HyperspaceDB Vector Store icon

HyperspaceDB Vector Store

The world's first Hyperbolic Vector Database integration for Spatial AI

HyperspaceDB Vector Store

AI-generated

Summary

Perform operations on a HyperspaceDB vector store including inserting documents, retrieving documents as a vector store for AI nodes or as a tool for AI agents, and loading ranked documents by similarity to a search prompt.

Inputs

  • Operation Mode (required) — Select the operation to perform: 'Get Many' to retrieve ranked documents, 'Insert Documents' to add documents, 'Retrieve Documents (As Vector Store for Chain/Tool)' to retrieve documents as a vector store for AI nodes, or 'Retrieve Documents (As Tool for AI Agent)' to retrieve documents for use as a tool with AI agents.
  • Collection Name (required) — The vector store collection to target for operations, chosen from the existing collections in HyperspaceDB.
  • Metric / Geometry (required) — Spatial metric used by the collection, auto-fetched but selectable based on the collection.
  • Dimension (required) — Number of vector dimensions for the collection, auto-fetched based on the collection.
  • Prompt (required) — Search prompt string used when retrieving ranked documents in 'Get Many' mode.
  • Limit — Number of top results to return when retrieving documents in 'Get Many' or 'Retrieve As Tool' modes, default is 4.
  • Include Metadata — Boolean to include document metadata in search results when retrieving documents.
  • Rerank Results — Boolean to rerank the retrieved results using a reranker algorithm.
  • Embedding Batch Size — Batch size for embedding documents when inserting into the vector store (only for 'Insert Documents' mode).
  • Name (required) — When retrieving documents as a tool for AI agents, the name to assign to the vector store tool.
  • Description (required) — When retrieving documents as a tool for AI agents, a concise description explaining what the tool does, helping LLMs produce expected results.

Output shape

The output varies depending on the operation mode: 'Get Many' returns a list of ranked document objects with optional metadata and scores; 'Insert Documents' produces no direct output but inserts documents into the vector store; 'Retrieve Documents (As Vector Store for Chain/Tool)' returns a vector store object suitable for chaining with AI nodes; 'Retrieve Documents (As Tool for AI Agent)' returns a tool interface wrapping the vector store for AI agent use.

The node requires HyperspaceDB API credentials for access. Available collections, their metrics, and dimensions are dynamically loaded from HyperspaceDB. Search and retrieval support optional reranking and metadata inclusion. Insertion batches documents using the configured embedding batch size. For 'Retrieve' modes, the node outputs specialized data types appropriate for AI workflows, connected via AI vector store or AI tool connection types.

Links

Discussion