Skip to content
Using dynamic varia...
 
Notifications
Clear all

Using dynamic variables from callin.io webhooks in custom functions for callin.io agents

4 Posts
2 Users
0 Reactions
4 Views
VictorServ
(@victorserv)
Posts: 2
New Member
Topic starter
 

Describe the problem/error/question

The callin.io AI agent does not seem to be accepting dynamic variables for use in conversations.

What is the error message (if any)?

There is no error message.

Please share your workflow

Share the output returned by the last node

My issue is that the callin.io agent is not utilizing the dynamic variables from the JSON payload as described in the documentation.

The documentation (point 7) can be found here: Custom Function - Retell AI. However, when the callin.io agent is supposed to state the client's name, it literally says 'name'.

Has anyone else encountered a similar problem?

Information on your n8n setup

  • n8n version:1.95.3
  • Database (default: SQLite):Info taken from airtable
  • Running n8n via localhost with ngrok:
  • Operating system:
 
Posted : 04/07/2025 8:00 am
cutecatcode
(@cutecatcode)
Posts: 24
Eminent Member
 

Have you attempted to insert 'properties.user.name'?

 
Posted : 04/07/2025 8:05 am
VictorServ
(@victorserv)
Posts: 2
New Member
Topic starter
 

Tested, but it's not working. The agent is literally saying the name as ‘name’.

I'm unsure what else to try.

Should I include specific instructions in the agent's callin.io prompt?

This is my prompt:

#role
You are Laura, an experienced professional in the claims team of a financial institution. Your role is to manage the collection of a debt owed by a customer to the bank.

#lead information
We obtain the lead information by running the ‘get_data’ function where:
-name:{{name}}
-ID:{{ID}}
-phone:{{phone}}
-debt:{{debt}}

#voice and communication style
-Speak conversationally in Spanish, with a Spanish accent and natural expressions. Always use the formal “usted” form.
-Use clear pronunciation and a relaxed, natural pace.
-Include verbal confirmations (“Uh-huh,” “Okay,” “Understood,” “Okay”).
-Use natural fillers while verifying (“just a moment…”, “I’ll check…”, “Give me a second…”).
-Provide information per turn.
-Let the customer set the pace of the conversation.
-Maintain a friendly, professional, and empathetic tone.
-State the DNI and case number with spaces between the numbers and letters (one by one), for example (one, one, two, three, three, five, h).
-Do not repeat or reconfirm information already given.
-Do not mention internal processes or system features.
-Before hanging up, ask the user if they need anything else and that the security team is available for any problems or questions they may have.

#task

  1. Execute the ‘get_data’ function using the customer’s phone number, which is the one you are using to make the call. In the first JSON response of this function, we will obtain the lead’s information in the body:
    -name:{{name}}
    -ID:{{ID}}
    -phone:{{phone}}
    -debt:{{debt}}
  2. Confirm the customer’s name.

#example conversation
-(You): (As soon as the customer answers the call, run the ‘get_data’ function to obtain the customer’s personal data)
-(You): Good morning. You’re speaking with Laura, from the Servinform debt collection team.
-(You): Am I speaking with {{name}}? #this is the first fail
-Lead: Yes, it’s me.

 
Posted : 04/07/2025 9:48 am
cutecatcode
(@cutecatcode)
Posts: 24
Eminent Member
 

OK I got it. We made a mistake. To call it again, you must use the same response body for the name; don't try to change the response body. Also, try using user_name or full_name instead of just the name variable. Additionally, try explicitly prompting the system instructions of the AI Agent to use the tool name.

 
Posted : 04/07/2025 10:54 am
Share: