Launch Week 02 wrapped — explore all five launches

August 25, 2026

  • Enterprise
  • Self-Hosting

v2.5.2

This release adds Mantle project attribution support across Bedrock and evaluation usage flows, along with a few public API and UI updates. It also includes a fix for model validation timing and a malformed project ID rejection on the public API.

Highlights

  • Mantle project IDs can now be entered, read, and propagated through Bedrock model settings and pings.
  • Mantle evaluation and platform model usage are now attributed to a project.
  • Malformed Mantle project IDs are rejected on the public API.
  • Model validation now waits until the model id settles before running.

New Features

  • Bedrock Project ID Support — Bedrock model settings can now carry a Mantle project id, which is read from config and sent with Mantle pings.
  • Model Settings Project ID Input — The model settings UI now allows entering a Mantle project id.
  • Project Usage Attribution — Mantle evaluation usage and platform model usage are now attributed to a project.
  • Public API Validation — The public API now rejects malformed Mantle project ids.

Improvements

  • Public Endpoints Cleanup — New public endpoints were cleaned up as part of the release.
  • Version-Aware Paths — Version-aware path handling was updated in the application.
  • Workflows Public Endpoints — Workflows page public endpoints were added.
  • Onboarding Integration — The GitHub onboarding integration was extended.

Fixes

  • Model Validation Timing — Model validation now waits until the model id settles before running.
  • MCP Server Context — The evaluations creator now handles MCP server context correctly.
  • Metric Breakdown Card — The metric breakdown card issue was fixed.
  • Temporary Metric Viewer — The temporary metric collections viewer was removed.

Upgrade Notes

If you send Mantle project ids through the public API or Bedrock configuration, verify they are valid and present where attribution is expected before upgrading.

Built byConfident AI