holos-tda

Python bindings for holos, a Vietoris-Rips persistent homology engine. The package installs the holos-tda CLI, which prints ripser-compatible output.

SparseProgram reuses valid articulation-separated atoms and repairs invalid atoms. It reports exact class continuation and work, and writes portable program, trace, and intervention artifacts. The optional holos_tda.torch module computes strict finite H1 endpoint derivatives for distinct edge weights.

SparseIndex maintains immutable persistence versions over separator interfaces through the configured dimension. It composes relative filtered cores, applies atomic active-topology patches, and emits proof records for the separate checker.

kinetic_zigzag computes exact fixed-scale class dynamics for affine edge weights. It returns a HOLOSZZ artifact for the separate checker.

rips_points_classes, rips_condensed_classes, and rips_sparse_classes return source-bound H1 class records. Their _class circular functions check the active graph, interval, field, representative scale, and class identity. circular_coordinates_class accepts the same records with a square distance matrix. These bindings require an odd prime, such as 47.

intervene_cohomology finds one minimum-cost set of candidate edges across declared graph scenarios. It returns an exact plan or a checked bound. The separate checker repeats every fixed-scale cohomology and search step from the HOLOSCI artifact.

relative_coverage returns an exact fence-filling chain. synthesize_coverage and synthesize_affine_coverage find a minimum-cost sensor activation plan across communication states and bounded failures. They return a HOLOSCOV artifact for the separate checker. Physical coverage is conditional on the controlled-boundary domain, placement, fence, and communication assumptions.

pip install holos-tda
python -c "import holos_tda; print(holos_tda.rips_points([[0, 0], [1, 0], [0, 1]]))"

uvx holos-tda points.csv runs the CLI without installing.

Source and docs: https://github.com/t0rsion/holos