Skip to main content

3.18 Release Notes

3.18.11​

New Features​

  • Added model encyption for improved platform security

3.18.10​

New Features​

  • Metadata logging added to pod logs

Bug and Vulnerability Fixes​

  • Resolved fluentbit logging errors

3.18.9​

Improvements & Enhancements​

  • Updated prompt injection default guardrail

3.18.7​

New Features​

  • [Beta] Set guardrail classification thresholds for finer control over guardrail performance
  • [Beta] Generate billing reports based on token utilization

3.18.6​

Improvements & Enhancements​

  • Improved policy description readability

3.18.4​

Improvements & Enhancements​

  • Streaming support improvement for messages with longer token lengths

3.18.2​

New Features​

  • Support for editing, retraining, and versioning output content policies

3.18.1​

Improvements & Enhancements​

  • Improved policy creation page format

Improvements & Enhancements​

  • Fixed feedback column incorrectly populated issue

3.18.0​

New Features​

  • Role Based Access Control. Set up and manage user roles to control access to platform resources, such as models and policies.
  • Share models and policies directly with other platform users
  • Edit and retrain input content policies
  • Leave feedback on monitoring logs, use logs for retraining
  • Input content policy version control
  • Manually upload datapoints to policy training datasets
  • Build custom output content guardrails
  • Apply guardrails to streaming outputs
  • Specify custom metadata fields to filter by on the get/logs/ endpoint
  • Support for in-cluster synthetic data generation models