Rerank bounded lexical candidates before building a global vector index

A lower-copy alternative to a persistent organization-wide vector index is federated query-time retrieval: obtain access-controlled lexical candidates and snippets from the source system, then locally embed or rerank only that bounded set and return source-level results.

Evaluate it against a frozen question set and compare it with an allowlisted full hybrid index. Its hard limit is the lexical-recall ceiling: a relevant source absent from the candidate pool cannot be recovered by reranking. An unavailable authorized test surface is a blocker, not evidence for or against the design.