Introducing Openlayer Governance
Introducing Openlayer Governance: the fastest way to track and enforce rules for your AI systems. With custom frameworks, you can define rules like:
- Test every system for PII leakage + prompt injection
- Require monitoring, tracing, logs, and production alerts
- Enforce owner, risk level, and approval status
Rules are tracked across projects and workspaces so you can easily see progress and spot gaps. Collaboration is made easy through assigning rule owners, comment threads, and uploading evidence that can be accessed by the whole team.
Governance is now available to everyone on Openlayer, so you can increase trust in your AI systems today.
Features
- Platform. Introducing Governance mode, a new way to define and enforce policies and procedures across projects in your workspace
- Platform. Brand-new workspace overview and project inventory, which gives users a birds-eye view of project activity and enables better categorization and labeling
- Platform. Support for more 3rd-party data backends in Monitoring mode, including Databricks, Redshift, and Postgres
- Platform. Enable running an A/B test across two commits in Development mode or data sources in Monitoring mode to compare performance across different metrics
- UI/UX. Support for configuring directory sync in workspaces with SAML/SSO enabled via the UI
- Platform. Custom onboarding for users purchasing Openlayer through the AWS Marketplace
- API. Enable deleting custom metrics via the API
- SDKs. Support for tracing OpenAI systems in Python that use the .parse() method or Responses API
- SDKs. Support for tracing systems built with Google ADK in Python
- SDKs. Allow users to configure traces in Python by defining settings like timeouts and retries
- CLI. Users can now create data sources directly from the CLI
- CLI. Users can now delete custom metrics directly from the CLI
Improvements
- Docs. Several improvements to the Openlayer Docs, including a brand new test catalogue, more integration guides for LLM frameworks like Google ADK and 3rd-party data backends like Snowflake
- API. Several improvements to the Openlayer API, including removing the need to specify an API version for requests to certain endpoints
- CLI. Several improvements to the Openlayer CLI, including easier project selection and support for profile management
- Platform. Better experience using 3rd-party data backends like Snowflake and Databricks, including quicker ways to connect multiple tables, inferring schemas, and support for remote execution of tests to scale better to large data volumes
- UI/UX. Enhanced navigation header for more rapid and seamless movement between projects and modes
- UI/UX. Several enhancements made to tables throughout the app [
- On-Prem. Shipped various tooling and improvements for users deploying Openlayer on-premise, including the ability to leverage internal LLM gateways
- UI/UX. Switching between login and signup forms no longer resets inputs
- UI/UX. Enable filtering columns in the Data page by clicking on the column header
- UI/UX. New flow for accepting tests generated by the Assistant, allowing users to select a subset of suggested tests, refine them, or retry requests
- UI/UX. More tooltips everywhere throughout the app!
- UI/UX. New-and-improved in-app support experience, which allows users to send feedback and reach out for help to standard channels
Fixes
- Platform. Squashed several bugs preventing tests from executing correctly
- UI/UX. Results graph inside a monitoring test would sometimes prevent users from navigating back in time to historical results
- UI/UX. Results page would sometimes show test results from other data sources in Monitoring mode or commits in Development mode than the currently selected
- UI/UX. Authors would not render in certain notifications
- UI/UX. Navigating to app settings would trigger toasts containing error messages unexpectedly
- UI/UX. Issue preventing some users from deleting their accounts
