azure-containerregistry-py – skill DevOps & Hạ tầng cho agent AI

Azure Container Registry SDK for Python

Manage container images, artifacts, and repositories in Azure Container Registry.

Installation

pip install azure-containerregistry

Environment Variables

AZURE_CONTAINERREGISTRY_ENDPOINT=https://<registry-name>.azurecr.io

Authentication

Entra ID (Recommended)

import os
from azure.containerregistry import ContainerRegistryClient
from azure.identity import DefaultAzureCredential

client = ContainerRegistryClient(
    endpoint=os.environ["AZURE_CONTAINERREGISTRY_ENDPOINT"],
    credential=DefaultAzureCredential()
)

Anonymous Access (Public Registry)

from azure.containerregistry import ContainerRegistryClient

client = ContainerRegistryClient(
    endpoint="https://mcr.microsoft.com",
    credential=None,
    audience="https://mcr.microsoft.com"
)

List Repositories

client = ContainerRegistryClient(endpoint, DefaultAzureCredential())

for repository in client.list_repository_names():
    print(repository)

Repository Operations

Get Repository Properties

properties = client.get_repository_properties("my-image")
print(f"Created: {properties.created_on}")
print(f"Modified: {properties.last_updated_on}")
print(f"Manifests: {properties.manifest_count}")
print(f"Tags: {properties.tag_count}")

Update Repository Properties

from azure.containerregistry import RepositoryProperties

client.update_repository_properties(
    "my-image",
    properties=RepositoryProperties(
        can_delete=False,
        can_write=False
    )
)

Delete Repository

client.delete_repository("my-image")

List Tags

for tag in client.list_tag_properties("my-image"):
    print(f"{tag.name}: {tag.created_on}")

Filter by Order

from azure.containerregistry import ArtifactTagOrder

# Most recent first
for tag in client.list_tag_properties(
    "my-image",
    order_by=ArtifactTagOrder.LAST_UPDATED_ON_DESCENDING
):
    print(f"{tag.name}: {tag.last_updated_on}")

Manifest Operations

List Manifests

from azure.containerregistry import ArtifactManifestOrder

for manifest in client.list_manifest_properties(
    "my-image",
    order_by=ArtifactManifestOrder.LAST_UPDATED_ON_DESCENDING
):
    print(f"Digest: {manifest.digest}")
    print(f"Tags: {manifest.tags}")
    print(f"Size: {manifest.size_in_bytes}")

Get Manifest Properties

manifest = client.get_manifest_properties("my-image", "latest")
print(f"Digest: {manifest.digest}")
print(f"Architecture: {manifest.architecture}")
print(f"OS: {manifest.operating_system}")

Update Manifest Properties

from azure.containerregistry import ArtifactManifestProperties

client.update_manifest_properties(
    "my-image",
    "latest",
    properties=ArtifactManifestProperties(
        can_delete=False,
        can_write=False
    )
)

Delete Manifest

# Delete by digest
client.delete_manifest("my-image", "sha256:abc123...")

# Delete by tag
manifest = client.get_manifest_properties("my-image", "old-tag")
client.delete_manifest("my-image", manifest.digest)

Tag Operations

Get Tag Properties

tag = client.get_tag_properties("my-image", "latest")
print(f"Digest: {tag.digest}")
print(f"Created: {tag.created_on}")

Delete Tag

client.delete_tag("my-image", "old-tag")

Upload and Download Artifacts

from azure.containerregistry import ContainerRegistryClient

client = ContainerRegistryClient(endpoint, DefaultAzureCredential())

# Download manifest
manifest = client.download_manifest("my-image", "latest")
print(f"Media type: {manifest.media_type}")
print(f"Digest: {manifest.digest}")

# Download blob
blob = client.download_blob("my-image", "sha256:abc123...")
with open("layer.tar.gz", "wb") as f:
    for chunk in blob:
        f.write(chunk)

Async Client

from azure.containerregistry.aio import ContainerRegistryClient
from azure.identity.aio import DefaultAzureCredential

async def list_repos():
    credential = DefaultAzureCredential()
    client = ContainerRegistryClient(endpoint, credential)
    
    async for repo in client.list_repository_names():
        print(repo)
    
    await client.close()
    await credential.close()

Clean Up Old Images

Do not execute deletion directly from age or missing tags alone: untagged digests can still be deployed or referenced by multi-architecture manifests. Inventory dependencies and current workloads, present the exact candidate digests and obtain deletion authorization with a retention/recovery policy. Use a dry-run candidate report before the mutating loop below.

from datetime import datetime, timedelta, timezone

cutoff = datetime.now(timezone.utc) - timedelta(days=30)

for manifest in client.list_manifest_properties("my-image"):
    if manifest.last_updated_on < cutoff and not manifest.tags:
        print(f"Deleting {manifest.digest}")
        client.delete_manifest("my-image", manifest.digest)

Client Operations

| Operation | Description |

|-----------|-------------|

| list_repository_names | List all repositories |

| get_repository_properties | Get repository metadata |

| delete_repository | Delete repository and all images |

| list_tag_properties | List tags in repository |

| get_tag_properties | Get tag metadata |

| delete_tag | Delete specific tag |

| list_manifest_properties | List manifests in repository |

| get_manifest_properties | Get manifest metadata |

| delete_manifest | Delete manifest by digest |

| download_manifest | Download manifest content |

| download_blob | Download layer blob |

Best Practices

1. Use Entra ID for authentication in production

2. Delete by digest not tag to avoid orphaned images

3. Lock production images with can_delete=False

4. Clean up untagged manifests regularly

5. Use async client for high-throughput operations

6. Order by last_updated to find recent/old images

7. Check manifest.tags before deleting to avoid removing tagged images

When to Use

This skill is applicable to execute the workflow or actions described in the overview.

Limitations

• Use this skill only when the task clearly matches the scope described above.
• Do not treat the output as a substitute for environment-specific validation, testing, or expert review.
• Stop and ask for clarification if required inputs, permissions, safety boundaries, or success criteria are missing.