querier

	// for GET /proposal-readiness [review COMPL-2]. RLS scopes all three tables to the caller's org.
	ProposalReadiness(ctx context.Context, opportunityID *uuid.UUID) ([]ProposalReadinessRow, error)
	// Candidates whose most recent recorded decision is older than the retention window and whose PII is
	// not yet anonymized. The de identified slate_members outcomes are retained; only the candidate PII is
	// wiped (by AnonymizeCandidate). RLS scopes this to the caller's org.
	RetentionDueCandidates(ctx context.Context, dollar_1 int32) ([]uuid.UUID, error)
	SamSyncMeta(ctx context.Context) (SamSyncMetaRow, error)
	// Hybrid retrieval: pgvector cosine kNN fused with full-text rank via Reciprocal Rank Fusion (k=60).
	// RLS already scopes both legs to the caller's org. $1=query embedding, $2=limit, $3=text query.
	// The vector arm is MATERIALIZED so the relaxed_order HNSW output is re-sorted by TRUE cosine distance
	// before it is ranked: under hnsw.iterative_scan = relaxed_order the index returns rows slightly out of
	// distance order, and ranking them without the re-sort would mis-rank the nearest neighbors at scale
	// (pgvector 0.8 README; AWS pgvector 0.8 Aurora guidance). Every ORDER BY carries a unique tiebreaker
	// (id) so results are byte-stable for fixed inputs, the AEDT reproducibility mandate. The caller SETs
	// LOCAL hnsw.ef_search and hnsw.iterative_scan first.
	SearchCandidatesHybrid(ctx context.Context, arg SearchCandidatesHybridParams) ([]SearchCandidatesHybridRow, error)
	// Cosine nearest-neighbor over the skills ontology (powers semantic skill adjacency).
	SearchSkillsSemantic(ctx context.Context, arg SearchSkillsSemanticParams) ([]SearchSkillsSemanticRow, error)

Hybrid retrieval: pgvector cosine kNN fused with full-text rank via Reciprocal Rank Fusion (k=60).