Struct LongTermMemory 

pub struct LongTermMemory { /* private fields */ }

A persistent, graph-based Long-Term Memory store.

LTM is responsible for storing consolidated memories and the knowledge graph extracted from them. It uses indices for efficient querying.

Implementations

impl LongTermMemory

pub fn new(config: LtmConfig) -> Self

Creates a new, empty LongTermMemory with the given configuration.

pub fn store(&mut self, entry: MemoryEntry) -> Result<MemoryId>

Stores a MemoryEntry in the LTM, typically after consolidation from STM.

Arguments

• entry - The MemoryEntry to store.

Returns

A Result containing the MemoryId of the stored entry.

pub fn get(&self, id: &MemoryId) -> Result<Option<MemoryEntry>>

Retrieves a MemoryEntry by its ID.

pub fn remove(&mut self, id: &MemoryId) -> Result<()>

Removes a MemoryEntry from the LTM.

pub fn query(&self, query: &MemoryQuery) -> Result<Vec<MemoryResult>>

Queries the LTM for memories matching the given MemoryQuery.

pub fn add_entity(&mut self, entity: Entity) -> Result<EntityId>

Adds a new Entity to the knowledge graph.

Returns

A Result containing the EntityId of the newly added entity.

pub fn get_entity(&self, id: &EntityId) -> Option<&Entity>

Retrieves an Entity by its ID from the knowledge graph.

pub fn add_link(&mut self, link: Link) -> Result<()>

Adds a new Link between two entities in the knowledge graph.

pub fn get_links_from(&self, id: &EntityId) -> Vec<&Link>

Retrieves all outgoing links from a given entity.

pub fn get_links_to(&self, id: &EntityId) -> Vec<&Link>

Retrieves all incoming links to a given entity.

pub fn find_entities_by_type(&self, entity_type: &str) -> Vec<&Entity>

Finds all entities of a specific type.

pub fn find_related( &self, id: &EntityId, relation: Option<&Relation>, ) -> Vec<&Entity>

Finds all entities related to a given entity (1-hop neighbors).

Arguments

• id - The EntityId of the starting entity.
• relation - An optional Relation to filter the links by.

pub fn traverse( &self, start: &EntityId, max_depth: usize, ) -> Vec<(&Entity, usize)>

Traverses the knowledge graph starting from a given entity (using BFS).

Arguments

• start - The EntityId to start the traversal from.
• max_depth - The maximum depth to traverse.

Returns

A vector of tuples, where each tuple contains a reference to an Entity and its depth from the starting entity.

pub fn semantic_search( &self, query_embedding: &Embedding, limit: usize, ) -> Vec<(&Entity, f32)>

Performs a semantic search over entities using embedding vectors.

Arguments

• query_embedding - The embedding vector to search with.
• limit - The maximum number of results to return.

Returns

A vector of tuples, where each tuple contains a reference to a matching Entity and its cosine similarity score. Returns an empty vector if embeddings are disabled.

pub fn memory_count(&self) -> usize

Returns the number of memory entries stored in the LTM.

pub fn entity_count(&self) -> usize

Returns the number of entities in the knowledge graph.

pub fn link_count(&self) -> usize

Returns the total number of links in the knowledge graph.

pub fn memory_usage(&self) -> usize

Returns the estimated memory usage of the LTM in bytes.

pub fn clear(&mut self) -> Result<()>

Clears all data from the LTM, including memories, entities, and links.