Team SWIRL is excited to announce the release of SWIRL AI Search 4.3, Enterprise Edition — delivering more transparency, feedback, and global language support to enterprise AI search.
What’s New in SWIRL 4.3
Deep Linking from RAG & Chat to Sources
Transparency is key in enterprise AI. With SWIRL 4.3, every RAG (retrieval-augmented generation) and Chat response now comes with direct citations and deep links to the original sources. This makes it faster and easier for users to validate answers and trace information back to authoritative data.
User Ratings for RAG & Chat Responses
Feedback just got simpler. Users can now rate RAG and Chat Assistant responses directly in the Galaxy UI. These ratings are captured inside SWIRL, providing organizations with actionable insights through the SWIRL Activity Analyzer for quality tracking, governance, and continuous improvement.
Advanced Japanese Language Support
With SWIRL 4.3, we’ve introduced modern Japanese language processing across the entire platform — from query parsing and re-ranking to highlighting, RAG, and AI Assistant prompts. Enterprises can now deliver more accurate and relevant AI search experiences in Japanese.
Expanded Connector Support
SWIRL 4.3 introduces support for even more enterprise content systems:
- M-Files
- Better Regulation
- Custom SharePoint Collections
- Outline Team Knowledge Base
This expands SWIRL’s ability to deliver real-time, federated search across diverse enterprise repositories.
Additional Features & Improvements
- High Relevancy Option: Preserve native Elastic/OpenSearch ranking when only one source is used
- Logs Download: Export Celery-Worker and Django logs directly from the Admin UI
- Mapping Enhancements: Improved handling of unknown keys and quotes in mappings
- Assistant Query Support: AI Assistant can now query using SQL, MongoQL, Elastic/OpenSearch, and more
UI & Experience Enhancements
- Hidden RAG options if no AIProvider is active
- Exported Chat PDFs now include the SWIRL logo
- Smarter feedback when no results are relevant enough to trigger RAG
- Optional cookie-consent banner
- Fixed relevancy star assignment in non-Relevancy mixers
RAG & Search Improvements
- Follow-up questions now appear as Optional AI Instructions instead of firing new searches
- More reliable threshold handling with RoundRobinThreshold
- Improved resilience when SearchProviders are renamed
Technical Fixes
- Resolved SharePoint/OneDrive link issues with pound signs
- Validated compatibility with Apache Tika 3.2.1 and Python 3.13.5
- Replaced WebSocket interface for RAG with REST
Upgrading & Documentation
Our documentation site has been fully updated with details on all the new features and improvements.
Get Support
- 📧 Email: support@swirlaiconnect.com
- 🏢 Enterprise customers: open a support ticket
SWIRL 4.3 continues our mission to deliver secure, federated, and real-time AI search without ETL, duplication, or data movement. With transparency, feedback, and expanded global language capabilities, enterprises can move faster and more confidently into the future of AI.