https://github.com/lazauk/aifoundry-hub-sdk
Integrating Azure AI Foundry hub with existing Storage and Key Vault resources in Azure Python SDK.
https://github.com/lazauk/aifoundry-hub-sdk
ai azure foundry hub key-vault python sdk storage studio
Last synced: about 1 month ago
JSON representation
Integrating Azure AI Foundry hub with existing Storage and Key Vault resources in Azure Python SDK.
- Host: GitHub
- URL: https://github.com/lazauk/aifoundry-hub-sdk
- Owner: LazaUK
- License: mit
- Created: 2024-10-21T13:15:33.000Z (7 months ago)
- Default Branch: main
- Last Pushed: 2025-01-20T14:11:19.000Z (4 months ago)
- Last Synced: 2025-04-05T23:42:30.976Z (about 1 month ago)
- Topics: ai, azure, foundry, hub, key-vault, python, sdk, storage, studio
- Language: Jupyter Notebook
- Homepage:
- Size: 36.1 KB
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Azure AI Foundry: Creating Hub resource with Azure Python SDK
This repo explains how to use the **_Azure Python Software Development Kit (SDK)_** to create an Azure AI Foundry _hub_ resource. It demonstrates two scenarios: creating a Hub with bare minimum configuration (automatically provisioned storage and key vault), and creating a Hub with existing Azure Storage and Azure Key Vault resources.
> [!WARNING]
> You can create an Azure AI Foundry hub with an existing Azure Storage account only if:
> - The Storage account uses the _Standard_ performance tier, not _Premium_,
> - Both _Blob_ and _File_ service capabilities are enabled,
> - If storage is of _Azure Data Lake Storage_ type, then its _Hierarchical namespace_ feature is disabled.
## Table of contents:
- [Pre-requisites](https://github.com/LazaUK/AIFoundry-Hub-SDK#pre-requisites)
- [Scenario 1: Hub creation with bare minimum configuration](https://github.com/LazaUK/AIFoundry-Hub-SDK#scenario-1-hub-creation-with-bare-minimum-configuration)
- [Scenario 2: Hub creation with existing dependent resources](https://github.com/LazaUK/AIFoundry-Hub-SDK#scenario-2-hub-creation-with-existing-dependent-resources)
## Pre-requisites
1. To follow these steps, you'll need:
- An Azure subscription with an active Entra ID account,
- An Azure Storage account (for Scenario 2),
- An Azure Key Vault (for Scenario 2),
- Python 3.11 or later.
2. You can install the required ```Azure ML```, ```Azure Identity```, ```Azure Storage``` and ```Azure Key Vault``` Python SDK packages using pip command:
``` PowerShell
pip install -r requirements.txt
```
## Scenario 1: Hub creation with bare minimum configuration
1. To create a Hub resource in Azure, you need to initialise **MLClient** first. The ```DefaultAzureCredential``` class allows your code to check available identities in your environment and retrieve the relevant access token automatically:
``` Python
client = MLClient(
credential = DefaultAzureCredential(),
subscription_id = subscription_id,
resource_group = resource_group
)
```
2. You can then instantiate the Hub object with desired property values, such as _name_, _resource_group_ and Azure region _location_:
``` Python
# Instantiate Hub object
ai_hub_config = Hub(
name = ai_hub_name,
display_name = ai_hub_display_name,
description = ai_hub_description,
location = ai_hub_location,
resource_group = ai_hub_resource_group
)
```
3. You can create a new Hub resource by using the _begin_create()_ function of MLClient's _workspaces_ class with the previously created Python _Hub_ object:
``` Python
ai_hub = client.workspaces.begin_create(ai_hub_config).result()
```
4. If successful, you should see the function's output similar to this:
``` JSON
The deployment request Demo_AI_Hub_1-xxxxxx was accepted. ARM deployment URI for reference:
Creating Key Vault: ( ) .. Done (22s)
Creating Storage Account: ( ) Done (23s)
Creating AzureML Workspace: (Demo_AI_Hub_1 ) ..... Done (36s)
Total time : 1m 2s
```
5. As shown in the above step, the **Azure ML SDK** automatically provisioned dependent **Azure Storage** and **Azure Key Vault** resources. To get detailed information about the newly created Hub resoutrce, you can print its properties as shown below:
``` Python
ai_hub_json = json.dumps(ai_hub._to_dict(), indent=4)
print(ai_hub_json)
```
6. This should generate JSON output similar to this:
``` JSON
{
"name": "Demo_AI_Hub_1",
"location": "swedencentral",
"id": ,
"resource_group": "AAA_AIHUB",
"description": "Demo AI Hub with bare minimum configuration",
"display_name": "Demo AI Hub 1",
"hbi_workspace": false,
"storage_account": ,
"key_vault": ,
"tags": {
"createdByToolkit": "sdk-v2-1.21.1"
},
"public_network_access": "Enabled",
"identity": {
"type": "system_assigned",
"principal_id": "XXX",
"tenant_id": "YYY"
},
"managed_network": {
"isolation_mode": "disabled",
"outbound_rules": []
},
"enable_data_isolation": true,
"default_resource_group": ,
"associated_workspaces": []
}
```
## Scenario 2: Hub creation with existing dependent resources
1. If you already have an existing **Azure Storage** account, you can retrieve its Resource ID either from the Azure portal, or by using the **Azure Management SDK for Storage**:
``` Python
storage_client = StorageManagementClient(
credential = DefaultAzureCredential(),
subscription_id = subscription_id
)
storage_account_id = storage_client.storage_accounts.get_properties(
resource_group,
ai_hub_storage_account
).id
```
> [!Note]
> The returned Storage account's **Resource ID** should look like this: _/subscriptions//resourceGroups//providers/Microsoft.Storage/storageAccounts/_
2. If you already have an existing **Azure Kay Vault**, you can retrieve its Resource ID either from the Azure portal, or by using the **Azure Management SDK for Key Vault**:
``` Python
key_vault_client = KeyVaultManagementClient(
credential = DefaultAzureCredential(),
subscription_id = subscription_id
)
key_vault_id = key_vault_client.vaults.get(
resource_group,
ai_hub_key_vault
).id
```
> [!Note]
> The returned Key Vault's **Resource ID** should look like this: _/subscriptions//resourceGroups//providers/Microsoft.KeyVault/vaults/_
3. You can use the Resource IDs from the above steps to set the values of ```storage_account``` and ```key_vault``` properties of a _Hub_ resource:
``` Python
ai_hub_config = Hub(
name = ai_hub_name,
display_name = ai_hub_display_name,
description = ai_hub_description,
location = ai_hub_location,
resource_group = ai_hub_resource_group,
storage_account = storage_account_id, # Existing storage account's Resource ID
key_vault = key_vault_id # Existing key vault's Resource ID
)
```
4. The next step would be to create a new Hub resource with the **Azure ML SDK**:
``` Python
ai_hub = client.workspaces.begin_create(ai_hub_config).result()
```
5. As our Hub's configuration references existing Storage and Key Vault resources, Azure will create only the Hub resource this time:
``` JSON
The deployment request Demo_AI_Hub_2-xxxxxx was accepted. ARM deployment URI for reference:
Creating AzureML Workspace: (Demo_AI_Hub_2 ) ...... Done (36s)
Total time : 37s
```
6. You can retrieve the Hub's properties with the following code:
``` Python
ai_hub_json = json.dumps(ai_hub._to_dict(), indent=4)
print(ai_hub_json)
```
7. The generated output will confirm that the newly created Hub is configured to use existing Storage and Key Vault resources:
``` JSON
{
"name": "Demo_AI_Hub_2",
"location": "swedencentral",
"id": ,
"resource_group": "AAA_AIHUB",
"description": "Demo AI Hub with existing dependent resources",
"display_name": "Demo AI Hub 2",
"hbi_workspace": false,
"storage_account": ,
"key_vault": ,
"tags": {
"createdByToolkit": "sdk-v2-1.21.1"
},
"public_network_access": "Enabled",
"identity": {
"type": "system_assigned",
"principal_id": "XXX",
"tenant_id": "YYY"
},
"managed_network": {
"isolation_mode": "disabled",
"outbound_rules": []
},
"enable_data_isolation": true,
"default_resource_group": ,
"associated_workspaces": []
}
```