Skip to main content

ChatLiteLLMRouter

LiteLLM is a library that simplifies calling Anthropic, Azure, Huggingface, Replicate, etc.

This notebook covers how to get started with using Langchain + the LiteLLM Router I/O library.

from langchain_community.chat_models import ChatLiteLLMRouter
from langchain_core.messages import HumanMessage
from litellm import Router
model_list = [
    {
        "model_name": "gpt-4",
        "litellm_params": {
            "model": "azure/gpt-4-1106-preview",
            "api_key": "<your-api-key>",
            "api_version": "2023-05-15",
            "api_base": "https://<your-endpoint>.openai.azure.com/",
        },
    },
    {
        "model_name": "gpt-35-turbo",
        "litellm_params": {
            "model": "azure/gpt-35-turbo",
            "api_key": "<your-api-key>",
            "api_version": "2023-05-15",
            "api_base": "https://<your-endpoint>.openai.azure.com/",
        },
    },
]
litellm_router = Router(model_list=model_list)
chat = ChatLiteLLMRouter(router=litellm_router, model_name="gpt-35-turbo")
messages = [
    HumanMessage(
        content="Translate this sentence from English to French. I love programming."
    )
]
chat(messages)
AIMessage(content="J'aime programmer.")

ChatLiteLLMRouter also supports async and streaming functionality:​

from langchain_core.callbacks import CallbackManager, StreamingStdOutCallbackHandler
await chat.agenerate([messages])
LLMResult(generations=[[ChatGeneration(text="J'adore programmer.", generation_info={'finish_reason': 'stop'}, message=AIMessage(content="J'adore programmer."))]], llm_output={'token_usage': {'completion_tokens': 6, 'prompt_tokens': 19, 'total_tokens': 25}, 'model_name': None}, run=[RunInfo(run_id=UUID('75003ec9-1e2b-43b7-a216-10dcc0f75e00'))])
chat = ChatLiteLLMRouter(
    router=litellm_router,
    model_name="gpt-35-turbo",
    streaming=True,
    verbose=True,
    callback_manager=CallbackManager([StreamingStdOutCallbackHandler()]),
)
chat(messages)
J'adore programmer.
AIMessage(content="J'adore programmer.")

Was this page helpful?