Skip to content

Scope and Non-Goals

The scope of bijux-canon-index is to make search behavior explicit enough to defend. It is not a general home for “things that happen around retrieval.”

Scope Map

flowchart LR
    prepared["prepared input"]
    scope["index scope"]
    results["replayable retrieval behavior"]
    refuse["source cleanup, claim meaning, and run authority stay out of scope"]

    prepared --> scope --> results
    scope --> refuse

This page should make index feel narrow in a useful way. It owns search behavior end to end, but it should stop before explaining what results mean or whether a whole run counts.

In Scope

  • vector execution and backend coordination tied to prepared ingest output
  • replayable retrieval behavior and provenance-rich search results
  • index-facing contracts that downstream reasoning and runtime flows rely on

Non-Goals

  • normalizing source material before search begins
  • deciding what retrieved evidence means for a claim or verification step
  • deciding whether a whole run is acceptable or durable under runtime policy

Scope Check

If the change can only be explained by saying “search needs it somewhere,” the ownership argument is not strong enough yet.

Design Pressure

If index starts collecting adjacent concerns just because they touch search, the package turns into a traffic junction instead of a clear retrieval surface. The non-goals keep the search contract reviewable.