Filter prompts by their attached source, so you can run targeted bulk actions across a document.

Mark a document as prompt-enabled the moment you create it, and use CoPrompter to prompt-enable an eCTD regulatory template.

Single-file downloads now keep the original filename — no timestamp or ZIP wrapper — and multi-file downloads produce a tidy, clean-named ZIP.

Select ranges with Shift+click across the web-portal grids and the Word add-in pickers.

Export audit logs to PDF, and enjoy a CSV export that now gives every detail field its own column.

Cancel an AI PK report after you've requested it.

Inserting a raw TFL .rtf from Sources now automatically formats CSR in-text tables.

• The Sources display now orders document and section names for easier reading.

• Removed the Summarize and Past Tense prompt task types in the Word add-in to reduce clutter.

• Refreshed the auto-attach sources interface.

Add-in & document management

• Existing prompts and portal-visible sources now display correctly in the Word add-in.

• Add-in prompts are visible again in multi-user, multi-project setups.

• The Storage Configuration tab now appears without requiring both CAD and CoA admin roles.

• The List Settings modal no longer fails silently when saving or loading.

• The "About" screen now shows the correct version number in the web app and API.

Generation & sources

• PDF figures inserted from the Sources tab now render correctly instead of breaking into fragments.

• "Generate All" on header prompts no longer shows false "General exception" errors after a successful run.

• The source-selection checkbox no longer clears when you change the search query.

• Veeva imports no longer fail when a document name contains periods, and bulk downloads of Veeva-stored sources no longer crash.

• Batch ZIP downloads no longer rename the files inside them.

Narratives & formatting

• Conditional Text now works with integer columns.

• Data Display now correctly suppresses table-based sections when a source value is blank.

• Scroll-bar restyling now applies correctly.

• Improved authentication and transaction performance, and fixed a memory leak in source parsing.

• Consolidated numeric-ID handling in the Word document state service and hardened the API health check.

• Prevented out-of-memory worker crashes and restored log-noise suppression in document processing.

Reference LLM attachments by name directly in a prompt, rather than by position.

The SharePoint connector now tells you clearly when a site can't be reached or a search returns no results, instead of failing quietly.

• The "Show Details" view in Search now reports an accurate document count.

• Updated table-tag processing for document ingestion.

• Removed a leftover Google Tag Manager tracking snippet.

• New users can register again after a system wipe-and-reinstall.

• The model list now refreshes automatically once a new model finishes creating — no page reload needed.

• Cleaned up leftover UI clutter for a tidier workspace.

• Fixed a character-encoding bug in CSV exports.

• Codex now shows a clear "no valid datasets to load" message instead of failing silently.

• The "Contact us" link now points to the shared AI support mailbox rather than a list of individuals.

Vyasa branding and tab labels have been replaced with Certara.AI across the Portal, Curate, and Layar, and logos have been refreshed in Composer and Codex.

The support address is now AI-Support@Certara.com in the Portal and Layar.

• Updated the mailing address in community emails to the Radnor, PA headquarters.

• Resolved published CVEs in the Cortex API and closed out a related security-vulnerability review.

• The PII-scanner toggle is now controlled by an environment variable for easier per-environment configuration.

• Removed the deprecated "Concepts" feature from the API.

• Enabled a token-aware paragraph splitter with corrected sentence-chunk ordering for cleaner ingestion, and removed the Large Language Model entry from a Cortex-driven dropdown.

• Quieted noisy startup logs.

• Conversation history is now passed correctly through a chat session.

• Began consolidating Connect API's data-fabric logic into the Cortex API, simplifying the backend.

• Service accounts can now be renamed and managed by any admin.

• Removed the custom Map and Reduce prompts from the MapReduce node.

• Updated Curate Document Set handling in workflows.

• Fixed tensor parallelism on nvidia-container-toolkit 1.18+.

• Investigated job scheduling on Windows and refreshed the master-patents data connector.