Langchain Embedding Backend

Feature request

The feature will create a new backend that utilizes LangChain's embedding interface to embed queries via the use of different supported LangChain models.

Motivation

As LangChain currently has an unified interface on the embedding text, which also covers some of the backends currently supported by BERTopic, it would be an interesting direction if we can have all different integrations from LangChain to be brought into BERTopic. This further expands what models that BERTopic can support and when such integration becomes mature, it may potentially help reducing the customized backend and development effort.

Appendix:
https://python.langchain.com/docs/integrations/text_embedding/

Your contribution

Will help the development of adding the Langchain backend. The implementation will take reference on OpenAI's backend(bertopic/backend/_openai.py). More than welcome to let me know if any other requirements that can help this feature.

Start by reading bertopic/backend/_openai.py and the LangChain text-embedding integrations listed in the linked documentation. Compare the existing OpenAI backend with the requirements for a LangChain embedding backend, then verify that supported LangChain models can embed queries through the new integration.