ONNX example is not a really full ONNX example

As I notice in example we use exported onnx model only to do prediction, but generation of embeddings still done using pytorch and python.
Which is very limited approach if for example I want to deliver to the system that uses onnx runtime only.
Is there a way to export onnx for calculating embeddings as well?

Start by reviewing the repository's ONNX example and the current PyTorch/Python embedding-generation path. Determine how embeddings are produced and what export boundary the example uses. Done means the example supports generating embeddings through an ONNX Runtime-only workflow, rather than using PyTorch and Python for that step.

Written by the indexing model from the issue text.