Looking at the chat example where the LLM is helping to return an item, I’m trying to get my head around one would train something like this.
For example, with step 1, is the LLM to output an action with that sql?
for step 2, does the LLM output the type of api to call and the parameters?
for step 3, is the model already trained such that it knows what a valid email looks like?
