I use the Vapi client-side web sdk to create a call like this:

import Vapi from "@vapi-ai/web";
const vapi = new Vapi("your-public-key-or-jwt");
const call = await vapi.start({ metadata: {...}, model: { provider: 'custom-llm', url: '...' }, .... });

Later on, after the call starts, i need to update the metadata or the model url for this assistant. Problem is, the assistant is not persisted, so I can't use the standard "update assistant" api (I don't think). Is there any way to update the metadata that is sent to the custom-llm? Or update the url for the custom-llm model?

Use the assistant-request webhook to modify the assistant’s configuration in real-time, responding with the updated metadata or model URL in the server’s reply.