Skip to content

deploy_customization.md

Latest commit

 

History

History
122 lines (84 loc) · 3.95 KB

deploy_customization.md

File metadata and controls

122 lines (84 loc) · 3.95 KB

Deployment Customization

This document describes how to customize the deployment of the Video-Agents-Foundry-Solution. Set these values before running azd up.

Customizing AKS Node Pools

The solution deploys an AKS cluster with four node pools. You can customize the VM sizes for each pool:

VM Size Selection

During deployment, the preprovision hook provides an interactive menu to select VM sizes with live quota checking. Alternatively, set environment variables before running azd up:

azd env set SYSTEM_VM_SIZE Standard_D4a_v4
azd env set WORKLOAD_VM_SIZE Standard_D32a_v4
azd env set DEEPSTREAM_GPU_VM_SIZE Standard_NC24ads_A100_v4
azd env set INFERENCE_GPU_VM_SIZE Standard_NC24ads_A100_v4

Other AKS Settings

Set the Kubernetes version:

azd env set KUBERNETES_VERSION 1.32

Set the maximum node count for the DeepStream GPU node pool:

azd env set DEEPSTREAM_GPU_MAX_NODE_COUNT 1

Customizing AI Foundry

By default, the solution provisions an Azure AI Foundry project with a model deployment. You can customize or disable this.

Disable AI Foundry

If you don't need the AI Foundry component:

azd env set CREATE_FOUNDRY_PROJECT false

Change the AI Model

Change the model name (default: gpt-5.2):

azd env set AI_MODEL_NAME gpt-5.2

Change the model version (default: 2025-12-11):

azd env set AI_MODEL_VERSION 2025-12-11

Change the model deployment capacity:

azd env set AI_MODEL_CAPACITY 1

Customizing Resource Names

By default, this template uses a naming convention with unique strings to prevent naming collisions. The following environment variables control resource naming via the main.bicepparam file:

Variable Description
AZURE_ENV_NAME Environment name prefix (default: azd-foundry-solution)
AZURE_RESOURCE_GROUP Resource group name
AZURE_LOCATION Azure region (default: eastus2)

To override, run azd env set <key> <value> before running azd up.

Enabling and Disabling Resources

User Role Assignment

By default, the deployment creates role assignments for the current user. To disable:

azd env set CREATE_ROLE_FOR_USER false

AI Foundry Project

To skip provisioning the AI Foundry project, hub, and model deployment:

azd env set CREATE_FOUNDRY_PROJECT false

Environment Variable Reference

All configurable parameters with their defaults:

Variable Default Description
AZURE_ENV_NAME azd-foundry-solution Environment name
AZURE_RESOURCE_GROUP (empty - auto-generated) Resource group name
AZURE_LOCATION eastus2 Azure region
KUBERNETES_VERSION 1.32 AKS Kubernetes version
SYSTEM_VM_SIZE (selected interactively) System node pool VM
WORKLOAD_VM_SIZE (selected interactively) Workload node pool VM
DEEPSTREAM_GPU_VM_SIZE (selected interactively) DeepStream GPU node pool VM
INFERENCE_GPU_VM_SIZE (selected interactively) Inference GPU node pool VM
DEEPSTREAM_GPU_MAX_NODE_COUNT 1 Max nodes in DeepStream pool
CREATE_FOUNDRY_PROJECT true Deploy AI Foundry
AI_MODEL_NAME gpt-5.2 AI model name
AI_MODEL_VERSION 2025-12-11 AI model version
AI_MODEL_CAPACITY 1 Model deployment capacity
CREATE_ROLE_FOR_USER true Create RBAC role assignments
MEDIA_STREAMER_ENABLED true Enable media streamer (RTSP camera + sample agent job in post-provision)
STORAGE_SKU_NAME Standard_LRS Storage account SKU (Standard_LRS, Standard_ZRS, Standard_GRS). ZRS availability is checked during preprovision.