Good news! I kindly suggest you to search and read, all what you are looking for is already here 
If you’re self-hosting, the configuration depends entirely on how you’re doing it and, specifically, on what you’re hosting; setting up Litellm is different from Ollama, Qwen or HuggingFace direct interferences.
If your endpoint is Open API compatible just choose Open AI and set your credentials / endpoint. You can saw the default LLMs created for a guide and/or check the specific documentation for the models and your proxy/provider to avoid guessing your settings.