We’re excited to announce the release of Version 1.18.0, delivering major enhancements across the Certara.AI ecosystem. This update introduces expanded model support, multi-modal capabilities, performance optimizations, and critical security improvements—reinforcing our commitment to flexibility, scalability, and reliability across all modules.
Certara GPT (certara-gpt)
Features:
Plug-and-play integration for Hugging Face LLMs and OpenAI open-source models on Hopper series
Improved keyword tokenizer configuration for metadata properties in Weaviate
Parameterized GPU memory and context settings for vLLM deployments
Integrated BGE reranker model for improved retrieval augmented generation
Refined deployment process and removed deprecated reranker modules
Increased text cap on map-reduce pipeline to support larger documents.
CoAuthor API
Features:
Placeholder support for source content placement
Universal subject ID naming for generated narratives
Enabled 21 CFR Part 11 compliance for electronic records and signatures in CoAuthor
Added support for formatted answers (including symbols, superscript, and subscript) in CoAuthor prompts/chats for Phoenix documents
Updated status endpoint to return total number of prompts and number completed
Fixes:
Table extraction bug resolved
Dataset attachment re-enabled
Login failures due to mixed-case emails fixed
Bulleted list formatting in Word integration corrected
Fixed CoAuthor document not retaining prompt-applied styles after inserting formatted answers
Resolved issues with customer PDFs failing to load due to parsing errors
Error Handling:
Improved status reporting for file processing issues
Fixed narrative generation errors related to field validation and coloring
Performance:
Removed token limit parameter for optimized LLM call handling
CoAuthor Web
User Experience Enhancements:
Improved attachment handling
Enhanced prompt UI behavior
Consistent markdown formatting
Bulk select in license modal
Visual feedback for “Apply to Document” actions
Cortex API
Multi-Modal Support:
Added image processing capabilities with compatible LLMs
Search Enhancements:
Reranker multiplier query parameter
Enhanced Certara connector data in search results
Integration:
New endpoints for Synapse backend
Improved error responses and flow saving logic
Performance Tuning:
Configurable thread count for better scalability
Certara Model Server
Features:
Users can now deploy their own fine-tuned embedding models
Performance:
Reranker model now runs on GPU for faster RAG operations
Fixes:
Resolved build/runtime issues from merge conflicts
Cortex Data
Clinical Trials Fixes:
Missing text issue resolved
Full stack updater implemented to fix data format inconsistencies
Vyasa Account
User Management Enhancements:
Updated UI messaging for account locking
Auto-refresh after bulk CSV uploads
Security:
Invite code validation improved during registration
Synapse
Testing & QA:
Comprehensive integration testing suite
Example flow for Filter Node functionality
Security Updates:
Enhanced session key validation
SSRF vulnerabilities addressed
Bug Fixes:
Flow name validation improved
HTTP response handling corrected
Certara GPT Web
Enhancements:
Improved composer functionality
Interface refinements