/ docs/ARCHITECTURE.md

Full pipeline DAG,
per-process, with side-inputs.

The home page shows the five-stage simplified view. This page shows the full technical DAG as Nextflow actually runs it — every process, every subworkflow, every side-input. The critical path is the Main column. The Atlas column feeds reference data from the Zenodo tarball. The Annotate column is an optional dimmed branch (--skip_annotate by default).

Full per-process Nextflow DAG for the APPLY workflow. Three columns: atlas reference inputs on the left, the main critical path in the centre (PREPROC subworkflow, SIMPER_MATCH, three correction layers IS/RIE/ref-filter, four parallel decompositions, three evaluation modules, REPORT), and a dimmed optional ANNOTATE branch on the right. — Fig. 2 · APPLY workflow, full per-process DAG, v0.1.0-dev. docs/ARCHITECTURE.md ↗

§ Reading the DAG

Each box is a Nextflow process; each process wraps one function in framework/*.py (plus its container digest and pinned atlas hash — see pipeline_info/provenance.yaml).

Solid black arrows are the critical path. Dashed soil-tone arrows are atlas side-inputs (reference data from the Zenodo tarball). The dimmed right column (ANNOTATE) runs only if you pass --skip_annotate false — decomposition does not depend on it. The ★ marks the primary decomposition method (default: fc_weighted_bc).

For full SOP-level traceability (every module ↔ function ↔ test ↔ SOP step) see the repository's ARCHITECTURE.md.

Full pipeline DAG, per-process, with side-inputs.

Full pipeline DAG,
per-process, with side-inputs.