Test strategy¶

Lab testing proves every transition and every refusal in the path from evidence need to follow-up. A successful planning or handoff scenario does not establish readiness, physical execution, QC acceptance, or evidence promotion.

Evidence layers¶

Layer	Contract under test	Representative suite
design	endpoints, contrasts, controls, replication, blocking, protocols, dependencies, and acceptance	`tests/design/`
planning	assay construction, priorities, queue, batching, scheduling, stable order, and deferred work	`tests/planning/`
readiness	resources, capacity, staffing, budget, risk, provenance, controls, conditional and blocked outcomes	`tests/readiness/`
handoff	authority, custody, instructions, artifact identity, target mapping, loss, rejection, and round trip	`tests/handoffs/`
observation	plan linkage, values, missingness, deviations, failure class, and immutable record	`tests/outcomes/`
QC and reliability	acceptance rules, controls, dispersion, reproducibility, partial and failed results	outcome and benchmark tests
reconciliation	requested/observed delta, rerun, redesign, hold, promotion, and history	`tests/reconciliation/`
package boundary	Foundation primitives, Core signatures, Knowledge grounding, Intelligence feedback, Runtime output	`tests/package/`

Transition proof¶

flowchart TD
    C["changed Lab contract"] --> P["valid transition"]
    C --> B["blocked or refused transition"]
    P --> S["serialization and identity"]
    B --> S
    S --> D["downstream artifact interpretation"]
    D --> H["history remains reconstructable"]

Run the focused lifecycle family first, then the complete package suite for public models, handoffs, persistence, or cross-package changes:

uv run --project packages/bijux-proteomics-lab \
  pytest -q packages/bijux-proteomics-lab/tests

Required operational pressure¶

Include missing controls, exhausted material, unavailable instrument, capacity conflict, incompatible batch, absent authority, target-system rejection, partial observation, deviation, failed QC, irreproducibility, biological non-support, inconclusive outcome, and promotion refusal.

Tests stop at the package boundary: they can validate a handoff and a returned observation contract, but they cannot prove physical instrument execution unless independent external evidence is supplied and linked.