Crate alaya 

Alaya

A neuroscience and Buddhist psychology-inspired memory engine for conversational AI agents.

Alaya (Sanskrit: alaya-vijnana, “storehouse consciousness”) provides three memory stores, a Hebbian graph overlay, hybrid retrieval with spreading activation, and adaptive lifecycle processes — all without coupling to any specific LLM or agent framework.

Quick Start

use alaya::{AlayaStore, NewEpisode, Role, EpisodeContext, Query};

let store = AlayaStore::open_in_memory().unwrap();

// Store an episode
store.store_episode(&NewEpisode {
    content: "Rust has zero-cost abstractions.".to_string(),
    role: Role::User,
    session_id: "session-1".to_string(),
    timestamp: 1700000000,
    context: EpisodeContext::default(),
    embedding: None,
}).unwrap();

// Query memories
let results = store.query(&Query::simple("Rust")).unwrap();
assert!(!results.is_empty());

Structs

AlayaStore

The main entry point. Owns a SQLite connection and exposes the full store / query / lifecycle API.

Category

CategoryId

ConsolidationReport

Episode

EpisodeContext

EpisodeId

ForgettingReport

Impression

ImpressionId

Interaction

Input to the perfuming process. The agent constructs this from whatever interaction format it uses (Signal message, Discord message, HTTP request, etc.)

KnowledgeFilter

Link

LinkId

MemoryStatus

NewEpisode

NewImpression

NewSemanticNode

NoOpProvider

A no-op provider for when no LLM is available. Consolidation and perfuming simply skip the LLM-dependent steps.

NodeId

NodeStrength

PerfumingReport

Preference

PreferenceId

PurgeReport

Query

QueryContext

ScoredMemory

SemanticNode

TransformationReport

Enums

AlayaError

LinkType

NodeRef

PurgeFilter

Role

SemanticType

Traits

ConsolidationProvider

The agent provides this trait to support intelligent consolidation. Alaya never calls an LLM directly — the agent owns the LLM connection.

Type Aliases

Result