How to fix LangChain deprecation warnings

LangChain has reorganized its package structure in v0.2+ to improve modularity and maintainability. Deprecated imports like from langchain.llms import OpenAI or from langchain.chat_models import OpenAI trigger deprecation warnings. These warnings appear when your code uses outdated import paths or client instantiation patterns, such as:

from langchain.llms import OpenAI
client = OpenAI()
response = client("Hello")

This triggers warnings because LangChain now requires imports from langchain_openai and langchain_community packages, and updated usage patterns.

Update your imports to the new packages and adjust client usage to the current LangChain v0.2+ API. For example, replace deprecated imports with:

from langchain_openai import ChatOpenAI

client = ChatOpenAI(model_name="gpt-4o")
response = client("Hello")
print(response)

This works because ChatOpenAI is the current class for chat models, and model_name replaces older model parameters. Similarly, for embeddings and vectorstores, use OpenAIEmbeddings and FAISS from langchain_openai and langchain_community respectively.

To avoid deprecation warnings in production, always pin your LangChain version to the latest stable release and review the official changelog for breaking changes. Use automated tests to detect warnings early. Implement retry logic and fallback models to handle API changes gracefully. Regularly update your dependencies and refactor code to align with the latest SDK patterns.