Unable to Connect Google Vertex AI Chat Model node to Deployed Hugging Face Model

[Koong]

I deployed a model from Hugging Face using Google Model Garden, and I want to connect it to the Google Vertex AI Chat Model node for inference. However, I am unable to establish the connection successfully.

Has anyone managed to integrate a custom-deployed Hugging Face model with Vertex AI Chat Model node? If so, what are the necessary steps or configurations to make it work?

Any guidance would be greatly appreciated!

[Dandy_Dow]

Hey Koong, let’s try getting that Hugging Face model working with Vertex AI through n8n. Here’s the idea:

1. First, check if the model is properly deployed in Vertex:

   Go to Vertex AI → Model Registry

   Make sure the model has an active endpoint

   Grab the Project ID, Region, and Endpoint ID

2. Try the Vertex AI node (if it recognizes the model):

   Use a credential with Vertex AI User permissions

   Fill in the model ID and your project info

   But if it’s a custom model, the node might not detect it

3. Plan B: Use a raw HTTP Request to the endpoint:

   Set the URL like this:
   https://{region}-aiplatform.googleapis.com/v1/projects/{project-id}/locations/{region}/endpoints/{endpoint-id}:predict

Headers:

{
“Authorization”: “Bearer YOUR_TOKEN”,
“Content-Type”: “application/json”
}

Body:

{
“instances”: [
{ “content”: “your input here” }
]
}

Before plugging it into n8n, test it with curl or Postman to make sure the endpoint responds properly.

If you need help formatting the request or setting up the auth, just let me know.

Dandy