seamm_bsse

An engine-agnostic library for N-fragment, charge-aware counterpoise (BSSE) corrections: given a cluster + fragment assignment + per-fragment charge, generate_job_specs produces the 2N + 1 job specs a full-cluster-basis correction needs (the full cluster; each fragment in the full cluster basis with the rest ghosted; each fragment alone in its own basis), and combine assembles the 2N + 1 results into a counterpoise-corrected total energy and gradient. validate_fragments checks a fragment list for overlapping atoms, unsupported open-shell fragments, and consistency with a known cluster charge/multiplicity.

No quantum chemistry, no engine-specific input/output handling -- that stays in each engine's own SEAMM step. orca_step's BSSE sub-step is the first consumer: it drives real ORCA jobs through this library's job specs and combine(), generalizing what used to be a two-fragment, neutral-singlet-only ORCA Compound-script correction to an arbitrary number of fragments with independent per-fragment charge. Validated against real ORCA on water/FEC/EC dimers (reproduces the old, trusted Compound-script numbers to SCF-noise level), charged two-body ion/water pairs (Na+/Cl-, Na+/H2O, Cl-/H2O, agreeing with approximate literature binding energies to a few kcal/mol), and a real N = 3 Na+/Cl-/H2O trimer. A Psi4 sub-step is planned next, as a thin wrapper around Psi4's own native bsse_type='cp' driver rather than a second consumer of this library's job-spec/combine() path -- Psi4 already does its own N-fragment CP orchestration internally, so wrapping it gives an independent cross-check of the ORCA path instead of the same arithmetic run twice.

See docs/developer_guide/campaigns/2026-08-03/bsse_architecture.rst for the full design: the physics, the confirmed ORCA ghost-indexing detail that makes combine() padding-only, what changed in orca_step, the validation results, and the milestone sequence.

Motivating plan: ~/Sites/mlff-training/2026-07-28_ion-bsse-plan/.