Hi everyone! I'm trying to create an inbound assistant for appointment setting. In that context, I want to capture the leads email address to send a Cal.com confirmation. However, I find that the assistant gets the email wrong 95% of the time. Even when the user spells it out letter by letter. For example when my friend tried it, 'yannicziane@gmail.com' turned into 'janneckziane@gmail.com'. The assistant also didn't seem to understand it when he spelled it out. We also tried saying 'Yannic is written with a Y, double N, and C at the end'. I can see in the logs that the transcriber sometimes gets it right, but then the assistant response is something completely different. But also the transcriber get's it wrong a lot on its own. Why is that? Should I put modelOutputInMessagesEnabled = true, so the LLM can capture the email better than the transcriber alone? What are the best practices to reliably capture emails? Of course I could also just capture the phone number. But if I have an assistant they call via the Web Widget (SDK) then I don't have that info. I think this should be doable and would greatly appreciate some input on this. Thanks!

@tom We are currently having an office hour, can you please join it? We will help you directly. https://discord.gg/WPQuUJkF?event=1290804148354416710

sorry, i've just seen this now.

@tom you have to ask user to spell email character by character and then confirm/validate it back the same way and it will work.

Please provide an example of your system prompt for the email confirmation/validation process you recommend.

@tom / @KarlKing 1. Requesting the Email Address: - "May I have your email address, please? Could you spell it out for me, character by character?" 2. Collecting the Spelled Email Address: - Listen carefully as the customer spells out their email address, noting each character. 3. Confirming the Email Address: - "Just to confirm, that's [repeat the email address, spelling it out character by character], correct?" - Example: "Just to confirm, that's j - o - h - n - dot - d - o - e - at - e - x - a - m - p - l - e - dot - c - o - m, correct?" 4. If the Customer Confirms: - "Thank you." 5. If the Customer Indicates It's Incorrect: - "I'm sorry about that. Could you please spell your email address for me again?"

Hi @Shubham Bajaj I have this issue too and it's not able to understand my name, email or phone number when spelling, is there a solution or update?

@User this issue hasnt been SOLVED and I need some help, for my french version dialer

Hey @Jetsoccer Game and @Youri please create support tickets for your issues, and use the following template for the issue description. [template] - The call ID - When exactly this happened (the timestamp) - What response you expected to get - What response you actually got instead This would really help me figure out what went wrong!

man I have a simple api to use I built for my cold calling clients in solar please try it at: https://mailtor.hyperbolicai.io/ let me know and I will make it work for you