pkgs/store/vecstore

pkgs/store/vecstore Defines vector similarity index and provides exact memory index and HNSW approximate nearest-neighbor implementation. Agent memory and recall use it to search for similar content by embedding.

Go API References

Core structure and implementation

Symbol	Function
Index	Define vector add, search, delete and index lifecycle.
Match	The expression matches ID, distance and metadata.
Memory / NewMemory	Provides in-process accurate vector search.
HNSW / NewHNSW	Provide HNSW approximate index.
HNSWConfig	Configure dimension, distance and graph parameters.
OpenHNSW	Open or create a persistent HNSW index from the Object Store.
LoadHNSW / LoadHNSWWithOptions	Restore HNSW from serialized stream.
CosineDistance	Calculate cosine distance.

Ownership Boundary

VecStore does not generate embeddings, nor determine model, chunk or recall policy. Embedding dimension, normalization, resource ID, result rearrangement, object name and save timing belong to the caller.