Skip to content

RAGOps v2.5.0

Choose a tag to compare

@thangldw thangldw released this 19 Nov 15:49
· 18 commits to main since this release
095935b

What's Changed

  • Add pluggable semantic_grounding evaluator by @thangldw in #63
  • Add blind relabeling workflow for the grounding benchmark by @thangldw in #64
  • Add sticky PR comment to the release gate action by @thangldw in #65
  • Add release_tools bump/check and derive test versions from pyproject by @thangldw in #66
  • CI: Python 3.14/3.15 and macOS, add mypy type check by @thangldw in #67
  • Raise coverage gate to 88%, add CLI/mutation tests, fix two loader crashes by @thangldw in #68
  • Move CLI helpers into cli_helpers.py by @thangldw in #69
  • Add property/tamper tests and fix loader crashes they found by @thangldw in #70
  • Add interactive HTML report explorer by @thangldw in #71
  • Add ragops init to build a first scenario from your own logs by @thangldw in #72
  • Import Langfuse, OpenAI and Anthropic logs in ragops init by @thangldw in #73
  • Add HaluEval public-benchmark case study (strict advantage did not reproduce) by @thangldw in #74
  • strict_grounding: number words and reversed names; document HaluEval error breakdown by @thangldw in #75
  • Add RAGTruth case study: strict grounding is near chance on long answers by @thangldw in #76
  • Remove all mypy ignore_errors overrides by @thangldw in #77
  • Raise coverage gate to 91% and fix loader crashes found by mutation tests by @thangldw in #78
  • Add optional NLI scorer example and measured results by @thangldw in #79
  • Release 2.4.0 by @thangldw in #80
  • NLI probe: evidence selection and aggregation variants (no gain) by @thangldw in #81
  • Measure explorer highlights against RAGTruth span labels by @thangldw in #82
  • ragops init --candidate: first comparison from two log sets by @thangldw in #83
  • Explain failed gates in plain language by @thangldw in #84
  • calibrate: suggest a pass threshold from human labels by @thangldw in #85
  • Add policy-suggest: generate a slice-aware release policy from a baseline by @thangldw in #86
  • Add weekly benchmark workflow by @thangldw in #87
  • Add mutation testing and tests that kill surviving mutants by @thangldw in #88
  • Release 2.5.0 by @thangldw in #89
  • Pin release gate semantics with mutation-tested policy_v2 tests by @thangldw in #90

Full Changelog: v2.3.1...v2.5.0