Hi, I have agent_response = agent.chat(question). Does anyone know why there are differences between agent_response.response and agent_response.sources[0].content? Is agent_response.sources[0].content the agent's output after using the tools given whereas agent_response.response is the agent's output without restricting its reasoning to the tools?
Exactly! The source is the raw tool output (i.e. from a query engine possibly, or any other tool). The agent then takes that and basically re-interprets that response to answer the user, based on chat history
