Support for custom HF embedding model

[tranhoangnguyen03]

I see here that you are encoding the user query and compare them against the encoded utterances in each route and map the query to the closest utterance>route. Rather than using the Cohere or OpenAI embedding models and suffer the latency and expenses from calling their apis, it would be great to be able perform routing use a locally deployed embedding model from HuggingFace. This also opens up the opportunity to use custom embedding model finetuned for specific domain (health-care in my case).

Thanks for the great project. And I can't wait to see it develop further.

[jamescalam]

We want to add this as soon as possible, we're working on it!

[jamescalam]

solved with #90