Skip to main content

Overview

Each app can have one or more named vector stores — collections of embeddings you manage directly, useful when you want full control over chunking and retrieval instead of using the higher-level RAG API.
1

Create an index

POST .../create-index with your embedding dimension and distance metric.
2

Upsert vectors

POST .../upsert with { id, vector, payload } points.
3

Query

POST .../query with a raw vector (or query text for Bedrock Knowledge Base–backed apps).

Base URL

vectorName is a namespace you choose per app — e.g. "docs", "products" — allowing one app to maintain multiple independent vector collections.

Authentication

Widget tokens are also accepted as a fallback for browser-embedded use cases.

Resources

Indexes

Create, list, inspect, and delete vector collections.

Vectors

Upsert and query vector points.