Release notes

Highlights for recent releases are summarised here. For the complete list of changes see the project changelog on GitHub.

v0.10.2 (reject integration)

  • Published ADR-029 and the companion implementation plan so the project has a definitive invocation/strategy/output story for reject orchestration while keeping reject=False behavior untouched.

  • Added the policy-driven RejectPolicy enum, RejectResult envelope, and RejectOrchestrator registry, wiring them through the prediction/explanation entry points so per-call and explainer-level defaults can opt into reject-aware outputs without breaking changes.

  • Documented the new usage patterns in docs/improvement/reject_policy_usage.md and linked the accompanying release-note summary for downstream consumers.

v0.9.0

  • Documentation CI now enforces notebook linting, docstring coverage (≥94%), and optional-extras placement checks so telemetry, PlotSpec, and plugin tooling stay clearly labelled as opt-in guardrails.

  • Release automation exercises the aggregated external-plugins extra to confirm the curated fast-explanation bundle remains an optional install with the expected dependencies.

  • Runtime performance controls—the calibrator cache, multiprocessing toggle, and vectorised perturbations baked into the core explainer—stay opt-in in v0.9.0. Follow :doc:../how-to/tune_runtime_performance to enable them deliberately, honour the new CE_CACHE/CE_PARALLEL environment overrides, and record rollback steps in your change log.

  • For the full milestone scope—including calibrated-explanations-first messaging and governance updates—see the v0.9.0 release plan <https://github.com/Moffran/calibrated_explanations/blob/main/docs/improvement/RELEASE_PLAN_v1.md#v090-documentation-realignment--targeted-runtime-polish>_.

v0.8.0

  • Adopted the role-based documentation information architecture defined in ADR-022.

  • Introduced PlotSpec as the default renderer with telemetry coverage for fallback chains.

  • Expanded telemetry payloads to include preprocessing snapshots and plugin provenance.

  • Published a practitioner-focused interpretation guide that explains calibrated predictions, interval semantics, alternative rules, and telemetry provenance; README quick-starts and notebooks now link to it as the primary learning path.

Refer to the CHANGELOG.md <https://github.com/Moffran/calibrated_explanations/blob/main/CHANGELOG.md>_ for historical entries and patch release details.