Skip to main content

Interface: CognitiveRetrievalResult

Defined in: packages/agentos/src/memory/core/types.ts:249

Properties​

diagnostics​

diagnostics: object

Defined in: packages/agentos/src/memory/core/types.ts:252

candidatesScanned​

candidatesScanned: number

confidence?​

optional confidence: RetrievalConfidenceSummary

escalations?​

optional escalations: string[]

factSupersession?​

optional factSupersession: object

Step-5: post-retrieve FactSupersession pass diagnostics. Populated only when a bench adapter or downstream consumer ran FactSupersession.resolve() over the retrieved traces.

factSupersession.droppedIds​

droppedIds: string[]

factSupersession.llmLatencyMs​

llmLatencyMs: number

factSupersession.notes?​

optional notes: string[]

factSupersession.parseOk​

parseOk: boolean

hyde?​

optional hyde: object

Step-4: when a HybridRetriever runs with a hydeRetriever attached, the first ~120 chars of the generated hypothesis are surfaced here for post-hoc analysis of which queries benefited from expansion.

hyde.hypothesis​

hypothesis: string

policyProfile?​

optional policyProfile: MemoryRetrievalProfile

retrievedTypedTraces?​

optional retrievedTypedTraces: ScoredMemoryTrace[]

Stage E: optional Hindsight typed-network output as canonical-shaped scored traces. When the manager is configured with typedNetwork and the variant supports retrieval-side activation ('full'), the manager delegates to a TypedNetworkRetriever which performs seed-finding (proper-noun + quoted-string entity extraction, case-insensitive intersection), spreading activation, and top-K ranking. Top-K results are surfaced as ScoredMemoryTrace[] for drop-in compatibility with the canonical retrieval pipeline (bank-prefixed content, namespaced IDs typed-network:<factId>, sourceType 'typed_network').

Absent when typed-network is not configured. Empty when the retriever found no seed matches in the typed-network store.

Phase 4.3 MVP: surfaced in diagnostics but NOT merged into the primary retrieved ranking. Phase 4.4 fusion lands when consumers wire the merged ranking. See 2026-04-26-hindsight-4network-observer-design.md.

scoringTimeMs​

scoringTimeMs: number

splitOnAmbiguous?​

optional splitOnAmbiguous: object

Step-6: when HybridRetriever runs with splitAmbiguousThreshold set, the bottom fraction of traces by first-pass rerank score are split at sentence boundaries and rescored. Replacements are recorded here for post-hoc analysis.

splitOnAmbiguous.candidateCount​

candidateCount: number

splitOnAmbiguous.replacedIds​

replacedIds: string[]

splitOnAmbiguous.threshold​

threshold: number

stageIds?​

optional stageIds: object

Per-stage ranked trace IDs for the hybrid retrieval pipeline (dense → sparse → merged → reranked → final). Populated by HybridRetriever so downstream consumers can compute per-stage retrieval-quality metrics (Recall@K, NDCG@K, MRR) and attribute losses to the stage that caused them. Absent for non-hybrid retrieval paths.

stageIds.dense​

dense: string[]

stageIds.final​

final: string[]

stageIds.merged​

merged: string[]

stageIds.reranked​

reranked: string[]

stageIds.sparse​

sparse: string[]

suppressed?​

optional suppressed: "weak_hits"

totalTimeMs​

totalTimeMs: number

vectorSearchTimeMs​

vectorSearchTimeMs: number

partiallyRetrieved​

partiallyRetrieved: PartiallyRetrievedTrace[]

Defined in: packages/agentos/src/memory/core/types.ts:251

retrieved​

retrieved: ScoredMemoryTrace[]

Defined in: packages/agentos/src/memory/core/types.ts:250