Knowledge Bases

Knowledge bases let you upload documents and data that agents can search and reference when answering questions. This is called RAG (Retrieval-Augmented Generation) — it grounds the agent's responses in your actual data instead of relying solely on the LLM's training.

---

Creating a Knowledge Base

Click the + Create button to set up a new knowledge base:

1. Give it a name and description
2. Upload your files (PDF, DOCX, TXT, CSV, MD, and more)
3. Files are automatically processed — chunked into sections and embedded for semantic search
4. Attach the knowledge base to an agent to start using it

How RAG Works

User asks a question ↓ Agent searches knowledge base for relevant chunks ↓ Matching content is included as context in the LLM prompt ↓ Agent generates a response grounded in your data

This means the agent can answer questions about your specific documents, policies, products, or any other uploaded content — with references to the source material.

Supported File Types

Managing Content

• Add files — upload new documents at any time
• Remove files — delete documents that are no longer needed
• Re-process — re-embed documents if processing settings change
• Monitor status — check the processing status of each file (pending, processing, ready, failed)

Quick Tips

• Smaller, focused knowledge bases perform better than large general ones
• Use clear document titles and headings — they improve chunk quality
• Check processing status — files need to finish embedding before they're searchable
• Attach a knowledge base to multiple agents if they need the same information
• Update documents regularly to keep the agent's context current
• For best results, remove duplicate or contradictory content