Skip to content

Vector Stores

Create and manage vector stores for knowledge storage, semantic search, and RAG capabilities.

Overview

Vector stores are the backend databases that store document embeddings and enable semantic search across your knowledge repositories. When a user asks a question, the vector store is queried to find the most relevant document chunks based on vector similarity.

Vector stores are managed under Settings > Data Management > Vector Stores.

Creating a Vector Store

Currently, Cognitive Search (Azure Cognitive Search) is the supported vector store type.

  1. Navigate to Settings > Data Management > Vector Stores
  2. Click "Create Vector Store"
  3. Configure the Cognitive Search settings (see below)
  4. Click "Save"

Cognitive Search Configuration

Field Required Default Description
Vector Store Name Yes — A descriptive name for this vector store
Connect Credential Yes — Select a Cognitive Search credential configured under Settings > Credentials
Embeddings Yes — Select an embedding model configured for the workspace
Index Name Yes — The name of the Azure Cognitive Search index to use
Query Type Yes Vector Search The search query type to use
Semantic Configuration Yes default Name of your semantic configuration
Content Key Yes content Content storage key — the field that stores the document text
Vector Field Yes embedding Field used for vector search
Metadata Key Yes metadata Metadata storage key. Should be of ComplexType, containing subfields: document_id, source_id, source, url, author

Additional Parameters

Additional Fields

Define custom fields for your vector store index. Each field requires:

Column Description
Property The field name
Field Type The data type of the field

Click "Add" to create additional fields as needed.

Managing Vector Stores

Vector Stores List

The Vector Stores list page displays all configured stores with:

Column Description
Vector Store Name Name of the vector store
Type Store type (Cognitive Search)
Created On Date and time created
Actions Edit, Delete

Using Vector Stores

Once configured, vector stores are used by:

  • Knowledge Repositories — each repository is backed by a vector store for document indexing and search
  • Retrieval Tool — queries the vector store to find relevant context during conversations
  • Context Memory Tool — "Memory" type repositories use vector stores for persistent memory storage

Vector Store Permissions

Vector store management is controlled through RBAC:

  • Create, Read, Update, Delete

Configure these under Workspaces > Roles > Settings > Vector Stores. See RBAC & Roles for details.