Vector Search

Vector Search lets you query the Vector Catalog with natural language or embeddings to retrieve relevant smart bites. Results support RAG (retrieval-augmented generation), agent grounding, and semantic search over contracts, policies, and other unstructured documents.

Overview

  • Run semantic queries (e.g. “What is the refund policy?”) and get ranked chunks.
  • Filter by metadata (source, date, document type).
  • Use results in your app or pass them to an LLM for grounded answers with source lineage.

Key tasks

TaskGuide
Understand Vector SearchOverview
How semantic retrieval worksSemantic retrieval
Scope queries with filtersFiltering
Improve relevance and recallBest practices

Tutorials and concepts

  • Vector Catalog — The store that Vector Search queries.
  • Agents — Ground answers using search results.