Add Anyscale endpoints

[plaguss]

Description

This PR adds anyscale endpoints on top of the OpenAI integration.

Would close #121

[alvarobartt]

LGTM, but shouldn't we add a check to ensure that OpenAILLM and AnyscaleLLM are not used within the same Pipeline? As AFAIK that would fail since the OPENAI_BASE_URL is modified as well as the OPENAI_API_KEY, right?

[alvarobartt]

LGTM, but shouldn't we add a check to ensure that OpenAILLM and AnyscaleLLM are not used within the same Pipeline? As AFAIK that would fail since the OPENAI_BASE_URL is modified as well as the OPENAI_API_KEY, right?

I guess it would work fine as long as we don't rely on the environment variables, right? But unsure on how that's internally handled by OpenAI, will the provided key via arg override whatever is on the environment variable?

[gabrielmbmb]

LGTM! Could we add some unit tests to the OpenAILLM class similar to the ones from VertexAILLM?

[plaguss]

LGTM, but shouldn't we add a check to ensure that OpenAILLM and AnyscaleLLM are not used within the same Pipeline? As AFAIK that would fail since the OPENAI_BASE_URL is modified as well as the OPENAI_API_KEY, right?

I guess it would work fine as long as we don't rely on the environment variables, right? But unsure on how that's internally handled by OpenAI, will the provided key via arg override whatever is on the environment variable?

It seems that the arg variable takes precedence in the OpenAI API, so it should work fine as long as we don't use both OpenAILLM and AnyscaleLLM and use the same names for the environment variables. I've updated the docs to take this into account, it will look for ANYSCALE_API_KEY instead of OPENAI_API_KEY in case someone prefers the environment variables.