I am trying to replace the system generated transfer message AND have the assistant speak upon transfer. No matter what I try, I cannot get rid of "Transferring the call now" and the destination assistant remains SILENT. I've set "message" to "", and "firstMessageMode": "assistant-speaks-first", and "firstMessage": "Would you like to receive a free estimate?" { "squad": { "members": [ { "assistantId": "ce32564b-84e2-49b0-9ec4-de526b1a797e", "assistantDestinations": [ { "type": "assistant", "assistantName": "AcmeSquadEstimate", "message": "", "description": "If the customer expresses interest in obtaining an estimate, pricing, or quotation for their project." }, { "type": "assistant", "assistantName": "AcmeSquadHelp", "message": "", "description": "The customer expresses interest to modify an existing appointment, has inquiries..." } ] }, { "assistantId": "e498b863-cbc5-4691-9c10-84f365497aa2", "assistantOverrides": { "firstMessageMode": "assistant-speaks-first", "firstMessage": "Would you like to recieve a free estimate?" } }, { "assistantId": "e90cebea-aaba-4e6c-80d6-c7a5cac0e724", "assistantOverrides": { "firstMessageMode": "assistant-speaks-first", "firstMessage": "I will collect some information from you and get a staff member to help. Can I have your full name please?" } } ] }, "phoneNumberId": "7f6b8628-dc1c-4f3a-9a97-9a2c779a77f1" } https://cdn.discordapp.com/attachments/1278372669276028949/1278372669527425055/Technical_Diagrams_1.pdf?ex=66d090d3&is=66cf3f53&hm=c91361037f096ae2fb8efb5998f21ec96fcc866506e372ea8282ffddda26f56c&

@mindofman Maybe you can help here. I know you kind of have the same problem I think.

@ACME Mike @chowderr yeah I've been having the same issue. @Sahil told me to set "message" to "" also. I've been having another issue with calling the squad that has been keeping me from testing this, but I'm sad to hear if it isn't working because I really had my hopes up. A small bright side is that Sahil mentioned they will be adding this to the dashboard sooner than later, so maybe that will resolve it.

Thanks for letting me know it's not just me...I've been hopeful it's a configuration issue - hopefully we can get some input from @Sahil or @Shubham Bajaj .

@ACME Mike some info on message content. This is the message to say before transferring the call to the destination. If this is not provided and transfer tool messages is not provided, default is "Transferring the call now". If set to "", nothing is spoken. This is useful when you want to silently transfer. This is especially useful when transferring between assistants in a squad. In this scenario, you likely also want to set

assistant.firstMessageMode=assistant-speaks-first-with-model-generated-message

for the destination assistant.

With the json provided above, here is a call ID: 1bd1982f-e3e0-4db1-9036-78b83ed47f02

logs

🔵 21:03:54:722 Read from Canonical Knowledge (8e7dedcb9c0fabecfe9eece7e1b14c2e5bb621d5fbc575017b03720cf5c16303)... Content: { "hit": true, "completion": "I'll transfer you to our Estimate Assistant who can help you with that. One moment, please.", "documents": [], "metadata": {} } 🔵 21:03:54:722 Canonical Cache Hit 91 (cacheAborted: undefined) Cache Result: I'll transfer you to our Estimate Assistant who can help you with that. One moment, please.

After removing the KB files, the transfer works successfully as expected. However, the KB files are an important feature of the assistant. Could you please provide more information regarding how to set semanticCachingEnabled to false. This squad is created via webhook response following the information found in Swagger for SquadMemberDTO. https://cdn.discordapp.com/attachments/1278372669276028949/1278831027971948544/SquadMemberDTO.png?ex=66d2e474&is=66d192f4&hm=786b6b051ebd186e9bd56e31aca32d54b838583d9b73b17ed3a06e3622d30531&

I am late to this, but you are saying having the KB attached was causing the problem?

Yes - partly at least...I do not know how to implement @Shubham Bajaj 's suggestion of setting semanticCachingEnabled to false. I don't see an option for that in SquadMemberDTO. Right now, removing the KB files fixes the transfer issue - problem is, I need KB files 🤣 . Maybe the semanticCachingEnabled setting is the answer? But how to do it?

@ACME Mike use this and let me know how it goes.

json
{
    "squad": {
        "members": [], //your exisiting members config
        "membersOverrides": {
          "model": {
            "semanticCachingEnabled": false
          }
        }
    },
    "phoneNumberId": "<placeholder>"
}

Okay, I've set "semanticCachingEnabled": false and added back in my KB docs. I am back to my original issue of getting the generic "I'm transferring the call to..." messages. I'm providing the json I used below.

can you check your KB files once because it is being generated from the KB content.

I have (10) KB files, but I am not aware of any content that could cause the assistant trouble. I've tested the assistant with only (1) KB file at a time and found the assistant to make the transfer correctly with 4 of the files. However, if I take 2 of the files that work and attach them both, the transfers no longer transfers correctly.

If understood correctly your trying to say when you use more files the assistant isn’t working as expected if thats case then can you share the recent call ids.

I'm also finding the behavior inconsistent. Attaching a KB file that previously worked, now doesn't. Here is a call with one KB file (#9) that did not transfer correctly: b08fb411-1abb-4079-80a7-2ff79e81c384 Here is a call with a different KB file (#4) that did transfer correctly: 94cc8b08-9bb8-4a75-b058-a6de28a4a57a This call is with two KB files (#4, #5) that work individually, but not when both used together: 790f816d-01ec-4813-b1d1-5b476c380563

the reason is simple when context is missing from system prompt the KB is invoked.

Okay, thank you

I'm loving this back and forth dialog. Can I jump in here and go back to the "assistant-speaks-first-with-model-generated-message" part of the question. I can trasfer with no message no problem. But when the next assistant in line answers it does not speak first. At the moment I have no transfer message and no first message thinking that this firstmessagemode would prompt the AI to speak first... but it isn't. It still requires me to say something to start the second in line assistant.

I'm obviously struggling with this stuff myself, but when you set

firstMessageMode: assistant-speaks-first-with-model-generated-message

, you might try removing the

"firstMessage": ""

. It looks like you're telling it to speak first, then telling it to say nothing.

can you create a new support ticket, closing this ticket for now.