AgentOps What's new
Explore major updates by month, or read the full GitHub release notes for all changes and fixes.
October 2026
| Release |
Highlights |
| v0.16.0 |
Review recorded evaluation-checkout commits and tracked configuration changes alongside regressions in reports and Doctor, and browse evaluated-version history in Cockpit. These comparisons do not prove a regression's cause or verify the deployed source. |
September 2026
| Release |
Highlights |
| v0.15.0 |
Added recipe-based azd evaluation integration for azd ai eval, preserving AgentOps results and threshold gates without changing existing azd ai agent eval recipes. The new path requires the preview azure.ai.evaluations extension, which was not yet publicly available at release. |
August 2026
| Release |
Highlights |
| v0.10.0 |
Read evaluation datasets directly from private Azure Blob Storage or ADLS Gen2 using your existing Azure identity instead of managing local copies. Objects are limited to 100 MiB and require data-reader permissions and network access, with no SAS or account-key support. |
July 2026
| Release |
Highlights |
| v0.8.0 |
Inspect evaluator pass rates, raw scores and failed evaluation events in the Agent behavior tab of the Foundry operations workbook. The tab reads compatible Application Insights telemetry and shows missing-data diagnostics, but does not instrument agents or schedule evaluations. |
June 2026
| Release |
Highlights |
| v0.6.0 |
Turn production traces from Azure Monitor or Application Insights into reviewable evaluation datasets with telemetry imports. Preview rows before writing them and review labels before gating a release, since self-similarity detects drift rather than proving correctness. |