Skip to main content

Class: CognitiveMechanismsEngine

Defined in: packages/agentos/src/memory/mechanisms/CognitiveMechanismsEngine.ts:141

Orchestrates 8 cognitive mechanisms across the memory pipeline lifecycle.

When cognitiveMechanisms config is present on CognitiveMemoryConfig, an instance is created during initialization. Existing pipeline files call the lifecycle hooks at the appropriate points.

If HEXACO traits are provided, mechanism parameters are personality-modulated: emotionality → reconsolidation drift, conscientiousness → RIF strength, openness → involuntary recall probability + novelty boost, honesty → source skepticism, agreeableness → emotion regulation, extraversion → FOK surfacing.

Constructors​

Constructor​

new CognitiveMechanismsEngine(config, traits?): CognitiveMechanismsEngine

Defined in: packages/agentos/src/memory/mechanisms/CognitiveMechanismsEngine.ts:147

Parameters​

config​

CognitiveMechanismsConfig

traits?​

HexacoTraits

Returns​

CognitiveMechanismsEngine

Methods​

getConfig()​

getConfig(): ResolvedMechanismsConfig

Defined in: packages/agentos/src/memory/mechanisms/CognitiveMechanismsEngine.ts:270

Get resolved config for diagnostics.

Returns​

ResolvedMechanismsConfig

onAccess()​

onAccess(trace, currentMood): void

Defined in: packages/agentos/src/memory/mechanisms/CognitiveMechanismsEngine.ts:159

Called by MemoryStore.recordAccess(). Applies reconsolidation drift to the trace's emotional context.

Parameters​

trace​

MemoryTrace

currentMood​

PADState

Returns​

void

onConsolidation()​

onConsolidation(traces, llmFn?): Promise<{ gistedCount: number; regulatedCount: number; sourceDecayedCount: number; }>

Defined in: packages/agentos/src/memory/mechanisms/CognitiveMechanismsEngine.ts:234

Called by ConsolidationLoop.run() after step 5 (compact). Runs temporal gist, source confidence decay, and emotion regulation.

Parameters​

traces​

MemoryTrace[]

llmFn?​

(prompt) => Promise<string>

Returns​

Promise<{ gistedCount: number; regulatedCount: number; sourceDecayedCount: number; }>

onEncoding()​

onEncoding(trace, traceEmbedding): void

Defined in: packages/agentos/src/memory/mechanisms/CognitiveMechanismsEngine.ts:221

Called by EncodingModel.encode(). Classifies the trace as schema-congruent or schema-violating and adjusts encoding strength accordingly.

Parameters​

trace​

MemoryTrace

traceEmbedding​

number[]

Returns​

void

onPromptAssembly()​

onPromptAssembly(allTraces, retrievedIds): object

Defined in: packages/agentos/src/memory/mechanisms/CognitiveMechanismsEngine.ts:248

Called by MemoryPromptAssembler. May inject an involuntary recall memory into the assembled context.

Parameters​

allTraces​

MemoryTrace[]

retrievedIds​

Set<string>

Returns​

object

involuntaryMemory​

involuntaryMemory: MemoryTrace | null

onRetrieval()​

onRetrieval(results, allCandidates, retrievalCutoff, queryEntities): object

Defined in: packages/agentos/src/memory/mechanisms/CognitiveMechanismsEngine.ts:169

Called by MemoryStore.query() after scoring. Applies retrieval-induced forgetting to competitors and detects FOK signals.

Parameters​

results​

ScoredMemoryTrace[]

allCandidates​

CandidateTrace[]

retrievalCutoff​

number

queryEntities​

string[]

Returns​

object

Suppressed trace IDs and metacognitive signals.

metacognitiveSignals​

metacognitiveSignals: MetacognitiveSignal[]

suppressedIds​

suppressedIds: string[]

setClusterCentroids()​

setClusterCentroids(centroids): void

Defined in: packages/agentos/src/memory/mechanisms/CognitiveMechanismsEngine.ts:265

Update cluster centroids (called after consolidation derive step).

Parameters​

centroids​

Map<string, number[]>

Returns​

void