Release notes
v1.2.1 | Sept 29, 2025
Fixes
- Fixed critical vulnerabilities.
- Resolved several bugs to improve overall stability.
v1.2.0 | Aug 26, 2025
New features
RBAC (Sharing)
You can now share resources such as workflows, evaluations, or tests directly in the UI through integration with the Authz service.
MRM artifacts and actionability
MRM Workflows now produce various output artifacts:
- Summary and report: Generate a comprehensive workflow report that you can ingest and explore in H2OGPTe.
- Corpus patch: A collection of FAQs based on the problematic prompts and their expected answers, which can be added to H2OGPTe collections to improve the RAG performance.
- Guardrails configuration: A downloadable configuration (JSON file) for custom guardrail, which you can embed into your application using the H2O Eval Studio Python client.
New MRM features
- The following MRM workflow features were introduced:
- Generated test validation: Validate the quality of the generated test suite using various metrics.
- Failure clustering: Automatically detect common failure patterns of the model host.
- Workflow statuses: Every workflow step now can report various colored statuses, such as failures, warnings or being stale, which means user should verify the outputs and re-run certain steps to get up-to-date results.
- MRM workflows now support other model hosts, including Bedrock and OpenAI-compatible APIs.
- MRM workflows are now accessible directly through H2OGPTe collection UI.
- Various UX improvements were made to workflows and their individual steps, including enhancements for robustness testing.
Feedback
- Submit and view feedback for this page
- Send feedback about H2O Eval Studio to cloud-feedback@h2o.ai