We're trying to break the complexity of our main info gathering assistant using a squad. The main assistant is tasked with gathering caller information but, if the caller asks something we would like to use a separate assistant in the squad to answer the question with some complex behaviour. I've implemented a squad in which the main assistant transfers the call successfully to the FAQ assistant, but we require that the caller doesn't feel like it's being transferred to another assistant. Removing the message in transferDestinations and in the first message seems to work, but the FAQ assistant keeps waiting for caller input although I've configured its firstMessageMode as 'assistant-speaks-first-with-model-generated-message' and instructed in the FAQ assistant system message to answer all the caller questions done by the caller through all the conversation. but in the first message of the caller it keeps doing the same greeting than the main assistant. The desired behaviour is_ 1- main assistant handles the call flow normally gathering caller info. 2- caller asks a question. 3- main assistant transfers to the FAQ assistant 4- FAQ assistant answers the caller question and keeps answering all caller questions until done. 5- When done FAQ assistant transfers back to main assistant. 6- main assistant continues with the call flow when it was left. From 1 to 3 we have everything right, but in 4 the caller must say something for the FAQ assistant to answer, which is not the desired behaviour. Are we forgetting something or is the desired behaviour possible with Squads?

@moidesko It sounds like you've made good progress with your squad implementation. To achieve the desired behavior where the FAQ assistant proactively speaks first without waiting for caller input, you should ensure that the

firstMessageMode

for the FAQ assistant is set correctly. However, you've mentioned that you've set it to 'assistant-speaks-first-with-model-generated-message', which is not one of the available options you provided in the context. Based on the context you've given, the available options for

firstMessageMode

are 'assistant-speaks-first' and 'assistant-waits-for-user'. To have the FAQ assistant speak first without waiting for the caller to say something, you should set the

firstMessageMode

to 'assistant-speaks-first'. This should prompt the FAQ assistant to initiate the conversation with the caller after the transfer. If you've already set the

firstMessageMode

to 'assistant-speaks-first' and the FAQ assistant is still not initiating the conversation, it's possible that there may be a configuration issue or a misunderstanding of how the setting is applied. In that case, please double-check the configuration of your FAQ assistant to ensure that it is set to 'assistant-speaks-first'. If the issue persists, it may be necessary to reach out to support for further assistance. Would you like me to check the configuration for you, or is there anything else I can assist you with regarding your squad setup?

If I set the firstMessageMode as 'assistant-speaks-first' the FAQ assistant reads the firstMessage, but we actually need the FAQ assistant to answer in the first message the question done by the caller to the main assistant prior to the transfer. Is that possible?

@moidesko To achieve the behavior where the FAQ assistant answers the caller's question immediately after the transfer without a generic greeting, you would need to ensure that the FAQ assistant has access to the context of the caller's question. This typically involves passing the caller's last question or intent to the FAQ assistant as part of the transfer process. Here's what you can do: 1. When the main assistant decides to transfer the call to the FAQ assistant, include the caller's last question or the detected intent in the transfer payload. 2. Configure the FAQ assistant to parse this incoming payload and generate a response based on the provided question or intent. 3. Set the

firstMessageMode

to 'assistant-speaks-first' for the FAQ assistant, but customize the first message to be the answer to the caller's question instead of a generic greeting. This way, when the transfer occurs, the FAQ assistant immediately addresses the caller's question with an appropriate response, creating a seamless experience for the caller. If your current setup does not support customizing the first message based on the transfer payload, you may need to explore additional configurations or reach out to the support team for guidance on how to implement this behavior. Would you like me to guide you through checking if your current setup can be adjusted to support this, or do you need further assistance with another aspect of your squad setup?

Could you give me an example configuration to achieve passing the question to the FAQ assistant, generating a response and puting it in the firstMessage with squads?

You need to use assistant overrides to set the firstMessageMode either to empty string or some other content.

I've achieved the desired behaviour changing the firstMessageMode to model generated, setting the firstMessage to empty and instructing in the main assistant to say a customized greet in the first message and continuing with the interrupted conversation when transferred back. Thank you very much!

As of now it's not there but will raise the feature request for it.

Ok, thank you very much!