What's Changed
* Added "Safe AI with Flexible Evaluation" Tests (SafeTest) for three hazard types.
* We now include a `source_id` in PromptWithContext, linking that prompt back to the ID used in the original data source.
* Reworked how Secrets work internally, making their use much more explicit.
* You can now see all possible secrets with `list-secrets` command.
* Lots of error messaging improvements, including:
* If you are missing any secrets necessary a run, you are notified immediately. Error message provides guidance for how to add them.
* TogetherAI provides more detailed error messages.
* When exceptions happen in SUTs, report what the SUT was working on.
* HuggingFaceSUT can now use [Gated Models](https://huggingface.co/docs/hub/en/models-gated)
**Full Changelog**: https://github.com/mlcommons/newhelm/compare/v0.2.0...v0.2.1