Issues with litellm embedding configuration

[sidnarayanan]

I noticed that my embeddings kept getting rate-limited without retries. On investigation:

• embed_kwargs is a field, but is not consumed anywhere: https://github.com/Future-House/llm-client/blob/main/llmclient/embeddings.py#L77-L80. We should probably remove it
• config["kwargs"] appears to be the way to do it, and we do have a validator to set a default, but validate_default is not set, and so it is not called:

>>> LiteLLMEmbeddingModel().config
{}

For now, if one wants to retry, do something like:

LiteLLMEmbeddingModel(config={"kwargs": {"num_retries": 3, "timeout": 60}})

I can put a patch in later, but wanted to solicit feedback on if anyone wants embed_kwargs or if we can put everything in config.

[sidnarayanan]

FWIW, I've been seeing embedding rate limits in my PQAL logs recently. Not sure how paper-qa handles rate limits.