v0.2.1

Release v0.2.1

Fixed

• Removed unused matplotlib import: Fixed ModuleNotFoundError in CI tests caused by unused import from matplotlib.axes import Subplot in tempest/steps/resample.py

Changes from v0.2.0

• All v0.2.0 breaking API changes (n_effective -> n_particles, removed n_active/metric)
• Documentation updates

v0.2.0

Release v0.2.0

Breaking Changes

• Renamed API parameters: → , removed and parameters
• Updated volume variation semantics: Removed sigmoid transformation, now returns raw CV

Documentation Updates

• Updated all documentation to reflect new API
• Updated all example scripts to use new parameter names
• Updated README and user guides

Migration Guide

Release 0.1.5

Full Changelog: v0.1.4...v0.1.5

Version 0.1.4

Bug fix release: Fixed critical packaging bug that caused ModuleNotFoundError for tempest.steps submodule after pip install.

Version 0.1.3

Changed default n_steps from 5 to 1

Release v0.1.2

Release v0.1.2

Fixed Critical Packaging Bug

Issue: Users getting ModuleNotFoundError: No module named 'tempest.steps' after pip install tempest-sampler

Root Cause: The package configuration in pyproject.toml was only packaging the top-level tempest directory, excluding subpackages like tempest.steps/, tempest.students/, etc.

Solution: Updated packaging to use automatic package discovery with [tool.setuptools.packages.find] to include all subpackages.

Changes

Fixed

• Critical packaging bug: Fixed ModuleNotFoundError: No module named 'tempest.steps' error that occurred after pip install
  • Root cause: pyproject.toml was only packaging the top-level tempest directory, excluding subpackages
  • Solution: Updated packaging configuration to use automatic package discovery

Added

• Integration test tests/test_package_install.py to verify all submodules are importable
• New CI workflow .github/workflows/test-wheel.yml to test pip-installed package in isolation
  • Builds wheel and tests imports in clean environment
  • Tests across Python versions 3.8, 3.10, 3.12
  • Catches packaging issues that unit tests miss

Changed

• Bumped version to 0.1.2

Verification

Tested with clean installations on all supported Python versions:

• ✅ from tempest.sampler import Sampler works
• ✅ from tempest.steps import Reweighter, Trainer, Resampler, Mutator works
• ✅ Sampler instantiation without errors

Installation

pip install tempest-sampler==0.1.2

All subpackages now properly included in distribution.

Release v0.1.1

[0.1.1] - 2026-01-23\n\n### Added\n- Bayesian model comparison example (linear vs oscillatory models) with Bayes factor calculation and interpretation guide\n- Standalone executable example script for interactive use\n- Adaptive MCMC steps with per-dimension scaling\n\n### Changed (Breaking)\n- n_active parameter now defaults to None instead of 256 (automatically computed as n_effective // 2)\n- Users no longer need to specify n_active in most cases\n- n_active=0 is no longer valid and will raise an error\n\n### Fixed\n- Documentation inaccuracies and wording fixes in parameter selection guide

v0.1.0

Initial release of tempest-sampler package