What is the LiteLLM SDK supposed to do?

[ratkins]

A cursory look at the Github front page leaves me with the impression that it provides a unified interface to multiple (hosted and local) LLMs.

Deeper investigation into the docs seems to imply there are different LiteLLM APIs for different model providers. I am confused.

(What doesn't help is that I can't find rendered versions of the Python SDK API docs. What types does completion() return? When should I use completion() and when should I use text_completion() are both questions I can't find answers to.)

[krrishdholakia]

Hey @ratkins,

.completion -> unifies to OpenAI's /chat/completions API

.text_completion -> unifies to OpenAI's /completions API

.responses -> unifies to OpenAI's /responses API

Does this help?

Where in docs / readme would this have been helpful to have?

[ratkins]

It helps somewhat; I think what misdirected me is that in the docs, "LiteLLM AI Gateway" and "Supported Endpoints" are at the same level as "LiteLLM Python SDK". "Supported Endpoints" is only relevant if you're using LiteLLM as a gateway, so "Supported Endpoints" should be under "Lite LLM Gateway" not at the top level.

Further, there are no actual API docs (specifying method signatures and return types) under the "LiteLLM Python SDK" heading which is where I would have expected to find them.

Under "LiteLLM Python SDK / Responses API" there's the line "Use litellm.responses() for advanced models that support reasoning content like GPT-5, o3, etc.", that is helpful. I'd like to see a "Completion API" and "Text Completion API" (etc) heading at that same level with a similar one or two lines explaining when to use those APIs. As well as the rest of the rendered Pydoc for those classes.