Package substance

Package-substance evidence tests whether each distribution has a coherent reason to exist. Canonical products must own domain behavior, Foundation must remain a narrow shared kernel, Agentic Proteins must remain a forwarding bridge, and maintainer tooling must own real repository-policy checks. Counts expose structural pressure; they do not measure scientific quality or release readiness by themselves.

flowchart LR
    inventory["source-module inventory"] --> classify["classify owned, forwarding, and thin modules"]
    classify --> role{"matches declared boundary role?"}
    role -->|no| block["package-substance blocker"]
    role -->|yes| evidence["retain inventory and policy verdict"]
    evidence -. "one input to" .-> release["release readiness"]

Read the metrics

Metric	Meaning	It does not prove
source modules	Python modules classified in the package source tree	that every module is public or scientifically important
owned logic	modules with substantive behavior assigned to this package	correctness, transfer, or sufficient test evidence
forwarders	modules that intentionally route to another canonical owner	parity for every caller or safe retirement
thin modules	small modules that may be valid seams or may indicate fragmented ownership	architectural debt without reviewing the named module
all diagnostic thresholds	owned-logic and thin-module counts are inside the inventory thresholds	whole-repository readiness; some diagnostic pressure is not release-blocking for every boundary role

Current boundary inventory

Package	Boundary role	Source modules	Owned logic	Forwarders	Thin modules	All diagnostic thresholds
agentic-proteins	compatibility_bridge	120	0	112	0	yes
bijux-proteomics-core	canonical_product	1337	1204	0	133	no
bijux-proteomics-dev	maintainer_support	188	160	0	28	no
bijux-proteomics-foundation	shared_kernel	39	37	0	2	yes
bijux-proteomics-intelligence	canonical_product	80	33	0	2	yes
bijux-proteomics-knowledge	canonical_product	76	54	0	22	yes
bijux-proteomics-lab	canonical_product	39	29	0	10	yes
bijux-proteomics-runtime	canonical_product	189	155	0	34	yes

The inventory contains 5 canonical products, 1 shared kernel, 1 compatibility bridge, and 1 maintainer support package.

Interpret the verdict

• A canonical product is blocked when its release identity outruns its owned domain behavior or hides unresolved ownership.
• Foundation is blocked when shared infrastructure grows into scientific, execution, evidence, decision, or presentation policy.
• Agentic Proteins may be forwarding-heavy because compatibility is its declared role; bridge-owned product logic is still a blocker.
• Maintainer tooling must contain tested policy behavior rather than only shell entrypoints or empty package structure.
• Thin-module counts trigger review pressure. A thin module is acceptable when it protects a durable seam and suspect when it fragments one owner.

The current inventory reports 1 package-substance policy finding. A clean substance verdict closes only this structural gate; scientific evidence, runtime replay, security, documentation, and release governance remain independent.

Reproduce the evidence

The machine-readable inventory is retained at docs/08-bijux-proteomics-maintain/bijux-proteomics-dev/package-substance.csv. The policy and freshness checks live in packages/bijux-proteomics-dev/tests/quality/architecture/test_package_substance.py.

.venv/bin/python -m bijux_proteomics_dev.quality.architecture.package_substance --check

A stale report, role mismatch, or unresolved policy finding is evidence to repair the owning package boundary. It must not be silenced by changing the generated summary or lowering the threshold without an architectural decision.