Class QdrantL2Store

Qdrant-backed cold semantic store (L2). Memories are points keyed by id with the embedding as the vector and the entry fields as payload; search filters by agentId for tenant/agent isolation and re-ranks the raw similarity with a recency blend (mirroring the Python actrone_memory QdrantStore).

Implements

L2Store

Index

Constructors

Methods

upsert search delete deleteAgentMemories

Constructors

constructor

new QdrantL2Store(client: QdrantLike, opts?: QdrantL2Options): QdrantL2Store
Parameters
- client: QdrantLike
- opts: QdrantL2Options = {}
Returns QdrantL2Store
- Defined in src/stores/qdrant.ts:80

Methods

upsert

upsert(entry: MemoryEntry): Promise<void>
Parameters
- entry: MemoryEntry
Returns Promise<void>
Implementation of L2Store.upsert
- Defined in src/stores/qdrant.ts:87

search

search(params: L2SearchParams): Promise<MemoryEntry[]>
Parameters
- params: L2SearchParams
Returns Promise<MemoryEntry[]>
Implementation of L2Store.search
- Defined in src/stores/qdrant.ts:115

delete

delete(memoryId: string): Promise<void>
Parameters
- memoryId: string
Returns Promise<void>
Implementation of L2Store.delete
- Defined in src/stores/qdrant.ts:183

deleteAgentMemories

deleteAgentMemories(agentId: string): Promise<void>
Delete every memory for an agent (local right-to-erasure).
Parameters
- agentId: string
Returns Promise<void>
Implementation of L2Store.deleteAgentMemories
- Defined in src/stores/qdrant.ts:191