I am having an issue setting up a custom llm. I'm trying to set up a workflow on n8n as my custom llm and have entered the test webhook url in the custom llm URL field on my assistant. I'm using a self-hosted instance of n8n which is on a GCP VM. I know the webhook is accessible because I can ding it from postman. When I call the assistant, it says the greeting but as soon as i speak it disconnects and says the room was deleted. The log just says 404 status code and (no body). The issue is that it does not ever hit the url. n8n just keeps waiting for a test event. I also tried setting the assistant back to using openai, and created a tool that uses that same webhook url (defined the tool as n8n test and said to call this tool when i ask to test n8n). I called the assistant, asked it to test n8n and voila! it hit the webhook in n8n. So, what am i missing? is there supposed to be something in the model field? i tried GPT-3.5-TURBO but that made no difference. I know the n8n workflow is not totally set up to complete a chat, first step is to just get it to hit the webhook url. Can you help?

@BrentPope could you share the call id so I can take a look?

Never mind, i figured it out. I didn't realize vapi had added /chat/completions to my webhook url. I added that in n8n and now it hits the server.

{ "name": "My Assistant", "model": { "provider": "custom-llm", "url": "", "model": "my-cool-model", "messages": [ { "role": "system", "content": "You are an assistant." } ], "temperature": 0.7 } }

@Angel Martinez please create a new #1211483291191083018 ticket for your issue.

How

Go to Support Channel and create your own issue.