One Interface, Different Backends

Normalize the call shape

LiteLLM is a Python library that gives many model providers one common interface shaped like OpenAI's chat-completion API. Application code sends the same message structure and receives a normalized response, while configuration chooses Ollama or another backend, the service that actually performs the work.

For local Ollama, a model identifier such as ollama/qwen2.5:0.5b makes the provider explicit. Provider portability does not mean models behave identically; prompts, context sizes, and output quality still need tests.

Analogy: A travel adapter standardizes the plug shape. It does not make every country's power grid identical.