Introducing langchain-azure-storage: Azure Storage integrations for LangChain

Microsoft

Oct 30, 2025

We're excited to introduce langchain-azure-storage, the first official Azure Storage integration package built by Microsoft for LangChain 1.0. As part of its launch, we've built a new Azure Blob Storage document loader (currently in public preview) that improves upon prior LangChain community implementations. This new loader unifies both blob and container level access, simplifying loader integration. More importantly, it offers enhanced security through default OAuth 2.0 authentication, supports reliably loading millions to billions of documents through efficient memory utilization, and allows pluggable parsing, so you can leverage other document loaders to parse specific file formats.

What are LangChain document loaders?

A typical Retrieval‑Augmented Generation (RAG) pipeline follows these main steps:

Collect source content (PDFs, DOCX, Markdown, CSVs) — often stored in Azure Blob Storage.
Parse into text and associated metadata (i.e., represented as LangChain Document objects).
Chunk + embed those documents and store in a vector store (e.g., Azure AI Search, Postgres pgvector, etc.).
At query time, retrieve the most relevant chunks and feed them to an LLM as grounded context.

LangChain document loaders make steps 1–2 turnkey and consistent so the rest of the stack (splitters, vector stores, retrievers) “just works”. See this LangChain RAG tutorial for a full example of these steps when building a RAG application in LangChain.

How can the Azure Blob Storage document loader help?

The langchain-azure-storage package offers the AzureBlobStorageLoader, a document loader that simplifies retrieving documents stored in Azure Blob Storage for use in a LangChain RAG application. Key benefits of the AzureBlobStorageLoader include:

Flexible loading of Azure Storage blobs to LangChain Document objects. You can load blobs as documents from an entire container, a specific prefix within a container, or by blob names. Each document loaded corresponds 1:1 to a blob in the container.
Lazy loading support for improved memory efficiency when dealing with large document sets. Documents can now be loaded one-at-a-time as you iterate over them instead of all at once.
Automatically uses DefaultAzureCredential to enable seamless OAuth 2.0 authentication across various environments, from local development to Azure-hosted services. You can also explicitly pass your own credential (e.g., ManagedIdentityCredential, SAS token).
Pluggable parsing. Easily customize how documents are parsed by providing your own LangChain document loader to parse downloaded blob content.

Using the Azure Blob Storage document loader

Installation

To install the langchain-azure-storage package, run:

pip install langchain-azure-storage

Loading documents from a container

To load all blobs from an Azure Blob Storage container as LangChain Document objects, instantiate the AzureBlobStorageLoader with the Azure Storage account URL and container name:

from langchain_azure_storage.document_loaders import AzureBlobStorageLoader

loader = AzureBlobStorageLoader(
    "https://<your-storage-account>.blob.core.windows.net/",
    "<your-container-name>"
)

# lazy_load() yields one Document per blob for all blobs in the container
for doc in loader.lazy_load():
    print(doc.metadata["source"])  # The "source" metadata contains the full URL of the blob
    print(doc.page_content)  # The page_content contains the blob's content decoded as UTF-8 text

Loading documents by blob names

To only load specific blobs as LangChain Document objects, you can additionally provide a list of blob names:

from langchain_azure_storage.document_loaders import AzureBlobStorageLoader

loader = AzureBlobStorageLoader(
    "https://<your-storage-account>.blob.core.windows.net/",
    "<your-container-name>",
    ["<blob-name-1>", "<blob-name-2>"]
)

# lazy_load() yields one Document per blob for only the specified blobs
for doc in loader.lazy_load():
    print(doc.metadata["source"])  # The "source" metadata contains the full URL of the blob
    print(doc.page_content)  # The page_content contains the blob's content decoded as UTF-8 text

Pluggable parsing

By default, loaded Document objects contain the blob's UTF-8 decoded content. To parse non-UTF-8 content (e.g., PDFs, DOCX, etc.) or chunk blob content into smaller documents, provide a LangChain document loader via the loader_factory parameter.

When loader_factory is provided, the AzureBlobStorageLoader processes each blob with the following steps:

Downloads the blob to a new temporary file
Passes the temporary file path to the loader_factory callable to instantiate a document loader
Uses that loader to parse the file and yield Document objects
Cleans up the temporary file

For example, below shows parsing PDF documents with the PyPDFLoader from the langchain-community package:

from langchain_azure_storage.document_loaders import AzureBlobStorageLoader
from langchain_community.document_loaders import PyPDFLoader  # Requires langchain-community and pypdf packages

loader = AzureBlobStorageLoader(
    "https://<your-storage-account>.blob.core.windows.net/",
    "<your-container-name>",
    prefix="pdfs/",  # Only load blobs that start with "pdfs/"
    loader_factory=PyPDFLoader  # PyPDFLoader will parse each blob as a PDF
)

# Each blob is downloaded to a temporary file and parsed by PyPDFLoader instance
for doc in loader.lazy_load():
    print(doc.page_content)  # Content parsed by PyPDFLoader (yields one Document per page in the PDF)

This file path-based interface allows you to use any LangChain document loader that accepts a local file path as input, giving you access to a wide range of parsers for different file formats.

Migrating from community document loaders to `langchain-azure-storage`

If you're currently using AzureBlobStorageContainerLoader or AzureBlobStorageFileLoader from the langchain-community package, the new AzureBlobStorageLoader provides an improved alternative. This section provides step-by-step guidance for migrating to the new loader.

Steps to migrate

To migrate to the new Azure Storage document loader, make the following changes:

Depend on the langchain-azure-storage package
Update import statements from langchain_community.document_loaders to langchain_azure_storage.document_loaders.
Change class names from AzureBlobStorageFileLoader and AzureBlobStorageContainerLoader to AzureBlobStorageLoader.
Update document loader constructor calls to:
1. Use an account URL instead of a connection string.
2. Specify UnstructuredLoader as the loader_factory to continue to use Unstructured for parsing documents.
Enable Microsoft Entra ID authentication in environment (e.g., run az login or configure managed identity) instead of using connection string authentication.

Migration samples

Below shows code snippets of what usage patterns look like before and after migrating from langchain-community to langchain-azure-storage:

Before migration

from langchain_community.document_loaders import AzureBlobStorageContainerLoader, AzureBlobStorageFileLoader

container_loader = AzureBlobStorageContainerLoader(
    "DefaultEndpointsProtocol=https;AccountName=<account>;AccountKey=<account-key>;EndpointSuffix=core.windows.net",
    "<container>",
)

file_loader = AzureBlobStorageFileLoader(
    "DefaultEndpointsProtocol=https;AccountName=<account>;AccountKey=<account-key>;EndpointSuffix=core.windows.net",
    "<container>",
    "<blob>"
)

After migration

from langchain_azure_storage.document_loaders import AzureBlobStorageLoader
from langchain_unstructured import UnstructuredLoader  # Requires langchain-unstructured and unstructured packages

container_loader = AzureBlobStorageLoader(
    "https://<account>.blob.core.windows.net",
    "<container>",
    loader_factory=UnstructuredLoader  # Only needed if continuing to use Unstructured for parsing
)

file_loader = AzureBlobStorageLoader(
    "https://<account>.blob.core.windows.net",
    "<container>",
    "<blob>",
    loader_factory=UnstructuredLoader  # Only needed if continuing to use Unstructured for parsing
)

What's next?

We're excited for you to try the new Azure Blob Storage document loader and would love to hear your feedback! Here are some ways you can help shape the future of langchain-azure-storage:

Show support for interface stabilization - The document loader is currently in public preview and the interface may change in future versions based on feedback. If you'd like to see the current interface marked as stable, upvote the proposal PR to show your support.
Report issues or suggest improvements - Found a bug or have an idea to make the document loaders better? File an issue on our GitHub repository.
Propose new LangChain integrations - Interested in other ways to use Azure Storage with LangChain (e.g., checkpointing for agents, persistent memory stores, retriever implementations)? Create a feature request or write to us to let us know.

Your input is invaluable in making langchain-azure-storage better for the entire community!