Creating collections

The vector database stores documents as embeddings for semantic search, powering AI agent knowledge bases and retrieval-augmented generation (RAG).

What is a collection?

A collection is a container for related documents. Each collection uses a specific embedding model to convert text into vectors for similarity search. You might create separate collections for different topics, projects, or data sources.

Navigate to Vector Database

In the dashboard sidebar, click Vector Database to view your existing collections and create new ones.

Creating a collection

Click Create Collection and provide the following details:

• Name — lowercase alphanumeric with hyphens or underscores, 3—50 characters
• Description — a brief description of the collection’s purpose
• Embedding model — the model used to generate vector embeddings

Embedding models

Choose an embedding model based on your content and language requirements.

Model	Dimensions	Best for
Amazon Titan Embed Text v1	1536	General English text
Amazon Titan Embed Text v2	1024	Improved general text
Cohere Embed English v3	1024	English-only, high accuracy
Cohere Embed Multilingual v3	1024	Multi-language support

Caution

Choose your embedding model carefully. You cannot change the embedding model after creating a collection. To use a different model, create a new collection and re-upload your documents.

Deleting collections

To delete a collection, open it from the Vector Database page and click Delete Collection. This removes all documents, embeddings, and metadata permanently. This action cannot be undone.

Next steps

• Uploading documents — add content to your collections
• Semantic search — query your collections using natural language