Just to add to Corentin P message : | ElevenLabs | Page 1

radiant cedar Apr 3, 2025, 2:52 PM

#

Hi Nicolas, the issue seems to be with your conversation setup

#

MissingDynamicVariableError: Missing required dynamic variables in tools: {'last_name', 'first_name', 'civility'}

#

You must have some misconfiguration issue in your webhook. In the config I can see that you have enabled conversation_initiation_client_Data_from_webhook, which is a good start

#

however, it looks like you have not provided that webhook. Dynamic variables need to be injected as part of conversation initiation, which I believe you are not currently doing, hence the calls cannot proceed

#

Do you have a custom server for Twilio?

late fog Apr 3, 2025, 2:59 PM

#

dynamic_variables are provided from https://ai-api.versity.online/api/call/init
We are only testing our agent using our Twilio number to avoid having this type of error

radiant cedar Apr 3, 2025, 3:00 PM

#

What do you mean? How is testing with your number different?

late fog Apr 3, 2025, 3:00 PM

#

Because when calling from Twilio it will gather dynamic variable using our API?

#

So it doesn't throw the missing dynamic variables error

radiant cedar Apr 3, 2025, 3:02 PM

#

@late fog do you want to jump on a call to screenshare ?

#

How are you providing the variables? I think there's something missing in that setup potentially

late fog Apr 3, 2025, 3:05 PM

#

Holy s*** We removed the dynamic variables from the bookAppointment tool and now the call go through

#

But why the usage of dynamic variable in a tool that should be used during the call is blocking the starting of the call ?

#

So we should only use LLM prompt instead in our tool arguments ?

radiant cedar Apr 3, 2025, 3:07 PM

#

It depends. If you want to inject client context to the conversation, then dynamic variables are perfect. BUT you need to do this always, and if you don't provide the context then the call will fail.

#

If you would like LLM to infer the value from the conversation before calling the tool, then LLM prompt will be more appropriate

#

DVs are almost universally better if you do have the information about the caller

late fog Apr 3, 2025, 3:08 PM

#

Ok so we should have an endpoint for the initial data that provide all dynamic_variables even if they are empty ? And is the endpoint called on all agent conversation or only on twilio ones ?

#

And did you have a chance to look on this one ? #📟│agents-chat message
(About the interruptions during the call that makes the agent hang up

#

)

radiant cedar Apr 3, 2025, 3:09 PM

#

endpoint will only be called for Twilio

late fog Apr 3, 2025, 3:10 PM

#

Ok so if you we want to use it on the web we should provide the dynamic variables using the 11labs library as i recall ?

radiant cedar Apr 3, 2025, 3:10 PM

#

you can configure it in the app

#

#

you need to add this webhook, so that we can call your endpoint when the conversation starts and get the data

late fog Apr 3, 2025, 3:12 PM

#

Yup I already did on our agent XCE4w736pIgSaSkAhmlT

But I didn't know it would make the other one crash if we don't have set dynamic variables.
It would be awesome if we could know on the call history what caused the call to crash

#

Thanks for your help anyway!

#

If you have time to look at this one ? #📟│agents-chat message
I think you wanted to transfer it to the person in charge but I don't know he had time to review it

#

(#📟│agents-chat message)

radiant cedar Apr 3, 2025, 3:16 PM

#

@late fog could you please check again with the missing dynamic variables case?

#

this should already be merged !

#

in conversation history you should be able to see websocket close reason

#

perhaps this is not working correctly for Twilio? 🤔

late fog Apr 3, 2025, 3:24 PM

#

radiant cedar perhaps this is not working correctly for Twilio? 🤔

Amazing!!

radiant cedar Apr 3, 2025, 3:24 PM

#

but... is it like that on your side as well?

#

I need to know

#

if there is a bug or not 😅

late fog Apr 3, 2025, 3:25 PM

#

Yup, no bugs, it's working perfectly 😄

radiant cedar Apr 3, 2025, 3:25 PM

#

awesome

#

then you should be better equipped to resolve these errors in the future 😉

late fog Apr 3, 2025, 3:27 PM

#

yes indeed!

#

But we just tried to interupt the agent and it hang up without any error 😅

#

https://elevenlabs.io/app/conversational-ai/history/TDCugJUmN1mjb4vnS1B4

ElevenLabs

AI Voice Generator & Text to Speech

Rated the best text to speech (TTS) software online. Create premium AI voices for free and generate text to speech voiceovers in minutes with our character AI voice generator. Use free text to speech AI to convert text to mp3 in 29 languages with 100+ voices.

radiant cedar Apr 3, 2025, 3:27 PM

#

sorry, what went wrong? 😄

#

you tried to interrupt and the conversation ended?

late fog Apr 3, 2025, 3:28 PM

#

When the agent is talking, the person interrupt him to ask something else and the call hang up

radiant cedar Apr 3, 2025, 3:28 PM

#

Ok, this is related to something that Sedat was trying to fix, I'll ping him

late fog Apr 3, 2025, 3:28 PM

#

Ok great

radiant cedar Apr 3, 2025, 3:29 PM

#

btw you have the error msg again in the conv history

#

right?

late fog Apr 3, 2025, 3:29 PM

#

Mate, you're amazing

#

We indeed have the error displayed now hehe

radiant cedar Apr 3, 2025, 3:30 PM

#

but unlikely you can do anything with it

#

I'll look into what's causing this error

late fog Apr 3, 2025, 3:31 PM

#

Yep but it's already good to know what caused the issue to come and bother you again 😂

#

Thanks

late fog Apr 3, 2025, 4:00 PM

#

Oh and if you have time to check why our agent XCE4w736pIgSaSkAhmlT starts the call by speaking really slowly and gradually speed up until it comes to our speed param

radiant cedar Apr 3, 2025, 4:06 PM

#

that sounds like an interesting bug 😅

#

I'll try to prioritize the interrupted message mismatch though

late fog Apr 3, 2025, 4:12 PM

#

It is right haha

#

Yep of course

radiant cedar Apr 4, 2025, 11:39 AM

#

oki, I have a fix that's approved, should be in production later today

#

hopefully you'll never see this "interrupted message" error again

late fog Apr 4, 2025, 12:43 PM

#

Thanks @radiant cedar .
Could you tell me when it's live ?
Because we are having interruptions (conv id : uZRKnPb69ahrGTEpkGAD / qtKrEYneOaYDnTVlYv0A / xDDEWR7TdRVwLFT5q6XG

radiant cedar Apr 4, 2025, 2:29 PM

#

not yet 🙂

#

it is literally being deployed to prod as I am typing this message. Please check in ~1h or so

raven hinge Apr 14, 2025, 1:45 PM

#

Hi everyone. Looking for someone who can help me understand why my injection of caller_name and other params in the inti payload is not working for outbound calls using my server. It works but it says hello can i speak to caller_name dont mind paying for the help please.

#Just to add to Corentin P message :