Skip to content

[Repo Assist] Release preparation: v0.15 #1696

Description

@github-actions

🤖 This is an automated analysis from Repo Assist. Maintainers should verify accuracy before acting.

The last release was v0.14 (November 2025). Since then, 124 commits have landed on main, including one breaking change, several new features, and a substantial number of bug fixes. A v0.15 release looks warranted.


⚠️ Breaking Changes

PR Description
#1688 Remove AutoGluon integration and BEST auto-assignment quality tier — users relying on AssignmentQuality.BEST or the AutoGluon-backed FCM must migrate to an explicit model assignment.

✨ New Features

PR Description
#1392 Add TabpfnEstimator for tabular causal inference
#1607 Add dedicated significance_level parameter and clearer significance output
#1399 Speed up refutation by parallelization
#1556 Add random_state to CausalEstimator for reproducible bootstrap CIs
#1557 Add random_state to DummyOutcomeRefuter
#1571 Add random_state to AddUnobservedCommonCause
#1573 Support CamelCase interpreter lookup in CausalEstimate.interpret
#1418 Fix identify all valid mediators in NIE/NDE estimation

🐛 Bug Fixes

PR Description
#1664 GCM: 27 silent bugs and crashes fixed (dtype issues, wrong class mappings, crashes)
#1632 Fix DummyOutcomeRefuter read-only array crash on pandas ≥ 3.0
#1609 Fix conditional effect estimation crash on pandas ≥ 3.0 (include_groups removed)
#1634 Fix distance_matching for int-encoded binary treatment
#1586 Fix ConfounderDistributionInterpreter crash on categorical confounders
#1585 Fix binary_flip crash on bool-dtype treatment/outcome
#1589 Add descriptive error for out-of-bounds interventions in weighting_sampler
#1563 Fix metric parameters not passed to NearestNeighbors
#1552 Raise ValueError instead of silent None when no valid estimand identified
#1546 Fix data-subset refuter index misalignment
#1579 Fix one_hot_encode crash on nested column structure failures
#1520 Fix TwoStageRegressionEstimator IV branch when instrumental variables missing
#1547 Fix conditional_MI entropy computation for discrete variables
#1611 Fix categorical_treatment_model crash on scikit-learn ≥ 1.7 (multi_class removed)

⚡ Performance

PR Description
#1676 Vectorise inner loops in marginal_expectation

📓 New Example Notebooks

PR Description
#1663 Causal discovery with FoundCause foundation model + GCM root cause analysis on simulated bottling line
#1524 Auditing CNN predictions for spurious correlations using chest X-ray data

📋 Release Checklist

  • Confirm version bump in pyproject.toml (currently 0.0.0 — looks like it's set dynamically)
  • Review and confirm breaking changes section (AutoGluon removal)
  • Draft GitHub Release notes (template above can be adapted)
  • Verify CI passes on main
  • Tag release and push to PyPI

Generated by 🌈 Repo Assist, see workflow run. Learn more.

Generated by 🌈 Repo Assist, see workflow run. Learn more.

To install this agentic workflow, run

gh aw add githubnext/agentics/workflows/repo-assist.md@11c9a2c442e519ff2b427bf58679f5a525353f76

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions