How to fix LangChain deprecated import langchain.embeddings
langchain.embeddings is deprecated. Replace it with from langchain_openai import OpenAIEmbeddings to import embeddings correctly. This aligns with LangChain v0.2+ SDK patterns and prevents import errors.code_error from langchain.embeddings import ... with from langchain_openai import OpenAIEmbeddings in your code.Why this happens
LangChain reorganized its package structure in v0.2+, deprecating the langchain.embeddings module. Attempting to import embeddings using from langchain.embeddings import OpenAIEmbeddings triggers an ImportError or a deprecation warning. This happens because embeddings were moved to a dedicated package langchain_openai to better modularize the SDK.
Example of broken code:
from langchain.embeddings import OpenAIEmbeddings
embeddings = OpenAIEmbeddings()
This import no longer works and will cause runtime errors.
from langchain.embeddings import OpenAIEmbeddings
embeddings = OpenAIEmbeddings()ImportError: cannot import name 'OpenAIEmbeddings' from 'langchain.embeddings'
The fix
Update your imports to use the new modular packages. Replace from langchain.embeddings import OpenAIEmbeddings with from langchain_openai import OpenAIEmbeddings. This matches the current LangChain v0.2+ SDK pattern and ensures compatibility.
Example of fixed code:
from langchain_openai import OpenAIEmbeddings
embeddings = OpenAIEmbeddings()
print(type(embeddings))<class 'langchain_openai.embeddings.OpenAIEmbeddings'>
Preventing it in production
To avoid similar import issues in production, always pin your LangChain version and review the official changelog for breaking changes. Use automated tests to catch import errors early. Consider adding retry or fallback logic if your app depends on dynamic imports or plugin loading.
Also, use the recommended import paths from the LangChain documentation and avoid deprecated modules.
Key Takeaways
- Always use
from langchain_openai import OpenAIEmbeddingsinstead of deprecatedlangchain.embeddings. - Pin LangChain versions and monitor changelogs to catch breaking changes early.
- Use automated tests to detect import errors before deployment.