KCAE.Preface:5 - Alternatives and the cost of the whole arrangement

A small stable dossier can be read directly. A large known-name corpus may need an exact reader and lexical index. A heterogeneous corpus with many unknown-source questions can justify persistent semantic retrieval at the outset. A structural question can benefit from explicit references or a qualified graph. A rapidly changing collection with few requests may favour direct or deferred interpretation because repeated preprocessing cannot be amortized. These are alternative answers to the workload, not maturity levels every installation must pass through.

Costs include preparation, extraction repair, storage, rebuilding, request latency, total model input, residual main context, human interpretation, interruption and the work of applying advice. Compare them over a declared time horizon. A source answer for one analyst need not require that analyst to learn the engineering framework. A local engineer may receive the method needed to construct a reusable service. Those are different acquisitions of capability and have different costs.

Persistent retrieval is particularly valuable when the same expensive source inspection can serve many later questions. Deferred interpretation is particularly valuable when questions are rare, change rapidly or need distinctions that prepared views did not anticipate. Hybrid arrangements make both available. They incur more maintenance and require truthful routing when only some views are ready. KCAE.INDEX and KCAE.EVAL explain how to make that choice testable.

The design preserves source return because compression is useful and incomplete. It separates candidate ordering from adequacy because the best result in a weak pool can still be unusable. It separates event observation from user interruption because many inexpensive internal inquiries should end quietly. It preserves query-time connections because usefulness depends on the work, while retaining stable structural links for navigation. These are the shared architectural reasons for the method set.