Hi,
I am having troubles getting embeddings with genai client for Golang. The texts i build the embeddings for are in Czech. The models the client is built for are in v1beta
and therefore are those available in Gemini AI - Gemini API | Google for Developers. So the best one for embeddings would be text-embedding-004
.
There is alternative option available in Vertex AI API - text-multilingual-embedding-002
however is not available in v1beta. See Text embeddings API | Generative AI on Vertex AI | Google Cloud
Now i have two options - use text-embedding-004
or go to use Vertex AI for embeddings to leverage the multilingual model.
Would anyone know if i actually need to use multilingual model with texts in languages other than English or text-embedding-004
supports all languages?
From the present experience i would say text-embedding-004
supports Czech poorly. The similarity search results are quite weird.