Skip to main content
Logic settings let you control how your agent searches, ranks, and generates responses. These settings are configured per agent.

​
Retrieval

Control how documents are searched:
  • Top K — Number of chunks to retrieve (default: 10)
  • Similarity Threshold — Minimum relevance score (0-1)

​
Reranking

After initial retrieval, reranking reorders results by relevance:
  • Enable Reranking — Toggle reranking on/off
  • Rerank Model — Choose the reranking model
  • Top N — Number of chunks to keep after reranking

​
Generation

Control the response generation:
  • Model — Select the LLM for response generation
  • Temperature — Creativity vs. precision (0-1)
  • Max Tokens — Maximum response length
Lower temperature values produce more factual, consistent responses. Higher values allow more creative answers.