v1.4.0
What's Changed
- refactor(actionagent): spell the MCP constants with the MCP acronym by @TonsOfFun in #368
- fix: load RubyLLM when the gem registers its acronym by @aoki-ryusei in #372
- feat: pin RubyLLM's backend with a platform option by @TonsOfFun in #374
- Stop average_score raising on comparison evaluation runs by @TonsOfFun in #399
- Correct the average_score comments flagged in review of #399 by @TonsOfFun in #401
- Dashboard review follow-ups: engine correctness, credentials, MCP, telemetry bodies, DX by @TonsOfFun in #402
- Add scenario suite evaluations with multi-model comparison by @TonsOfFun in #407
- feat: agent evaluations — scenario suites, model comparison, fault diagnosis by @TonsOfFun in #406
- Add APM-style Metrics dashboard with time-series charts and golden signals by @TonsOfFun in #409
- Finish the suite results view, the APM Metrics page and the report by @TonsOfFun in #410
- fix(actionagent): fit the run report frame to its content by @TonsOfFun in #411
- fix(actionagent): derive encryption keys before Rails configures the encryptor by @TonsOfFun in #412
- Release activeagent 1.4.0 and actionagent 1.3.0 by @TonsOfFun in #413
Full Changelog: v1.3.1...v1.4.0