Skip to content

Invariants

Invariants are the claims that must remain true for the package to stay worth trusting.

Invariant Model

flowchart LR
    claims["claim and evidence meaning"]
    confidence["confidence and contradiction handling"]
    boundary["helpers do not redefine knowledge truth"]

    claims --> confidence --> boundary

This page should make knowledge invariants feel like truth-handling rules, not just implementation habits. Reviewers should be able to see what has to stay stable for downstream consumers to trust evidence and uncertainty together.

Review Rules

  • claim and evidence meaning remains canonical across consumers
  • confidence and contradiction handling stay explicit enough to review
  • repository and serialization helpers do not redefine knowledge truth

First Proof Check

  • packages/bijux-proteomics-knowledge/tests
  • src/bijux_proteomics_knowledge/claims.py and evidence.py
  • src/bijux_proteomics_knowledge/confidence/segments.py and review.py

Design Pressure

Knowledge invariants weaken when contradiction handling or helper behavior can quietly redefine what the package treats as truth. The package has to keep meaning, uncertainty, and review surfaces visibly linked.