Tool not capturing all the information
# supportt
tom
09/02/2024, 7:05 AMHi everyone! Thanks for having me in the community.
I'm new to Vapi and built my first assistant today.
I want to develop it for realtors so it captures information from leads.
That includes personal information, intention (buy/sell/rent), and the object of interest.
It didn't really work though.
The assistant captured only the intention and the object (refer to screenshots). I also added screenshots of how I configured the tool.
The assistant is in German, but it should still be clear what I wanna achieve here.
Would really appreciate some help! Thanks a lot
https://cdn.discordapp.com/attachments/1280060954985496617/1280060955279101972/90b696048d4944509ec2a863932f76ce95af98ea8bb24551ab38c2f9505c1823.png?ex=66d6b52a&is=66d563aa&hm=7b82691397701f35bcca2333ed121b9ae902f64fe3db6c9079ff702c1d8b77ec&
https://cdn.discordapp.com/attachments/1280060954985496617/1280060955710980096/4a3ddac0c12f424ba96ffce6edd255d961952a09e6ba4037b4074c51f1e69765.png?ex=66d6b52a&is=66d563aa&hm=794a84157fc73d10ded01dc05b77dda79938cd93406afb458ec4c79e6ad1e0d2&
https://cdn.discordapp.com/attachments/1280060954985496617/1280060956038270976/b218487b24c94ee1b95fd334383ccc860e3c292890c24389a7f43995e06e1e9c.png?ex=66d6b52a&is=66d563aa&hm=12157c0ac92689abc5503e472611c7d65405f328f06b41ccdec3ace570311c1f&
v
Vapi
09/02/2024, 7:06 AM@tom
Hello and welcome to the VAPI community! It's great to hear that you're diving into creating your own assistant for realtors. I'm here to help you get your assistant up and running as you intend.
To assist you better, could you please provide me with more details about how you've set up your assistant? Specifically, I need to know how you've configured the assistant to capture personal information. This includes any input fields, voice prompts, or data processing logic you've implemented.
Once I have that information, I'll be able to guide you through the troubleshooting process or suggest modifications to ensure your assistant captures all the necessary information from leads.
Vapi
09/02/2024, 7:06 AMYou can continue this conversation by mentioning me in the message.
t
tom
09/02/2024, 7:20 AM@Vapi
I created a tool called 'LeadInformation' and added the properties 'Name', 'Beruf', 'Email', 'Objekt', 'Intention' and 'Nettoeinkommen'. I added a clear description each and configured them as strings.
In the system prompt I added example scripts and where appropriate added...
'## Tool Call:
Property 'Email' in Tool 'LeadInformation' ausführen.'
... so the assistant knows when to capture information.
In my test run it only captured the properties 'Intention' and 'Objekt'.
tom
09/02/2024, 7:41 AM@Vapi
tom
09/02/2024, 8:07 AMI've just come closer to the solution.
I can see in the webhook queue in Make that the missing items are stuck in there. The problem seems to be that I run the same webhook multiple times during the conversation.
For each run only a part of the information is captured (the one currently talked about in the call). So I have 5-6 webhook pings per call.
How can I design this better? So that Vapi captures all the variables but only sends the webhook after the call ended? Or do I have to create a separate webhook for each information I want to capture? That would be a bit inefficient tho.
tom
09/02/2024, 8:07 AM@Vapi
v
Vapi
09/02/2024, 8:07 AM@tom
It's great to hear that you've identified where the issue might be occurring. To streamline the process and ensure that all the information is captured before sending it through the webhook, you can consider the following strategies:
1. **Aggregate Data**: Instead of sending each piece of information separately, you can modify your assistant to collect all the necessary data during the conversation and store it in a temporary structure. Once the conversation is complete, you can then send all the collected data in a single webhook call.
2. **Post-Conversation Trigger**: Configure your assistant to trigger the webhook only after the conversation has ended. This can be done by setting a specific condition or using a 'conversation end' event to initiate the webhook.
3. **Batch Processing**: If your webhook provider supports batch processing, you can accumulate the data and send it in batches at specific intervals or triggers.
4. **Modify Webhook Logic**: If you have control over the webhook's logic, you could modify it to handle multiple inputs and wait for a complete set of data before processing.
5. **Session Management**: Implement session management within your assistant to keep track of the conversation state and ensure that all data points are captured before sending them out.
If you need assistance with implementing any of these strategies or have further questions, please let me know. I'm here to help you make your assistant as efficient as possible!
t
tom
09/02/2024, 8:09 AM@User @User Could a real person assist please? :) Thank you!
j
Jotham
09/02/2024, 8:55 AMIf you are happy to get this data once the call has finished you can use the call analysis feature.
https://docs.vapi.ai/assistants/call-analysis
Jotham
09/02/2024, 8:55 AM@tom
s
Shubham Bajaj
09/02/2024, 10:29 PMthanks @Jotham
Shubham Bajaj
09/02/2024, 10:30 PM@tom use the call analysis and at the end of the call you will get the information you required.
t
tom
09/03/2024, 9:18 AMthank you @Jotham and @Shubham Bajaj , will take a look at this !
tom
09/03/2024, 10:11 AM@Jotham hi, i created a structured data scheme and it captures everything correctly. but i don't quite understand how to send this to Make, i don't see an option to enter a webhook url
j
Jotham
09/03/2024, 10:13 AMIt should be in your end-of-call-report.
So send your end-of-call-report to your make webhook url by setting the serverUrl at the assistant level.
t
tom
09/03/2024, 10:13 AMj
Jotham
09/03/2024, 10:13 AMyeah thats right
t
tom
09/03/2024, 10:13 AMok got it thanks man
10 Views