Vector Store icon

Vector Store

Store, search and manage vectors on the n8n host filesystem via js-vector-store

Actions12

Build ANN Index

AI-generated

Summary

Builds and persists an IVF approximate-nearest-neighbor index for the selected collection. Configure the cluster count, probes per search, and dimensions used for centroid distance; the operation returns index build details.

Inputs

  • Collection (required) — Collection to index.
  • Num Clusters — K-means cluster count; defaults to 100 and must be at least 1.
  • Num Probes — Clusters to probe per search; defaults to 10 and must be at least 1.
  • Sample Dims — Dimensions used for centroid distance during K-means; defaults to 0, which uses the dimension configured in the credential. Must be 0 or greater.
  • Read Only — Defaults to false. When enabled, rejects this operation because index building is treated as a write.

Output shape

One item per input item, with collection, numClusters, numVectors, and built fields.

On success, built is true. The IVF index is persisted for the collection. Use it with Search by enabling Use ANN Index; that search mode is cosine-only.

Links

Discussion