Hello n8n Team and Community, I am facing a persistent issue with the HTTP Request node when trying to connect to the Google Vertex AI API. After an exhaustive debugging process, I believe this is a bug. **Environment:** * n8n Version: 1.110.1 (Cloud) * Google Cloud Project: Has Billing enabled and both “Gemini API” and “Vertex AI API” are active. * Authentication: Using a working `Google OAuth2 API` credential with `cloud-platform` scope. **Summary of Failures:** 1. **400 Bad Request on `:predict` / `:generateContent`:** When making a POST request to a model endpoint (e.g., `…/models/gemini-1.0-pro:predict`), the request is successfully authenticated, but fails with a `400 Bad Request`. The error is always: `Invalid JSON payload received. Unknown name \“\”: Proto fields must have a name.` This happens even with a minimal, hardcoded body: `{{ { “instances”: [ { “prompt”: “Hello” } ] } }}`. 2. **404 Not Found on listing models:** To debug, I tried to list the available models. A `GET` request to the models endpoint (`…/publishers/google/models`) fails with a `404 Not Found` for the entire URL path, even though the Vertex AI API is enabled for the project. This seems to indicate a fundamental issue in how the n8n HTTP Request node is communicating with the `aiplatform.googleapis.com` endpoints. The `Code` node is not an alternative as my instance does not have a “Credentials” tab for it to use OAuth2 credentials. Has anyone successfully connected to the Vertex AI API with OAuth2 using the HTTP Request node recently, or could this be a regression or bug? Thank you.
hello @Rafael_Xavier
Please, share the workflow.
You can select all nodes with ctrl+a, copy it with ctrl+c. Then past the content here after pressing button </> with ctrl+v.
Hey,
Have you solved your issue ? I am looking for connecting to Vertex API through a HTTP request as well.
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.