Notifications
Clear all
Topic starter
Hello,
I'm currently developing a system for a hotel that operates entirely via voice.
The proposed workflow is as follows:
- Initial communication with ElevenLabs.
- ElevenLabs processes and understands the user's query.
- A webhook is triggered to forward the query to an AI agent.
- After a brief pause, the AI agent provides a response, which is then sent back to the last node, keeping the call active on ElevenLabs.
However, the issue I'm encountering is:
When a second question is asked, the system fails to properly engage the AI agent again.
I would greatly appreciate assistance in resolving this.
Additionally, I am open to suggestions for a more optimal system structure!
Posted : 27/04/2025 10:20 pm
What do you mean by:
When I ask a second question, it can’t properly access the AI agent again.
The agent responds with an error? Or no response? Or do you mean it doesn’t have the context of the previous question? (You would use memory for that in some way)
What do you mean by:
asking elevenlabs a question
Do you mean sending a query for it to it or using the agents/conversation ai feature? Can you show some screenshots of the setup on elevenlabs side?
Posted : 28/04/2025 3:55 am
Topic starter
I mean
I have connected the AI agent ElevenLabs to the webhook. This agent in ElevenLabs has a tool called callin.io and only one parameter: the user question. I've instructed it that when it receives a user question, it should send it to the callin.io webhook. (This is just the initial concept and needs further refinement).
Then, the LLM provides an answer, which is sent back to the response webhook.
So, what I'm doing is initiating the workflow in callin.io and simultaneously starting the AI agent in ElevenLabs. When I pose a question to the agent, it triggers the webhook and generates a response. However, this only happens once. The call remains within ElevenLabs, but when I ask another question, for instance, about the type of room number...
it can't query the LLM again.
I hope this clarifies my situation:
This video demonstrates something similar to what I've implemented: https://youtu.be/pJYm7MOyMH0?si=WjbiyTJLzFgwuaeS
Posted : 28/04/2025 12:34 pm
This thread was automatically closed 90 days following the last response. New replies are no longer permitted.
Posted : 27/07/2025 12:35 pm