Hey All! I'm trying to implement a Twilio -> Vapi -> My Server pipeline. I'm close, but just hitting this one final issue. My use case necessitates the use of ephemeral assistants. I'm also trying to use an OpenAI Fine Tuned model. I have the former working with the standard gpt4 model, and the latter working via a separate pipeline that utilizes the Client SDK through the browser, but for the Twilio setup I'm not quite sure how to follow step 4 of the documentation (https://docs.vapi.ai/customization/custom-llm/fine-tuned-openai-models):

Execute a PATCH request to the /assistant endpoint and ensure that model.metadataSendMode is set to off.

I can make PATCH requests all day long, but AFAIK the ephemeral models don't have an ID. Is there some other mechanism I should be following in order to update an ephemeral assistant in flight? Is the use case I'm trying to accomplish supported? Thanks in advance for any and all assistance!

@bestestdev Hello! It's great to hear that you're integrating Twilio with Vapi and your server, and looking to use ephemeral assistants with a fine-tuned OpenAI model. You're correct that ephemeral assistants, by their nature, don't have a persistent ID that you can reference for a PATCH request. For your use case, you would typically configure the assistant's behavior before it's created, rather than updating it in flight. This means setting any necessary configurations, such as the use of a fine-tuned model, at the time of the assistant's creation. Since you're working with ephemeral assistants, you should ensure that the creation request for the assistant includes all the necessary configurations. This would involve specifying the fine-tuned model details as part of the assistant's initial setup. If the documentation or the API isn't clear on how to specify a fine-tuned model for an ephemeral assistant, or if you're encountering limitations with the current setup, it would be best to reach out to Vapi support for more detailed guidance on this specific use case. They may have additional insights or updates that could help you achieve your goal. Please let me know if there's anything else I can assist you with or if you would like me to help you reach out to support for more information on this.

@Vapi I have attempted to set metadataSendMode during the initial assistant request, but it doesn't seem to work. The fact that the documentation explicitly mentions doing a PATCH rather than setting it during creation also kind of indicates that it's probably necessary to do an update.

@bestestdev I understand your concern. If setting

metadataSendMode

during the initial request isn't working and the documentation suggests a PATCH request is necessary, this does present a challenge with ephemeral assistants that don't have a persistent ID. Given this situation, it seems there might be a gap in the documentation or a feature that needs to be addressed by the Vapi team. The best course of action would be to reach out to Vapi support directly for clarification on how to apply settings like

metadataSendMode

to ephemeral assistants, especially since you're working with a fine-tuned OpenAI model. Support should be able to provide you with the correct procedure or update their documentation to reflect the necessary steps for your use case. They might also provide insights into whether this is a feature currently in development or if there's an alternative approach you can take. Would you like me to assist you in reaching out to support for more detailed information on this issue?

@Vapi yes please!

@bestestdev Sweet. Have a great day!