customize

Guide Azure OpenAI model deployment with SKU, capacity, and RAI policy configuration.

Updated Jul 3, 2025

Search Tags:#azure #openai #model deployment #capacity planning #sku selection #rai policy #quota fallback

Ready to run

npx skills add https://github.com/isakawar/crmKvitkovaPovnya --skill customize-isakawar

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: customize
Source: https://github.com/isakawar/crmKvitkovaPovnya/tree/main/.agents/skills/microsoft-foundry/models/deploy-model/customize
Command: npx skills add https://github.com/isakawar/crmKvitkovaPovnya --skill customize-isakawar

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

It solves the problem of deploying Azure OpenAI models without losing control over which exact model version, capacity, SKU, and content filtering settings are used.

Core Features & Use Cases

Interactive deployment flow: Guides you through selecting model name, version, SKU (GlobalStandard/Standard/ProvisionedManaged/DataZoneStandard), and deployment name with validation.
Capacity planning with guardrails: Helps you choose an exact capacity within min/max/step limits and can fall back across regions if capacity isn’t available.
Operational configuration options: Supports RAI policy selection for OpenAI models, plus advanced options like dynamic quota, priority processing, and spillover.

Quick Start

Tell the assistant: "Customize an Azure OpenAI deployment for gpt-4o by choosing the best available version, GlobalStandard SKU, set a capacity for my expected load, select Microsoft.DefaultV2 RAI policy, and then deploy and show me the endpoint."

Frequently Asked Questions about customize

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I deploy a specific Azure OpenAI model version with exact capacity and SKU settings?▼

To deploy an Azure OpenAI model with exact settings, you need an interactive deployment flow that guides model and version selection, validates SKU choices like GlobalStandard, and enforces min/max capacity limits. This ensures you provision the precise TPM or PTU capacity required for your expected load.

Can I configure RAI policy and advanced deployment options when provisioning Azure OpenAI models?▼

Yes, configuring RAI policy and advanced deployment options is supported during Azure OpenAI model provisioning. You can select content filtering policies like Microsoft.DefaultV2 and enable operational features including dynamic quota, priority processing, and spillover to control model behavior.

What happens if my requested TPM or PTU capacity is not available in my Azure region?▼

When requested TPM or PTU capacity is unavailable in your Azure region, the deployment process can automatically handle quota constraints using cross-region fallback. This capacity planning guardrail queries live SKU availability and falls back across regions to ensure successful provisioning.

Which Azure OpenAI deployment SKUs are supported for custom capacity planning?▼

Supported Azure OpenAI deployment SKUs for custom capacity planning include GlobalStandard, Standard, ProvisionedManaged, and DataZoneStandard. Choosing the appropriate SKU determines your capacity model, whether you are provisioning TPM for shared throughput or PTU for dedicated managed capacity.

How do I automate Azure OpenAI deployment using Azure CLI and ARM REST when quota is constrained?▼

You can automate Azure OpenAI deployment under quota constraints by using Azure CLI and ARM REST to create deployments with cross-region fallback. This involves validating inputs, querying live SKU availability, and monitoring the provisioning state until completion or failure.

Does Azure OpenAI deployment enforce deployment name rules and input validation?▼

Yes, Azure OpenAI deployment strictly enforces deployment name rules and input validation before submitting any provisioning request. This ensures the specified model, version, SKU, and capacity parameters are formatted correctly and meet Azure requirements to prevent deployment failures.