Release notes from harness-evals
Articles about Release notes from harness-evals
Release v0.25.4 · harness/harness-evals
feat: report total and per-session cost and time in eval summaries (#
Harness-evals v0.25.1
Full Changelog: v0.25.0...v0.25.1
Release v0.25.0 · harness/harness-evals
feat: adapt Langfuse traces (#99)
V0.24.4: fix: skip oversized Langfuse traces during OTEL hydrate (#97)
a1a36b fix: bump pyjwt to >=2.14.0 for Trivy HIGH/CRITICAL CVEs 4d5604 fix: skip non-numeric turn latency metadata in TurnLatencyMetric 19a258 fix: coerce list-shaped assistant content in OTEL importer 93fccd fix: use Langfuse v1 observations on self-hosted 7da89a feat: report avg assistant-turn latency in golden eval summaries 541d1e fix: skip oversized Langfuse traces during OTEL hydrate
V0.24.2: AIEVALS-68: Price canonical model identities directly (#95)
328a30 fix: price canonical model identities directly
V0.24.1: feat: [AIPLAT-1634]: online QA conversation eval support (#94)
3f6849 fix: prefer child span usage cost over rolled-up root becc3f chore: keep session-merge cost docs vendor-agnostic e13f20 chore: strip product-specific wording from production-trace path 242c51 fix: [AIPLAT-1634]: sum per-trace costs on session merge d3d50c fix: bump anyio/httpx2/httpcore2 for Trivy HIGH/CRITICAL CVEs 349641 fix: keep core conversation/safety metrics unchanged 82addb chore: keep production-trace eval APIs vendor-agnostic e45df4 chore: [AIPLAT-1634]: drop unrelated bedrock/config wiring
Harness-evals v0.24.0
Full Changelog: v0.23.1...v0.24.0
Harness-evals v0.23.0
Full Changelog: v0.22.0...v0.23.0
Harness-evals v0.22.0
What's Changed Decision primitives (DP-1–DP-3): TypeSafe provider, Choice/Score/Noul metrics, composite batching by @sunilgattupalle in #30 Full Changelog: v0.20.2...v0.22.0
Harness-evals v0.20.1
Full Changelog: v0.20.0...v0.20.1