Cloudera Docs

RAG Studio - 2.1.0

This version of Cloudera AI Studios provides you with several new capabilities. Learn how the new features and improvements benefit you.

New Features/Improvements

RAG Studio installation in airgapped environments: Cloudera RAG Studio now features a containerized ML Runtime Image deployment method, streamlining installation in air-gapped environments by replacing legacy source code builds. This comprehensive image is pre-packaged with all necessary dependencies, such as Python and Node.js, to enable immediate execution upon deployment. For more information see, Installing RAG Studio in Air-gapped environments.

Fixed issues

  • Previously, the RAG Studio containers had multi-minute cold starts and downloaded large ML and CUDA dependencies, such as PyTorch, NVIDIA cu12, and RAG frameworks at Runtime because the prebuilt virtual environment was not utilized. This issue is now resolved and the Docker image is now updated to correctly use the build-time virtual environment. (DSE-54451)
  • Fixed an issue where Cloudera AI Workbench version format parsing failed on PVC 1.5.5 SP2 (for example, non‑PEP 440 version strings) and caused RAG Studio to freeze. (DSE-54589)
  • Previously, RAG Studio failed to list endpoints or complete model discovery when Cloudera AI Inference (CAII) was enabled, resulting in an AttributeError during endpoint processing. This issue is now resolved and this fix ensures reliable and performant endpoint listing and model discovery within RAG Studio. (DSE-54373)
  • Fixed an issue where RAG Studio had authentication errors (missing /tmp/jwt) that prevented accessing CAII REST API endpoints or Studio settings. (DSE-55009)