsearch before.) Known issue on 1.3.3.

UI/UX

• Deprecated warning added in Rules/Executors screen. Rules/Executors only work with old ESUI not GAIA.

Vulnerabilities

• Fixed critical, high, and medium vulnerabilities found by security scans on both Saga Server and Python Bridge. Remaining some vulnerabilites related to CIO Docker Images out of our control. 

Bug fixes

Server

• Better error message on start-up when connecting with NoSql provider (Elasticsearch/Opensearch).
• Error message about Processing Units mapping during start-up when using Processing Units.
• Multiple random ports opened by the Elasticsearch client library used by Saga.
• Load Processing Units only after Resources have been loaded. 
• State cannot be determined error when using SSO login.
• Connection error messages not showing correctly for maximum retries. 
• Fixed processing units mappings.  
• Elasticsearch connections not being released. 

Evaluations Page

• Statistics not being rendered after evaluating. Known issue on 1.3.3.

Framework/API

• Stop endpoint path fixed. 
• Dictionary process endpoint used by Aspire Saga Publisher is duplicating entries. 

Saga Parser

• Using the Python Bridge with SSL or basic authentication was not connecting. Known issue on 1.3.3.
• NPE error when enabling the Debug option. Known issue on 1.3.3.

Tags/Recognizers

• PostalCode Recognizer - country selected was not working recognizing postal codes.
• Intent/FAQ: Recognizers using the Python Bridge with authentication failing on the Recalculating vector option.
• Python Classification Recognizer - Minor issues and improvements in the UI.
• Check duplicate setting added to Entity and FAQ recognizer importers.
• Fixed [object Object] message recalculating with no entries for the Intent Recognizer. 

UI/UX

• Check Tag button in preview screen

• was not working

• . Known issue on 1.3.3.
• Server Down popup message constantly appearing. Known issue on 1.3.3.
• Refresh issue having multiple character splitters. 
• Export for Data Science - null error trying to export - dataset with special characters (&).
• Universal Importer - CSV for Intent Recognizer not loading dictionary. 

Note: you can find all the details in Jira here.