Using LangChain and gpt-3.5-turbo (ChatGPT API) with Azure OpenAI Service

Introduction #

In this post we briefly discuss how LangChain and gpt-35-turbo (the model behind ChatGPT) can be used with Azure OpenAI Service.

Tutorial #

First, create a .env and add your Azure OpenAI Service details:

OPENAI_API_KEY=xxxxxx
OPENAI_API_BASE=https://xxxxxxxx.openai.azure.com/

Next, make sure that you have gpt-35-turbo deployed and used the same name as the model itself for the deployment.

Let’s install/upgrade to the latest versions of openai and langchain via pip:

pip install openai --upgrade
pip install langchain --upgrade

This is super important, as older versions of the openai Python SDK do not support the API version needed to access gpt-35-turbo.

Finally, we can run our sample code:

By setting the openai configuration, we force LangChain (which uses the OpenAI Python SDK under the hood) to talk to Azure instead of OpenAI directly. By specifying openai.api_version = "2023-03-15-preview", we make sure that the openai SDK uses the latest version, which is compatible with the turbo model version.

Summary #

In this blog post, we discussed how we can use the ChatGPT API (gpt-35-turbo model) with Azure OpenAI Service and LangChain.