verbal-limeV

Using transient assistants while hiding prompts from web browser

Hi - has anyone used transient assistants in a way which avoids exposing the prompts & config to the web browser?
For example - Ambr uses VAPI in a way where you can just go to dev tools and see what prompts are driving it.
For many uses cases including mine it is not appropriate to leak the prompts in the way. And yet we need transient assistants (I assume) because the details of the configuration and prompts will differ on each use case.

How can you change core prompts for voice agents on each use, without exposing prompts to the web browser?
Was this page helpful?