Azure AI Projects Python SDK: Build AI Applications with Foundry

What This Skill Does

Build AI applications using the Azure AI Projects Python SDK (azure-ai-projects). Use when working with Foundry project clients, creating versioned agents with PromptAgentDefinition, running evaluations, managing connections/deployments/datasets/indexes, or using OpenAI-compatible clients. This is the high-level Foundry SDK - for low-level agent operations, use azure-ai-agents-python skill.

When to Use It

This skill is designed for Python developers working with Azure cloud services. Reach for it when you need to:

• Integrate Azure Ai Projects capabilities into your application
• Follow SDK best practices for authentication, error handling, and resource management
• Understand the correct API patterns and client initialization

Key Capabilities

Authentication

Azure SDK skills use DefaultAzureCredential for flexible authentication that works across development and production:

# Supports managed identity, CLI credentials, environment variables
# No hardcoded keys needed

Core Operations

The skill covers the primary operations for this service:

Client initialization — Proper setup with credential providers and configuration

Resource operations — Create, read, update, and delete operations specific to this service

Error handling — Azure-specific exception patterns and retry strategies

Async support — Both synchronous and asynchronous client patterns where available

Best Practices

• Use managed identity in production — Avoid storing credentials in code or config files
• Handle throttling gracefully — Azure services have rate limits; implement exponential backoff
• Log operations — Enable SDK logging for debugging without exposing sensitive data
• Pin SDK versions — Use specific versions in production to avoid breaking changes

Common Patterns

Resource Lifecycle

Most Azure SDK operations follow a consistent pattern:

Create a credential object

Initialize the service client with the credential and endpoint

Perform operations through the client

Handle responses and errors consistently

Configuration

Keep service endpoints, resource names, and other configuration in environment variables or Azure App Configuration rather than hardcoding them.

When NOT to Use

• Management plane operations — Use the Azure Resource Manager SDK for provisioning and lifecycle management
• Multi-cloud deployments — Consider cloud-agnostic abstractions if you need portability
• Simple HTTP calls — If you only need one API call, the REST API might be simpler than the full SDK

Related Resources

• Azure SDK for Python documentation
• Azure Ai Projects documentation