Gemini Embedding Error

[frendy_nicodemus]

Subject: Error: expected 768 dimensions, not 3072 when using Google Gemini Embedding

Body: Hi everyone,

I’m hitting a roadblock with my n8n workflow (version 2.1.5) while using the Google Gemini Embedding node (model: embedding-001).

The Issue: When trying to upsert data into my Vector Store, I receive the following error:

“Error inserting: expected 768 dimensions, not 3072. 400 Bad Request”

My Setup:

• n8n Version: 2.1.5

• Model: Google Gemini embedding-001

• Vector Database: Supabase

It seems like Gemini is outputting 3072 dimensions, but my vector index is configured for 768.

Any guidance would be much appreciated. Thanks!

[Anshul_Namdev]

Hi @frendy_nicodemus make sure the embedding dimensions match both in supabase and and store.

Consider reading these:

Scraping a website using crawl4AI and supabase

Describe the problem/error/question i’m trying to scrape a website which has nested xml using crawl4AI and supabase. The outputs at the nodes before the Supabase vector store seemed ok. When the data gets passed to the Supabase vector store, it seemed like the vector store and embedding are looping. kindly advise where i did wrong. tyvmia! What is the error message (if any)? Problem in node ‘Supabase Vector Store‘ Error inserting: expected 1536 dimensions, not 768 400 Bad Request Please shar…

Issue with Vector Dimension Mismatch in Supabase Vector Store: 1536 vs. 2048 Error

I am encountering an issue when trying to insert vectors into the Supabase Vector Store node in n8n. The vectors generated by Ollama embeddings are not matching the expected dimensions required by Supabase. The error I receive states that the expected dimensionality is 1536, but the vectors generated are 2048, resulting in a 400 Bad Request error. Error inserting: expected 1536 dimensions, not 2048 400 Bad Request Please share your workflow

just make sure that in supabase it is using vector(3072) and the error should disappear. Classic dimension mismatch error just make sure they match.

[frendy_nicodemus]

got it already.. thanks !

[Anshul_Namdev]

@frendy_nicodemus Glad it helped! Kindly mark that as a solution to let other’s know where to look for the right answer to your question. Cheers!