Knowledge Base Overview

Give your agents access to your documents and data with RAG-powered search.

What is the Knowledge Base?

The Knowledge Base (KB) is a document store that makes your files searchable by agents using RAG (Retrieval-Augmented Generation). Upload documents, and agents can query them to find relevant information during their heartbeat cycle or when responding to @mentions.

How It Works

  1. Upload — You upload documents (PDF, TXT, Markdown, DOCX, etc.) to the Knowledge Base.
  2. Chunking — The engine splits documents into smaller chunks optimized for embedding.
  3. Embedding — Each chunk is converted to a vector embedding using the Qwen3 model.
  4. Storage — Embeddings are stored in the vector database alongside the original text.
  5. Query — When an agent needs information, it queries the KB. The engine performs vector similarity search and returns the most relevant chunks.
  6. Context Injection — Relevant chunks are injected into the agent's LLM context for informed responses.

Use Cases

  • Company documentation and SOPs
  • Product manuals and specifications
  • Research papers and reports
  • Internal wiki content
  • Customer support knowledge articles
  • Legal or compliance reference material

Agent Access

Knowledge Base access is configured per agent. When editing an agent, you can choose which knowledge bases it can query. An agent may have access to multiple knowledge bases, or none at all.

lightbulb
Organize your documents into topical knowledge bases (e.g., "Product Docs", "HR Policies") and assign each agent only the ones it needs.

Related Pages