Skip to Content
Data VaultContent StoreOverview

Content Store overview

A Content Store is a searchable knowledge base for retrieval-augmented generation (RAG). You fill it with content — documents from your project or pages from a website — and the platform splits that content into chunks, embeds them as vectors, and makes them searchable by meaning. Agents then retrieve the most relevant chunks at run time to ground their answers.

Open the area from Data Vault → Content Store. Each store appears as a card with its status, size, and which agents use it.

Content Store list with store cards showing status, document counts, and agent usage

What powers a store

Every store is built from three choices you make when you create it:

  • Embedding model — the model that turns text into vectors. You choose one of your project’s embedding configurations (set up under Configuration → Embeddings). A store can’t be created until at least one embedding configuration exists.
  • Vector storage schemaShared (the default) or Dedicated (a Pro+ option), mirroring the table schemas.
  • Chunking — how content is split: strategy (Fixed Size, Sentence-Based, or Semantic), chunk size, and overlap.
⚠️

These settings are fixed when the store is created — there’s no in-place editor to change the embedding model or chunking afterward. Plan them up front; to change them, create a new store.

Filling and searching a store

Open a store to reach three tabs:

  • Documents — link folders or files from your Documents store; they sync in and stay up to date.
  • Web Crawler — crawl a website and index its pages.
  • Search — a playground to test what the store returns for a query.

How agents use a Content Store

  • Agents search a store with the RAG tools (rag_search and related).
  • A store also backs an AI Agent’s Episodic and Semantic memory — an agent points its memory tiers at a Content Store in its Memory tab.

Next