v0.2.0

0.2.0 (2026-11-04)

Breaking changes

• entropy: Rank single-orbital entropies over the whole window by default (#112)

  Code that builds EntropyConfig without ranking now thresholds over
  the whole window.
  To keep the previous selection, pass ranking="subset":

  # before (implicit subset)
  EntropyConfig(fraction=0.1, fraction_type="r")
  # after, same selection as before
  EntropyConfig(fraction=0.1, fraction_type="r", ranking="subset")

• projection: Reject AO labels that match no projection AO (#113)

  A label that matches no AO used to give an empty selection and now
  raises ValueError.
  Fix the label using the shells listed in the message. For a 4d/5d metal,
  either give the
  molecule an ECP or project onto the molecular basis:

  # before: silently empty when Ru has no ECP
  ProjectionConfig(ao_labels=["Ru 4d"])
  # after: give Ru an ECP when building the molecule (as in the paper) ...
  mol = gto.M(atom=..., basis={"Ru": "cc-pvdz-pp", "default": "def2-svp"}, ecp={"Ru": "cc-pvdz-pp"})
  # ... or project onto the molecular basis
  ProjectionConfig(ao_labels=["Ru 4d"], basis=None)

• projection: Default select_by_ao_projection to a contribution of 0.5 (#114)

  Calls to select_by_ao_projection without contribution (and without
  top_n) now keep
  only orbitals with a weight of at least 0.5. To keep every orbital, as
  before, pass
  contribution=0.0:

  # before (implicit 0.0: every orbital kept)
  select_by_ao_projection(mf, ao_labels=["C 2pz"])
  # after, same result as before
  select_by_ao_projection(mf, ao_labels=["C 2pz"], contribution=0.0)

Fixes

• projection: Warn when per-label thresholds do not match the labels (#116)

Documentation

• parameters: How to choose the AEGISS parameters (#117)
• uhf: Define the UHF active space as two independent lists (#115)

v0.2.0rc1

0.2.0rc1 (2026-11-04)

Breaking changes

• entropy: Rank single-orbital entropies over the whole window by default (#112)

  Code that builds EntropyConfig without ranking now thresholds over
  the whole window.
  To keep the previous selection, pass ranking="subset":

  # before (implicit subset)
  EntropyConfig(fraction=0.1, fraction_type="r")
  # after, same selection as before
  EntropyConfig(fraction=0.1, fraction_type="r", ranking="subset")

• projection: Reject AO labels that match no projection AO (#113)

  A label that matches no AO used to give an empty selection and now
  raises ValueError.
  Fix the label using the shells listed in the message. For a 4d/5d metal,
  either give the
  molecule an ECP or project onto the molecular basis:

  # before: silently empty when Ru has no ECP
  ProjectionConfig(ao_labels=["Ru 4d"])
  # after: give Ru an ECP when building the molecule (as in the paper) ...
  mol = gto.M(atom=..., basis={"Ru": "cc-pvdz-pp", "default": "def2-svp"}, ecp={"Ru": "cc-pvdz-pp"})
  # ... or project onto the molecular basis
  ProjectionConfig(ao_labels=["Ru 4d"], basis=None)

• projection: Default select_by_ao_projection to a contribution of 0.5 (#114)

  Calls to select_by_ao_projection without contribution (and without
  top_n) now keep
  only orbitals with a weight of at least 0.5. To keep every orbital, as
  before, pass
  contribution=0.0:

  # before (implicit 0.0: every orbital kept)
  select_by_ao_projection(mf, ao_labels=["C 2pz"])
  # after, same result as before
  select_by_ao_projection(mf, ao_labels=["C 2pz"], contribution=0.0)

Fixes

• projection: Warn when per-label thresholds do not match the labels (#116)

Documentation

• parameters: How to choose the AEGISS parameters (#117)
• uhf: Define the UHF active space as two independent lists (#115)

v0.1.0

0.1.0 (2026-10-28)

Features

• Additional localization procedure (#46)
• Added format (#54)
• Added resolution space without SOE (#55)
• Added beta components to plot_ao_contributions (#58)
• Add mo_composition (#59)
• Add clusters grouping to mo_ao_contributions (#61)
• Simple chemcore core identification; (#57)
• Started UHF integration (#62)
• Orbital index mutable setup (#65)

Fixes

• Improved plateau visibility
• Expose AO contributions and update outdated tutorials (#71)
• Use mulliken as default pop_method for Pipek-Mezey (#72)
• IAO shape, Pipek-Mezey occupied pop_method and plot output names (#76)
• scf: Log the mean-field energy after the SCF (#86)

Documentation

• Refining mkdocs (#64)
• localization: Explain why the DMRG window is not a frontier window for localized orbitals (#75)
• Fix docstrings that no longer match the code (#79)
• plotting: Explain how to use the MI plots with an external mutual-information matrix (#81)
• Trim the README to an entry point and fix stale docs pages (#77)
• tutorials: Use a CAS(18,18) window in 03/04 and fix stale tutorial text (#78)
• Deploy the site, run the doc code blocks and add a developer guide (#88)
• uhf: Validate the entropies against PySCF FCI instead of pyblock2 (#93)
• site: Show the Apache licence on the docs site (#101)
• citing: Cite AEGISS and the work it builds on (#102)
• contributing: Document the Apache-2.0 licence header (#104)

v0.1.0rc1

0.1.0rc1 (2026-10-28)

Features

• Additional localization procedure (#46)
• Added format (#54)
• Added resolution space without SOE (#55)
• Added beta components to plot_ao_contributions (#58)
• Add mo_composition (#59)
• Add clusters grouping to mo_ao_contributions (#61)
• Simple chemcore core identification; (#57)
• Started UHF integration (#62)
• Orbital index mutable setup (#65)

Fixes

• Improved plateau visibility
• Expose AO contributions and update outdated tutorials (#71)
• Use mulliken as default pop_method for Pipek-Mezey (#72)
• IAO shape, Pipek-Mezey occupied pop_method and plot output names (#76)
• scf: Log the mean-field energy after the SCF (#86)

Documentation

• Refining mkdocs (#64)
• localization: Explain why the DMRG window is not a frontier window for localized orbitals (#75)
• Fix docstrings that no longer match the code (#79)
• plotting: Explain how to use the MI plots with an external mutual-information matrix (#81)
• Trim the README to an entry point and fix stale docs pages (#77)
• tutorials: Use a CAS(18,18) window in 03/04 and fix stale tutorial text (#78)
• Deploy the site, run the doc code blocks and add a developer guide (#88)
• uhf: Validate the entropies against PySCF FCI instead of pyblock2 (#93)
• site: Show the Apache licence on the docs site (#101)
• citing: Cite AEGISS and the work it builds on (#102)
• contributing: Document the Apache-2.0 licence header (#104)