Hi, I am trying to put together a "simple" assistant which picks up the phone, answers questions. If it has answers replies back, if it doesn't have the answers the caller is looking for then it collects the info and sends me an email. I have made the tool in Vapi (no code as I'm green in it) connected it to Make. created the webhook, but I have trouble connecting the gmail to it. To be precise i dont know how to set up that the gmail only sends me the relevant info. Phone #, Email, message. Also i know probably it is a prompting issue, but i was wondering if you have any tip how to get better results if it is asking for phone numbers and email adresses. it looks to me that it cant handle if the caller is spelling the email, or the number. Would anyone be so kind and could help me in it? Thank you!

Hi guys, Im building an AI receptionist that is meant to handle missed calls during after-hours and answer common customer questions. As many businesses (I assume) get the same common X questions all the time, I have tried to put together a KB. It turned out that my business has about 5 different "areas" that have its own separate Q&As. My thinking process was the following: Putting everything into one prompt is going to cause latency and makes it really hard to update. I have decided to create 5 different KB documents and prompted the AI to use each KB based on the questions the caller asks. So now the latency is good, no big issues there, but now the AI is hallucinating a lot or using answers from the appropriate KB but only some. Some answers are clearly in the KB yet it says doesnt have answer for it. I might have to mention that I sent the KB through chat GPT and asked it to convert to an easy-to-follow format for the voice assistant. Do you have any helpful tips for me on how to handle this?