Litelm: LiteLLM Without the Bloat

litelm implements LiteLLM’s core routing, message translation, streaming, tool use, and embeddings in about 2,900 lines with only OpenAI and httpx as dependencies, removing the surrounding proxy, caching, and cost-tracking code. It supports 19 providers via a “provider/model” syntax, offers synchronous and asynchronous APIs matching LiteLLM’s signatures, and maps provider errors to a unified exception hierarchy. All core functions pass the maintainer’s compatibility audit, with 262 internal tests and full DSPy execution-path verification succeeding.