fascinating-indigo
fascinating-indigo•4mo ago

Fallback Message Repeats Unexpectedly During Warm Transfer

We're using warm-transfer-experimental mode with voicemailDetectionType: "transcript" for transferring calls to an AI agent. When the agent is unavailable or voicemail is detected, we fall back to a custom message, after which the assistant should continue the conversation with the caller. Here is the relevant configuration snippet:
destination: {
type: "number",
message,
number,
numberE164CheckEnabled: true,
callerId,
transferPlan: {
mode: "warm-transfer-experimental",
voicemailDetectionType: "transcript",
fallbackPlan: {
message: `It seems ${fullName || "team member"} is currently busy or on another call. Can I take a message?`,
endCallEnabled: false,
},
summaryPlan: { ... }
}
}
destination: {
type: "number",
message,
number,
numberE164CheckEnabled: true,
callerId,
transferPlan: {
mode: "warm-transfer-experimental",
voicemailDetectionType: "transcript",
fallbackPlan: {
message: `It seems ${fullName || "team member"} is currently busy or on another call. Can I take a message?`,
endCallEnabled: false,
},
summaryPlan: { ... }
}
}
Issue: In about 2 out of every 10 calls, when the fallback message is triggered: The fallback message plays successfully the first time. The assistant then continues the conversation with the caller as expected. But after a few seconds, the fallback message plays a second time, unexpectedly and interrupts the flow and the second playback gets cut off mid-sentence or the call hangs up abruptly. Expected Behavior: The fallback message should play once. After that, the assistant should remain in control and not repeat the fallback message or hang up unless explicitly instructed. @Shubham Bajaj @Kings_big💫
8 Replies
fascinating-indigo
fascinating-indigoOP•4mo ago
Also, when the person answers the call, the system provides a summary only after the person has spoken. It takes around 10 to 15 seconds before it starts giving the brief summary, and only then the caller is connected to the person. @Shubham Bajaj @Kings_big💫
sensitive-blue
sensitive-blue•4mo ago
Suppress repeat triggers using assistant hooks, ensure only one fallback executes, and consider adjusting the voicemail detection model or adding backend logic to track if the message already played..
fascinating-indigo
fascinating-indigoOP•4mo ago
I tried adjusting voicemail detection model, but had no improvement can you please help me out, as this is on production and its needed feature This is the main issue current I am facing @Kings_big💫 @Shubham Bajaj
sensitive-blue
sensitive-blue•4mo ago
To fix it, switch to warm-transfer-with-summary so the summary is pre-generated and plays immediately when the recipient answers..
fascinating-indigo
fascinating-indigoOP•4mo ago
I need to use warm-transfer-experimental so that the agent takes back the call if the specialist didn't answer, and if the specialist answers I need to provide a summary to the specialist Can you guys tell me how can I do it? @Kings_big💫 @Shubham Bajaj Our app is on production
sensitive-blue
sensitive-blue•4mo ago
Enable summaryPlan to speak a summary to the specialist, set fallbackPlan.endCallEnabled: false to let the assistant resume if unanswered, avoid holdAudioUrl to prevent call drops, and use Twilio callbacks to reliably handle unanswered calls and prevent duplicate fallback messages..
fascinating-indigo
fascinating-indigoOP•4mo ago
@Kings_big💫 It takes around 10 to 15 seconds before it starts giving the brief summary, and only then the caller is connected to the person. This long wait is not ideal for production app
sensitive-blue
sensitive-blue•4mo ago
Use warm-transfer-with-summary or pre-generate the summary before transfer to avoid the delay caused by real-time summary generation in warm-transfer-experimental..

Did you find this page helpful?