halo-model-workflow

name	halo-model-workflow
description	Guides edits to hmfast's differentiable JAX halo model (HaloModel, mass function, bias, concentration, profiles). Use when changing src/hmfast/halos/, power-spectrum methods, or halo-model consistency flags.

Halo model workflow

Before editing

1. Read src/hmfast/halos/halo_model.py for how Cosmology is wired (multiple _load_emulator calls at init) and how TophatVar / mcfit is instantiated.
2. Check MassDefinition, halo_mass_function, halo_bias, subhalo_mass_function, and concentration defaults — HaloModel.update() preserves the immutable-style pattern via pytree flatten/unflatten.

Implementation tips

• Keep new quantities JAX-friendly and consistent with x64 usage elsewhere.
• If adding parameters, thread them through update() and _tree_flatten / _tree_unflatten so the model remains a valid pytree where intended.
• Add or extend pytest cases in tests/test_halo_model.py (shapes, initialization, basic spectra).

Verification

Run from repository root:

pytest tests/test_halo_model.py -v

Extend tests if you add public methods on HaloModel.