Skip to main content

Release v1.16.0

K
Written by Kay Chen

We’re excited to announce the release of Version 1.16.0, which brings a wide range of enhancements across the Certara GPT ecosystem, including performance improvements, expanded API capabilities, UI refinements, and critical security updates. This release reflects our continued commitment to improving usability, scalability, and developer experience across all modules. Below is a detailed summary of the updates:

Certara GPT (certara-gpt)

  • Metadata Injection Enhancements: Added support for injecting metadata when using entire documents as context; improved document stub generation and delimiter handling.

  • Search Filters: Now supports nested metadata filtering.

  • Performance Improvements: Reduced worker requirements for high concurrency; significantly lowered failure rates under load.

  • Summarization: Introduced new parameters to customize map-reduce prompts in Composer/API.

  • Request Handling: Request cancellations now propagate correctly to the model server.

  • Tokenizer Refactor: Improved tokenizer management for better reliability.

  • Call Argument Refactor: Enhanced structure and robustness of GPT call arguments.

CoAuthor API

  • GPT Attachments: Users can now attach GPT-generated content to dataset files.

  • Searchable Children: Added searchable flag to child objects.

  • IQ Endpoint Updates: Refined endpoints and removed deprecated category field.

  • Prompt Enhancements:

    • Support for a fourth prompt type.

    • Improved parsing from content controls.

  • Phoenix Templates: Added support for Phoenix-style templates.

  • Bulk Downloads: Improved logic for recursive file and folder downloads.

  • Markdown Conversion: New endpoint for converting Markdown to Word format.

  • Debug Logging: Added scripts and improved logging structure.

CoAuthor Web

  • Attachment Handling:

    • Improved error handling for invalid attachments.

    • Dataset preview and max file size alerts.

  • Prompt UI Enhancements:

    • Scroll position memory.

    • Cursor placement synchronization.

    • Dirty form detection.

  • Markdown Formatting: Consistent formatting across add-in and Word document insertion.

  • Admin Dashboard: Bulk select added to license modal; other UI improvements.

  • Loading Animation: Visual feedback added for "Apply to Document" actions.

CoAuthor WebDAV

  • Security Fixes: Resolved known vulnerabilities.

Cortex API

  • RAG Fallback: Added fallback logic for bulk questions.

  • Prompt Enhancements: Support for nested conditions and non-mandatory fields.

  • URL Ingestion: Improved formatting preservation.

  • Async Improvements: Queued async requests with cancellation support.

  • Model Support: Added support for .tar and .tar.gz files.

Cortex

  • Security Updates: Upgraded dependencies to address CVEs.

  • Grobid Versioning: Updated and reverted Grobid versions for compatibility.

Vyasa Chemistry

  • Model Loading: Extended support for scikit-learn trained models.

  • Error Handling: Improved handling of invalid SMILES strings.

  • DevOps Integration: Streamlined build and deployment processes.

Curate

  • Prompt UI Tweaks: Enhanced layout and consistency.

  • Saved List Enhancements: Added copy options and improved descriptions.

  • Question Re-Prompting: Improved UI feedback for re-asking questions.

Vyasa Proxy

  • Health Check: Landing page now returns 503 for health check endpoint.

  • Build Automation: Docker builds now triggered on commits and tags.

Bug Fixes & Enhancements

  • Fixed Issues:

    • Nested filters in single documents.

    • GPT model packaging and unused imports.

    • Document parsing listener timeout.

    • Prompt/attachment updates not reflecting immediately.

    • CVEs across multiple services.

  • Refactored:

    • Tokenizer handling.

    • Logging and Docker build scripts.

    • Manifest and Helm chart configurations.

Did this answer your question?