SYNAPSE
What’s New
Layar Search Node — Search Layar documents directly within your flows using a dedicated node
Inline Document Grounding — Attach sources directly in the Prompt Node without requiring a separate node
Encrypted Flow Export — Export flows with optional password protection
About Synapse — New in-app informational modal accessible from the UI
Improvements
Flows now automatically save to the server
Session tokens refresh during autosave, preventing interruptions in long-running flows
Enhanced help text across key fields, including Number of Results, Output Type, and Document IDs
HTTP error responses now include detailed messages for easier debugging
Fixes
Resolved an issue where Unicode characters were corrupted during flow saves
Fixed a bug preventing flows from overwriting existing saved files
Inspect Node now correctly displays output from the Filter Node
Improved parsing of LLM outputs with relaxed JSON formatting
CORTEX
What’s New
Pipeline Execution Engine — New backend for executing pipelines
Layar Search Pipeline Node — Retrieve Layar documents dynamically during pipeline execution
SharePoint Search Pipeline Node — Search SharePoint content within pipelines
Pipeline Template Management — Create, search, update, and delete pipeline templates
Map-Reduce Mode — Execute pipeline branches in parallel for faster processing
Microsoft / SharePoint OAuth — Unified OAuth connector with shared token support
SMILES Structure Indexing — Automatically index extracted chemical SMILES structures into the vector database
Improvements
Streaming responses from OAI endpoints are now handled correctly without buffering issues
Emoji rendering improved for OAI endpoint prompts
Map-Reduce endpoint now supports raw text and tabular inputs
Deprecated model selections now return clear guidance instead of errors
Source identifiers are randomized to improve isolation between pipeline runs
RabbitMQ connections automatically recover after disruptions
Removed / Deprecated
Removed legacy chemistry integrations (chemprop, chemrxn)
Removed Triton-based NER integrations
Removed ManifoldCF live source connectors
Removed deprecated ontology and concept APIs
CERTARA GPT
What’s New
COLBERT Embedder — Support for late-interaction embeddings to improve retrieval quality
Fixes
Fixed validation issues with tabular documents
Corrected document filtering by date
COMPOSER
What’s New
Large Document Processing Modal — Real-time progress feedback for large document workflows
Improvements
Default timeout increased from 5 minutes to 3 hours to support long-running tasks
Fixes
Developer Settings now apply by default
Retry logic for large document processing works as expected
Tables are now compatible with map-reduce workflows in the UI
LAYAR
What’s New
Document Sets Pagination — Support for paginated views beyond 25 rows
Uploaded document titles no longer override user-defined names
Fixes
“Check All” now correctly toggles to “Uncheck All”
“Create Embeddings” option properly hidden for read-only users
“Date Added” filter now works accurately
Parent document previews load correctly
Fixed SharePoint connector integration issues
Removed
Deprecated NER and ML project model options
NER concept highlighting in the document view
Ontology tools and related navigation
CURATE
Improvements
Default extraction model updated to LLM-based model
Fixes
Corrected trial data filtering behavior
Removed
Column Summary visualizations
QA Fine-Tuning Projects section
Batches and My Assignments section
COAUTHOR
What’s New
Veeva Vault Integration — Full lifecycle support including import, export, OAuth, and metadata handling
Enhanced Attachment Validation —
validateAttachmentsnow returns source and section namesImproved Audit Log Retrieval — More reliable historical audit log searches
Improvements
Font formatting no longer defaults incorrectly to size 10
More robust async source deletion prevents partial states
Improved handling of long documents in WebKit browsers
Attachment validation is now scoped per project for better data isolation
Upload workers skip already-deleted sources
Prompt library upgraded to v5
Fixes
Corrected source count discrepancies with Veeva Vault
Resolved null pointer errors on deleted sources
Fixed system prompt and role-based access logic
Security
Applied additional CVE remediations
CONNECT API
Improvements
Improved OAuth stability
Added rollback script for OAuth migrations