Why Azure Embedding does not work for supabase vector store?

[Blueli]

Hello

I am trying to build a chat ai with my own document
I am using Azure openai API. embedding model text-embedding-ada-002

![Screenshot 2024-07-31 at 2.21.03 PM|690x498](upload://tbFalS8YqRqTAuBd2e2D0IJWHjU.png

But it does not work, and loads forever.

later I tried
Google Gemini embedding and it works.

And then I use Google Gemini Chat Model, it works well.

But I want to use Azure instead of Gemini. However, I don’t understand why Azure embedding would not work in this case.

Does anyone have any idea what may go wrong?
Thanks

[n8n]

It looks like your topic is missing some important information. Could you provide the following if applicable.

• n8n version:
• Database (default: SQLite):
• n8n EXECUTIONS_PROCESS setting (default: own, main):
• Running n8n via (Docker, npm, n8n cloud, desktop app):
• Operating system:

[aya]

Hi @Blueli , did you by any chance use the Gemini embedding model to generate the embeddings in the vector store? We recommend that you use the same embedding model (or a similar model with the exact same number of dimensions) for indexing and for retrieval.

You also mention that it ‘loads forever’ which could either mean that the service is low or you’re making a lot of iterations. In that case, you could try setting a lower limit in the retriever to see if that helps!

[Blueli]

Hello
So I retry to do to whole thing. The embedding by Azure does not seem to work. The number of dimensions on Supa Table is set to 1536, it should match the ai model.
The same vector table setup for Google Gemini is working only difference is the table number of dimensions is set to 768.

Would you have any idea what else here I should do for Azure model?

Thanks