Cloud Storage Sources

Load content from S3, GCS, SharePoint, GitHub, and Azure Blob into a knowledge base.

Register cloud storage providers on a Knowledge instance with content_sources. Each provider has .file() and .folder() methods that create content references you pass to knowledge.insert().

from agno.knowledge.knowledge import Knowledge
from agno.knowledge.remote_content import S3Config

knowledge = Knowledge(
    vector_db=vector_db,
    contents_db=contents_db,
    content_sources=[
        S3Config(
            id="company-docs",
            name="Company Documents",
            bucket_name="my-docs-bucket",
            region="us-east-1",
        ),
    ],
)

# Insert a single file
knowledge.insert(
    name="Q4 Report",
    remote_content=knowledge.content_sources[0].file("reports/q4-2025.pdf"),
)

# Insert an entire folder
knowledge.insert(
    name="Engineering Specs",
    remote_content=knowledge.content_sources[0].folder("specs/"),
)

Supported Providers

Provider	Config Class	Install
Amazon S3	S3Config	pip install boto3
Google Cloud Storage	GcsConfig	pip install google-cloud-storage
SharePoint	SharePointConfig	pip install msal requests
GitHub	GitHubConfig	pip install requests
Azure Blob Storage	AzureBlobConfig	pip install azure-identity azure-storage-blob

All configs are importable from agno.knowledge.remote_content.

Provider Configuration

S3Config

from agno.knowledge.remote_content import S3Config

s3 = S3Config(
    id="s3-docs",
    name="S3 Documents",
    bucket_name="my-bucket",
    region="us-east-1",
    aws_access_key_id="...",       # optional, falls back to default credential chain
    aws_secret_access_key="...",   # optional, falls back to default credential chain
    prefix="documents/",           # optional, default prefix for browsing
)

Field	Type	Default	Description
id	str	required	Unique identifier for this source
name	str	required	Display name
bucket_name	str	required	S3 bucket name
region	Optional[str]	None	AWS region
aws_access_key_id	Optional[str]	None	AWS access key. Falls back to default credential chain.
aws_secret_access_key	Optional[str]	None	AWS secret key. Falls back to default credential chain.
prefix	Optional[str]	None	Default prefix for browsing and listing

GcsConfig

from agno.knowledge.remote_content import GcsConfig

gcs = GcsConfig(
    id="gcs-docs",
    name="GCS Documents",
    bucket_name="my-gcs-bucket",
    project="my-gcp-project",
)

Field	Type	Default	Description
id	str	required	Unique identifier
name	str	required	Display name
bucket_name	str	required	GCS bucket name
project	Optional[str]	None	GCP project ID
credentials_path	Optional[str]	None	Path to GCP credentials file
prefix	Optional[str]	None	Default prefix

GitHubConfig

from agno.knowledge.remote_content import GitHubConfig

github = GitHubConfig(
    id="my-repo",
    name="My Repository",
    repo="owner/repo",
    token="ghp_...",
    branch="main",
)

Field	Type	Default	Description
id	str	required	Unique identifier
name	str	required	Display name
repo	str	required	Repository in owner/repo format
token	Optional[str]	None	GitHub personal access token (needs Contents: read)
branch	Optional[str]	None	Branch name
path	Optional[str]	None	Default path filter

SharePointConfig

from agno.knowledge.remote_content import SharePointConfig

sharepoint = SharePointConfig(
    id="sharepoint-docs",
    name="SharePoint Documents",
    tenant_id="...",
    client_id="...",
    client_secret="...",
    hostname="contoso.sharepoint.com",
    site_path="/sites/Engineering",
)

Field	Type	Default	Description
id	str	required	Unique identifier
name	str	required	Display name
tenant_id	str	required	Azure AD tenant ID
client_id	str	required	Azure AD application client ID
client_secret	str	required	Azure AD application client secret
hostname	str	required	SharePoint hostname
site_path	Optional[str]	None	Site path (e.g., /sites/Engineering)
site_id	Optional[str]	None	Full site ID
folder_path	Optional[str]	None	Default folder path

AzureBlobConfig

from agno.knowledge.remote_content import AzureBlobConfig

azure = AzureBlobConfig(
    id="azure-docs",
    name="Azure Blob Documents",
    tenant_id="...",
    client_id="...",
    client_secret="...",
    storage_account="mystorageaccount",
    container="documents",
)

Field	Type	Default	Description
id	str	required	Unique identifier
name	str	required	Display name
tenant_id	str	required	Azure AD tenant ID
client_id	str	required	Azure AD application client ID
client_secret	str	required	Azure AD application client secret
storage_account	str	required	Azure storage account name
container	str	required	Blob container name
prefix	Optional[str]	None	Default prefix

Requires the Storage Blob Data Reader (or Contributor) role on the storage account.

Inserting Content

Each config has .file() and .folder() methods that return content references for knowledge.insert().

# Single file
knowledge.insert(
    name="Architecture Doc",
    remote_content=s3.file("docs/architecture.pdf"),
)

# Entire folder
knowledge.insert(
    name="All Specs",
    remote_content=gcs.folder("specs/"),
)

# GitHub file from a specific branch
knowledge.insert(
    name="README",
    remote_content=github.file("README.md", branch="develop"),
)

# SharePoint file from a specific site
knowledge.insert(
    name="Policy",
    remote_content=sharepoint.file("Shared Documents/policy.pdf", site_path="/sites/HR"),
)

Browsing S3 Files

S3Config supports paginated file listing with list_files(). This is useful for building file pickers or exploring bucket contents before ingesting.

result = s3.list_files(prefix="reports/", limit=50, page=1)

for folder in result.folders:
    print(f"Folder: {folder['name']}")

for file in result.files:
    print(f"File: {file['name']} ({file['size']} bytes)")

print(f"Page {result.page} of {result.total_pages}")

Parameter	Type	Default	Description
prefix	Optional[str]	None	Path prefix filter. Overrides the config’s prefix.
delimiter	str	"/"	Folder delimiter
limit	int	100	Files per page (1-1000)
page	int	1	Page number (1-indexed)

An async variant alist_files() is also available with the same signature.

Multiple Sources

Register multiple providers on a single Knowledge instance.

knowledge = Knowledge(
    vector_db=vector_db,
    contents_db=contents_db,
    content_sources=[s3, gcs, github, sharepoint, azure],
)

# Insert from different sources
knowledge.insert(name="S3 Doc", remote_content=s3.file("doc.pdf"))
knowledge.insert(name="GitHub Doc", remote_content=github.file("README.md"))

When running with AgentOS, registered sources are exposed via the /knowledge/{id}/sources API endpoint for listing and browsing.

Next Steps

Task	Guide
Content types overview	Content Types
Filter search results	Filtering
Set up a vector database	Vector Databases