Azure Storage Blob

12.28.0 · active · verified Wed Mar 25

Official Azure Blob Storage client library for Python. Current version: 12.28.0 (Mar 2026). v12 (released 2019) is a complete rewrite from v2.x — BlockBlobService removed, replaced by BlobServiceClient/ContainerClient/BlobClient. Most LLM training data and older tutorials use the removed v2 BlockBlobService API. Recommended auth is DefaultAzureCredential (azure-identity) not connection strings in production.

Warnings

breaking BlockBlobService removed in v12. Raises ImportError: 'cannot import name BlockBlobService'. This is the #1 error when using LLM-generated Azure storage code.
Fix: from azure.storage.blob import BlobServiceClient — not BlockBlobService
breaking create_blob_from_path(), create_blob_from_text(), create_blob_from_stream(), get_blob_to_path(), get_blob_to_text() all removed in v12. Use upload_blob() and download_blob() instead.
Fix: blob_client.upload_blob(data) and blob_client.download_blob().readall()
breaking SAS token generation moved from class method to standalone function in v12. Old: service.generate_account_shared_access_signature(). New: generate_account_sas() top-level function.
Fix: from azure.storage.blob import generate_account_sas, ResourceTypes, AccountSasPermissions
gotcha upload_blob() raises ResourceExistsError if blob already exists. Default overwrite=False. Must pass overwrite=True to replace.
Fix: blob_client.upload_blob(data, overwrite=True)
gotcha download_blob() returns a StorageStreamDownloader — not bytes directly. Must call .readall() or .read() to get content.
Fix: content = blob_client.download_blob().readall()
gotcha Using connection strings in production exposes account keys. Microsoft recommends DefaultAzureCredential with managed identity for production deployments.
Fix: pip install azure-identity; use DefaultAzureCredential() as credential parameter.
gotcha Async client requires separate import: from azure.storage.blob.aio import BlobServiceClient. Mixing sync and async clients causes TypeError.
Fix: from azure.storage.blob.aio import BlobServiceClient for async usage.

Install

pip install azure-storage-blob Python (blob storage)
pip install azure-storage-blob azure-identity Python (with DefaultAzureCredential — recommended)

Imports

BlobServiceClient (v12 — connection string)

from azure.storage.blob import BlobServiceClient
import os

# From connection string
connect_str = os.environ['AZURE_STORAGE_CONNECTION_STRING']
blob_service_client = BlobServiceClient.from_connection_string(connect_str)

# Upload
container_client = blob_service_client.get_container_client('mycontainer')
with open('file.txt', 'rb') as data:
    container_client.upload_blob('myblob', data, overwrite=True)

# Download
blob_client = blob_service_client.get_blob_client(
    container='mycontainer', blob='myblob'
)
with open('downloaded.txt', 'wb') as f:
    f.write(blob_client.download_blob().readall())

BlockBlobService, create_blob_from_path, get_blob_to_path, get_blob_to_text all removed in v12. LLMs trained pre-2020 generate these patterns. Use BlobServiceClient + upload_blob/download_blob.

DefaultAzureCredential (recommended for production)

from azure.identity import DefaultAzureCredential
from azure.storage.blob import BlobServiceClient

# Passwordless auth — works locally (CLI/VS Code) and in Azure (managed identity)
credential = DefaultAzureCredential()
blob_service_client = BlobServiceClient(
    account_url='https://myaccount.blob.core.windows.net',
    credential=credential
)

# Upload
blob_client = blob_service_client.get_blob_client(
    container='mycontainer', blob='myblob'
)
with open('file.txt', 'rb') as data:
    blob_client.upload_blob(data, overwrite=True)

Connection strings expose account keys. Use DefaultAzureCredential + managed identity in production. Connection strings are acceptable for local dev only.

Quickstart

azure-storage-blob v12 with DefaultAzureCredential upload/download.

# pip install azure-storage-blob azure-identity
from azure.identity import DefaultAzureCredential
from azure.storage.blob import BlobServiceClient

# Production: passwordless auth
credential = DefaultAzureCredential()
blob_service_client = BlobServiceClient(
    account_url='https://myaccount.blob.core.windows.net',
    credential=credential
)

# Create container
container_client = blob_service_client.create_container('mycontainer')

# Upload from file
blob_client = blob_service_client.get_blob_client(
    container='mycontainer', blob='sample.txt'
)
with open('sample.txt', 'rb') as data:
    blob_client.upload_blob(data)

# List blobs
for blob in blob_service_client.get_container_client('mycontainer').list_blobs():
    print(blob.name)

# Download
downloader = blob_client.download_blob()
content = downloader.readall()
print(content)

view raw JSON →