Knowledge base
A knowledge base lets an agent answer from documents you upload - a price list, a policy, a product manual - rather than from what the model happens to know.
Uploading documents
Upload a document (PDF or text) under Knowledge base. It is split into chunks and indexed in the background, so a large file does not hold up the upload. A document becomes usable once it is marked ready.
Attaching it to an agent
A knowledge base is not used until an agent is pointed at it. On the agent, choose
which documents it may search and, if you want, how many passages to retrieve
(knowledge_base.top_k).
Different agents can be given different documents, so a support agent and a sales agent do not have to share one pile.
How retrieval works
When a question arrives, the passages most relevant to it are found and given to the model with the question. The agent answers from those passages.
Retrieval is hybrid: it matches on meaning and on exact keywords. Meaning alone is bad at exact tokens - an order number, a product code, a surname - and keyword matching is bad at paraphrase, so both run and the results are fused.
Say in the prompt what the documents cover, e.g. "Answer questions about pricing and delivery from the knowledge base. Never invent a price." This is the difference between an agent that looks things up and one that guesses.
What it does not do
A knowledge base answers questions. It does not take actions or read live data - a stock level or an order status has to come from a tool.