I have used firm constraints (and played around with the prompting) in the personality agent but nothing seems to stop the bot from referring people to the website for more information. It sometimes also makes up its own 404 URLs and refers people there! Is there any way, prompt or otherwise to stop this kind of thing. Like a temp for the agent to alter how much it edits your initial text? Thanks

write this "never send website address or link " to Knowledge agent context filed what you want , and don't forger to publish after changes

Thanks, I got it to work in the end by repeating the constaints twice and cross referencing in the instructions (although most of the instruction character limit is now used on constraints. I think not hitting publish was an obstacle - I didnt know that you needed to, I will bear that in mind. Thanks!

no, I mean like in the picture. In that field you give context to your KB agent, like the chatgpt prompt

Ok wow, I didnt realise that! Thank you so much!