Skip to main content

Release v1.20.0

D
Written by Daniel Schindler

Version 1.20 – Release Notes

We’re excited to share the 1.20 release, which brings improvements across performance, search and retrieval accuracy, reliability, and user experience. Highlights include enhanced retrieval capabilities, expanded auditing and security features, and multiple usability improvements across our products.


Cortex

Enhancements

  • Expanded retrieval capabilities to support the full Data Fabric, with improved controls over result limits.

  • Added support for a new ranking algorithm to improve response relevance.

  • Generation responses now include clearer completion status and richer reasoning details (for supported models like GPT-OSS and Llama 4).

Fixes & Improvements

  • Improved backward compatibility for embedding responses.

  • Stabilized embedding response formats to ensure consistent integrations.


CoAuthor

Enhancements

  • You can now search prompts directly within your document.

  • Source selection is easier with document search available when building prompts.

  • Expanded audit logging, including new APIs, persistent storage, and access tracking.

  • Added API key support for Phoenix integrations.

  • Updated default prompt settings to improve generation quality out of the box.

Fixes

  • Improved performance of the Word add-in.

  • More accurate PDF processing and text extraction.

  • Cleaner, more consistent button styling across the CoAuthor web app and Word add-in.

  • Better handling of structured and tabular data for more accurate responses.

  • Increased reliability across Narrative workflows, including dropdowns, previews, tables, headers and footers, sorting, and time displays.


Certara GPT

Enhancements

  • Ability to provide tokenization per metadata field for increased accuracy.

  • Improved performance and reliability through backend client refactoring.

  • Clearer model discovery with indicators for internal vs. external models.

  • Added improved reranking support for more relevant results.

  • Added the option to run the application using only external models.

  • Reduced unnecessary processing to improve overall efficiency.

  • Citations can be returned when using Certara Generative AI in Composer or via the API.

Fixes

  • Improved validation and type checking.

  • Increased ingestion reliability by avoiding unnecessary duplicate checks.

  • Fixed token handling issues affecting certain model types.

  • Corrected scoring behavior during reranking.

  • Reduced noise in system health logs.

  • Restored compatibility by reverting recent identifier structure changes.


Connect

Enhancements

  • Administrators can now create service accounts.

  • Updated welcome email subject lines for improved clarity.

Fixes

  • Prevented API client identifiers from appearing in logs.

  • Resolved a security issue related to invite code usage.

  • Ensured notification emails are sent only to valid user accounts.


Synapse

Enhancements

  • Added a dedicated search node for improved data retrieval workflows.

  • Introduced a new example flow demonstrating retrieval-augmented generation.

  • Added light and dark mode support.

  • Added an “About Synapse” information panel.

  • Enabled single sign-on (SSO) login support.

  • Added encryption options for exported flow files.

  • Introduced a new JavaScript Filter node for more flexible workflows.

  • Improved error handling by surfacing clearer error details in the user interface.


General

Fixes

  • Updated Java version to resolve critical CVEs.

Did this answer your question?