Chat Completions

Providers might offer params not supported by OpenAI (e.g. top_k). LiteLLM treats any non-openai param, as a provider-specific param, and passes it to the provider in the request body, as a kwarg. See Reserved Params

📄️ Structured Outputs (JSON Mode)

Quick Start

📄️ Pre-fix Assistant Messages

Supported by:

📄️ Drop Unsupported Params

Drop unsupported OpenAI params by your LLM Provider.

📄️ Prompt Formatting

LiteLLM automatically translates the OpenAI ChatCompletions prompt format, to other models. You can control this by setting a custom prompt template for a model as well.

📄️ Output

Format

📄️ Prompt Caching

For OpenAI + Anthropic + Deepseek, LiteLLM follows the OpenAI prompt caching usage object format:

📄️ Usage

LiteLLM returns the OpenAI compatible usage object across all providers.

📄️ Exception Mapping

LiteLLM maps exceptions across all providers to their OpenAI counterparts.

📄️ Streaming + Async

- Streaming Responses

📄️ Trimming Input Messages

Use litellm.trim_messages() to ensure messages does not exceed a model's token limit or specified max_tokens

📄️ Function Calling

Checking if a model supports function calling

📄️ Using Vision Models

Quick Start

📄️ Model Alias

The model name you show an end-user might be different from the one you pass to LiteLLM - e.g. Displaying GPT-3.5 while calling gpt-3.5-turbo-16k on the backend.

📄️ Batching Completion()

LiteLLM allows you to:

📄️ Mock Completion() Responses - Save Testing Costs 💰

For testing purposes, you can use completion() with mock_response to mock calling the completion endpoint.

📄️ Reliability - Retries, Fallbacks

LiteLLM helps prevent failed requests in 2 ways: