- Service Principal
- SAS Token
"""
Azure Blob Storage Content Source for Knowledge
================================================
Load files and folders from Azure Blob Storage containers into your Knowledge base.
Uses Azure AD client credentials flow for authentication.
Features:
- Load single blobs or entire prefixes (folders) recursively
- Supports any Azure Storage Account
- Automatic file type detection and reader selection
- Rich metadata stored for each file (storage account, container, path)
Requirements:
- Azure AD App Registration with:
- Application (client) ID
- Client secret
- Storage Blob Data Reader role on the storage account
- Storage account name and container name
Setup:
1. Register an app in Azure AD (portal.azure.com)
2. Assign "Storage Blob Data Reader" role to the app on your storage account
3. Create a client secret
4. Set environment variables (see below)
Environment Variables:
AZURE_TENANT_ID - Azure AD tenant ID
AZURE_CLIENT_ID - App registration client ID
AZURE_CLIENT_SECRET - App registration client secret
AZURE_STORAGE_ACCOUNT_NAME - Storage account name (without .blob.core.windows.net)
AZURE_CONTAINER_NAME - Container name
Run this cookbook:
python cookbook/07_knowledge/cloud/azure_blob.py
"""
from os import getenv
from agno.knowledge.knowledge import Knowledge
from agno.knowledge.remote_content import AzureBlobConfig
from agno.vectordb.pgvector import PgVector
# Configure Azure Blob Storage content source
# All credentials should come from environment variables
azure_config = AzureBlobConfig(
id="company-docs",
name="Company Documents",
tenant_id=getenv("AZURE_TENANT_ID"),
client_id=getenv("AZURE_CLIENT_ID"),
client_secret=getenv("AZURE_CLIENT_SECRET"),
storage_account=getenv("AZURE_STORAGE_ACCOUNT_NAME"),
container=getenv("AZURE_CONTAINER_NAME"),
)
# Create Knowledge with Azure Blob Storage as a content source
knowledge = Knowledge(
name="Azure Blob Knowledge",
vector_db=PgVector(
table_name="azure_blob_knowledge",
db_url="postgresql+psycopg://ai:ai@localhost:5532/ai",
),
content_sources=[azure_config],
)
if __name__ == "__main__":
# Insert a single file from Azure Blob Storage
print("Inserting single file from Azure Blob Storage...")
knowledge.insert(
name="DeepSeek Paper",
remote_content=azure_config.file("DeepSeek_R1.pdf"),
)
# Insert an entire folder (prefix)
print("Inserting folder from Azure Blob Storage...")
knowledge.insert(
name="Research Papers",
remote_content=azure_config.folder("testfolder/"),
)
# Search the knowledge base
print("Searching knowledge base...")
results = knowledge.search("What is DeepSeek?")
for doc in results:
print(f"- {doc.name}: {doc.content[:100]}...")
"""
Azure Integration: Blob Storage (SAS Token)
=============================================
Load files and folders from Azure Blob Storage containers using SAS token authentication.
Features:
- Load single files or entire prefixes (folders)
- Uses SAS (Shared Access Signature) token for authentication
Requirements:
- Azure Storage Account with a SAS token
Environment Variables:
AZURE_SAS_TOKEN - SAS token
AZURE_STORAGE_ACCOUNT_NAME - Storage account name
AZURE_CONTAINER_NAME - Container name
Run `uv pip install azure-storage-blob` to install dependencies.
"""
import asyncio
from os import getenv
from agno.knowledge.knowledge import Knowledge
from agno.knowledge.remote_content import AzureBlobConfig
from agno.vectordb.chroma import ChromaDb
azure_blob = AzureBlobConfig(
id="company-blob-sas",
name="Company Blob Storage SAS",
sas_token=getenv("AZURE_SAS_TOKEN"),
storage_account=getenv("AZURE_STORAGE_ACCOUNT_NAME"),
container=getenv("AZURE_CONTAINER_NAME"),
)
knowledge = Knowledge(
name="Azure Blob Knowledge (SAS)",
vector_db=ChromaDb(
collection="azure_blob_knowledge_sas",
),
content_sources=[azure_blob],
)
if __name__ == "__main__":
async def main():
await knowledge.ainsert(
name="Report",
remote_content=azure_blob.file("reports/annual-report.pdf"),
)
await knowledge.ainsert(
name="All Docs",
remote_content=azure_blob.folder("documents/"),
)
results = knowledge.search("What were the annual results?")
for doc in results:
print("- %s" % doc.name)
asyncio.run(main())
Run the Example
# Clone and setup repo
git clone https://github.com/agno-agi/agno.git
cd agno/cookbook/07_knowledge/cloud
# Create and activate virtual environment
./scripts/demo_setup.sh
source .venvs/demo/bin/activate
# Optiona: Run PgVector (needs docker)
./cookbook/scripts/run_pgvector.sh
python azure_blob.py