Where can I find the content an assistant has gene...
# supporth
hxrdxk
05/22/2025, 8:05 AMFor example, if an assistant is interrupted while speaking, where can I find what it would have said if it hadn't been interrupted?
hxrdxk
05/22/2025, 8:05 AM@User please answer
v
Vapi Support Bot
05/22/2025, 8:06 AMYou can listen for the
message Copy code
js
vapi.on("message", (message) => {
console.log("Assistant generated message:", message);
});v
Vapi
05/22/2025, 2:13 PMYou can find the LLM output in the in call logs.
h
hxrdxk
05/23/2025, 8:44 AMi want it before it speaks it during the call
hxrdxk
05/23/2025, 8:45 AMtranscript includes what it has spoken, i want what it will speak beforehand that is response generated by llm which it speaks
v
Vapi Support Bot
05/23/2025, 8:46 AMTo capture what the assistant is about to say before it speaks during a call, listen for the
message Copy code
js
vapi.on("message", (message) => {
console.log("Assistant generated message:", message);
});h
hxrdxk
05/23/2025, 8:46 AMhave a look
v
Vapi Support Bot
05/23/2025, 8:47 AMYou can listen for the
messageVapi Support Bot
05/23/2025, 8:47 AMTo access what the assistant is about to say before it speaks, listen for the
message Copy code
js
vapi.on("message", (message) => {
console.log("Assistant generated message:", message);
});v
Vapi
05/24/2025, 2:12 AMHey, you can access modal output through the modal output server event which is being sent to you as soon as modal generates the output and right before it is sent to the voice service provider. To select this event, go to the VAPI dashboard, then assistant of your choice, select advanced and scroll down to the server messages section, select modal.output server message, fill in your server URL if you haven't, and then you will get it on your server. Let me know if you require further help.
h
hxrdxk
05/26/2025, 11:12 AMhow can we configure model output to include structured output? as in response that which is already coming up + fields whataver we want
v
Vapi Support Bot
05/26/2025, 11:14 AMTo include structured data fields in your assistant's output, customize the
analysisPlanstructuredDataPromptstructuredDataSchema Copy code
json
{
"analysisPlan": {
"structuredDataPrompt": "Extract the user's name and intent from the call.",
"structuredDataSchema": {
"type": "object",
"properties": {
"userName": { "type": "string" },
"intent": { "type": "string" }
},
"required": ["userName", "intent"]
}
}
}call.analysis.structuredDatah
hxrdxk
05/26/2025, 11:20 AMnah, i am looking for this to be defined for model output in which one property vapi must fix which it has to speak and other one mmust be defined by developers according to our need.
v
Vapi Support Bot
05/26/2025, 11:21 AMCurrently, Vapi lets you capture the assistant's generated message before it is spoken by listening to the
messageh
hxrdxk
05/26/2025, 11:23 AMmain goal is to access additional structured data alongside the model output in real time during the call,
for example, if i want it to know that the answer talks about india then:
"response": "india is a great country",
"india":"true"
here response will be spoken by voice.
v
Vapi Support Bot
05/26/2025, 11:25 AMCurrently, Vapi does not support enforcing a real-time output format with both a fixed speech property and additional developer-defined fields during the call. You can capture the assistant's output as it is generated using the
messagev
Vapi
05/28/2025, 11:08 AMHey, you cannot get the structural data as it's doesn't exists. This is something that is not configurable through turn-taking models or models in general. So you have to make it work on your side.