Skip to main content
Performs a semantic search against a knowledge base using the input as a query. Returns the most relevant chunks. Two settings to configure:
  • Top K – how many chunks to return
  • Format – text (plain concatenated results) or json (structured with metadata)
This is the workflow-level equivalent of the KB Search tool. The difference: a tool lets the model decide when to search, while this node searches at a fixed point in the pipeline. Use the node when you always want context, the tool when the model should judge.