VectorChat Docs

Knowledge Bases

Ground your agents in accurate information by creating knowledge bases from documents, text, and websites.

Knowledge bases let you provide agents with specific information to reference when answering questions. Instead of relying solely on the AI model's training data, agents search your uploaded documents and return accurate, source-backed answers.

Knowledge bases list Screenshot: The knowledge bases overview page showing all knowledge bases with their names, document counts, and status.

How knowledge bases work

When you upload a document to a knowledge base, VectorChat processes it through several steps:

  1. Document processing — The file is converted to text and split into smaller sections (chunks).
  2. Embedding — Each chunk is converted into a numerical representation that captures its meaning.
  3. Storage — The embeddings are stored in a vector database for fast similarity search.

When an agent receives a question, it searches the connected knowledge bases for the most relevant chunks and includes them as context in its response.

Types of content

Files

Upload documents in common formats. VectorChat extracts the text content and indexes it automatically.

Text

Paste plain text directly into a knowledge base. Useful for quick notes, FAQs, or content that doesn't exist as a file.

Websites

Provide a URL and VectorChat will crawl the website, extract content from each page, and index it. Useful for documentation sites and help centers.

Shared knowledge bases

Knowledge bases are shared resources. Create one and attach it to multiple agents:

  • A single source of truth — update the knowledge base and all connected agents benefit.
  • No need to upload the same documents to each agent separately.
  • Can be scoped to your personal workspace or shared across an organization.

Next steps