[Feature Request] Add LMStudio compatibility to OpenAI embeddings.

[larspohlmann]

Current situation:

When I connect to a local LMStudio server to create embeddings the call sends an array of tokens to /v1/embeddings

LMStudio accepts only an array of strings. Otherwise it will throw this error: 'input' field must be a string or an array of strings

Workaround:

I was able to create that behaviour by editing the call to OpenAIEmbeddings() in config.py.

Currently it looks like this:
return OpenAIEmbeddings( model=model, api_key=RAG_OPENAI_API_KEY, openai_api_base=RAG_OPENAI_BASEURL, openai_proxy=RAG_OPENAI_PROXY, )

If I add the parameter "check_embedding_ctx_length=False"

return OpenAIEmbeddings( check_embedding_ctx_length=False, model=model, api_key=RAG_OPENAI_API_KEY, openai_api_base=RAG_OPENAI_BASEURL, openai_proxy=RAG_OPENAI_PROXY, )

It will work with LMStudio.

Proposal:

I propose to add an environment variable to be able to switch to this behaviour.