Gemini node don't use custom host for completions/embeddings.

[laxtiz]

Bug Description

I use the Cloudflare AI Gateway proxied for Google Gemini.
I noticed one thing: GooglePlamApi.credentials supports a custom host attribute, but LmChatGoogleGemini and EmbeddingaGoogleGemini nodes only use this host attribute to retrieve the model list, and do not use the custom host in chat completions and embeddings requests.
In contrast, LmChatOpenAi uses the custom base_url parameter.

To Reproduce

1. Create a GooglePlamApi credentials with a custom host (e.g., cloudflare ai gateway).
2. Use the LmChatGoogleGemini node in the workflow.
3. Chat with any Gemini model.
4. The logs only requests for list models and do not include chat completion requests.

Expected behavior

The ChatGoogleGenerativeAI class's constructor parameters include base_url, similar to the LmChatOpenAi node, where the base_url parameter is appended when creating an instance of the model.

Operating System

Debian 12.8

n8n Version

1.72.1

Node.js Version

20.18.0

Database

SQLite (default)

Execution mode

main (default)

[Joffcom]

Hey @laxtiz,

We have created an internal ticket to look into this which we will be tracking as "N8N-8041"