Does AI Agent with OpenAI Chat Model and 2 tools only look inside the tools for info?

[miguste]

I’m building a n8n flow in which an ai agent has to lookup information from my vector store, as well as my database, this is all working fine, but when I’m asking questions I wonder, does this flow (see screenshot below) only get (truthful) data from my sources? Or does it use the complete GPT library of data?

Thanks!

[solomon]

It uses the complete GPT library.
You can restrict that using two features:

Temperature of the model

This will reduce creativity and make the model respond in a more direct way.

Prompt

This is will tell the agent to be honest when it doesn’t find the answers.
You’re gonna have to test it a few times before you can be sure it’s not bringing his GPT knowledge.

.
If my reply answers your question, please remember to mark it as a solution.

[miguste]

Thank you! Would you recommend setting the temperatur to 0.0? And do you add both of your methods or pick just one?

[solomon]

I usually apply both, yes.

But finding the right temperature comes down to testing, @miguste

I got some good results using temperature at 0.2 for some projects, but each project is unique.