Expand AI Context beyound local documentation

Hi! Welcome to the forums!

:thinking:

do you have any context instructions? You can easily tell the model how it should handle user queries with a system message or an injected user message. you’re familiar with the playground, I imagine.

I’m not sure how well bifurcations or escalations work in langchain, but it would be fairly straight forward to catch a negative response (just ask for your output to be structured) and then decide what to do afterwards.

I’m not a great fan of langchain because it feels clunky and poorly documented. but if you find it useful, don’t feel discouraged!