Knowledge Base

Documentation Index

Fetch the complete documentation index at: https://docs.brainstormer.io/llms.txt

Use this file to discover all available pages before exploring further.

A knowledge base (KB) is a collection of documents and content sources your agent can search when responding to users. Instead of relying solely on what the underlying model already knows, your agent retrieves relevant information from your KB and uses it to produce accurate, grounded answers. You can link any number of knowledge bases to an agent, and each KB can be shared across multiple agents.

​

How it works

When a user sends a message to an agent that has a KB linked, Brainstormer searches the knowledge base for content relevant to the question. The most relevant passages are pulled into the agent’s context before it generates a response — a technique called retrieval-augmented generation (RAG). The agent cites the sources it used, so users can see exactly where information came from. See How RAG works in chat for details on citations and the sources panel.

​

What you can store

​

Multimodal content

Brainstormer uses multimodal embeddings, so content doesn’t have to be text-only to be searchable. Images, video, and audio files are embedded alongside text, meaning your agent can find and reference them in responses. When you add media content, you can choose how it is processed:

• Native — the media is embedded directly, enabling visual and audio search.
• Transcription — audio and video are converted to text first, which works reliably for any length of content.

You can also enable visual entity extraction on individual sources, which uses AI vision to identify people, brands, and products in images and video.

​

Analytics

Each knowledge base has an Analytics dashboard that shows you how the KB is being used: Use the knowledge gaps report to find topics where your KB is missing coverage, then add sources to fill those gaps.

​

Next steps