Azure API Management for AI Agents

Learn how to productionize Azure AI Foundry endpoints with API management, deploying OpenAI models across regions, implementing token-based pricing, rate limits, circuit breakers, and semantic caching.

Explore the full API lifecycle with Azure API management, including design, develop, publish, secure, monitor, and monetize APIs through a gateway, policies, and a developer portal.

Explore the design phase of the API lifecycle in Azure API Management, comparing design-first and code-first approaches, and learn to import OpenAPI definitions to a robust, tested API surface.

Use the public GitHub repository for the course labs, including the generative ai with api management content, with Jupyter notebooks and Markdown guides.

Deploy an Azure API Management instance, configure the resource group and East US region, select the developer pricing tier, and enable log analytics and application insights for API monitoring.

Explore how Azure API Management groups APIs into products in an API store, assigns subscription keys, and monetizes through tiered plans, with policies at global, product, API, and operation levels.

Build an http api from scratch in Azure API Management, define get people and get by id operations using a code-first approach, and test across starter and unlimited products.

Import a remotely hosted OpenAPI specification into Azure API Management to create and test Pet Store API. Apply a design-first approach to define paths and routing for LLMs function calling.

Secure API lifecycle with Azure API Management by protecting developer portal, gateway, and publisher access; use OAuth/OpenID Connect, social identity providers, RBAC, and private networking in a virtual network.

Explore the publish phase of the api lifecycle in azure api management, using the developer portal for self onboarding, discovery, testing, and subscription keys.

Publish the developer portal, enable cors, and sign up as a user to test APIs in console. Activate starter and unlimited subscriptions and run Star Wars and Pet Store APIs.

Configure rate limiting and caching policies in Azure API Management by creating a colors API from an OpenAPI spec, testing caching, and enforcing three calls per minute.

Explore how Azure API Management scales with multi-region deployments to achieve high availability and fault tolerance by replicating API backends and using Traffic Manager for health-based routing.

Monitor and analyze API lifecycle in Azure API Management, tracking requests and IP origins. Use Application Insights and log analytics to build dashboards showing usage, performance, throttling, and status codes.

Connect an application insights resource to an Azure API management instance, fetch logs from Log Analytics, and build dashboards showing API calls, status codes, regions, and response times.

Deploy an Azure OpenAI resource with a GPT-4o model and integrate it with Azure API Management, then monitor calls via Log Analytics and Application Insights.

Lab demonstrates calling the Azure OpenAI API from Python notebooks using API management, configuring environment variables, making HTTP requests and using the Azure SDK, plus image analysis with GPT-4.

Learn to build an http api from scratch in Azure api management by creating a custom backend with a runtime url that forwards requests to Azure OpenAI GPT-4.

Learn how to replace hard-coded API keys with named values in Azure API Management to keep credentials secure and centrally updated, reflecting changes across backends and policies.

Use a system assigned managed identity in API management to authenticate Azure OpenAI, assign the Cognitive Services OpenAI user role, and fetch tokens for the authorization header.

Explore the Azure AI Foundry landscape, comparing Azure OpenAI Studio and Azure Studio Hub, and learn how they enable testing, collaboration, AI agents, model catalog, and multimodal apps.

Deploy an Azure AI Foundry resource and stage GPT-4 and Deep Sea AR1 model, then layer API management on top to orchestrate multi-vendor LLMs via a single endpoint.

Learn to build an Azure AI Foundry API with Azure APIM, configure model deployments (GPT-4 and Deep Sea Carbon), and secure calls with a system-assigned managed identity.

Test Azure AI Foundry api calls from Python notebooks by configuring environment variables, hitting the GPT four and deep sea carbon models through api management, and observing rate limits.

Explore semantic caching with Azure OpenAI and Azure managed Redis to accelerate LLM responses by using vector embeddings, semantic search, and the text embedding A002 engine.

Explain how vector embeddings create high-dimensional semantic representations for retrieval augmented generation, using Ada 002’s 1536-d vectors and storage options like Azure Cognitive Search, Cosmos DB, and Redis vector similarity.

Learn to generate vector embeddings with Python using an Azure OpenAI text embedding model, deploying and calling embeddings.create. See how 1536-dimension embeddings capture semantic meaning for indexing and rag architecture.

Set up an Azure managed Redis with ready search and a text embedding A002 engine to power semantic caching for Azure OpenAI GPT four, delivering cache hits via vector similarity.

Learn to protect AI applications with Azure AI Content Safety Studio and Azure API Management, applying prompt shields, groundedness checks, and content safety policies to secure LLM workflows.

Deploy and test Azure Content Safety Studio within the AI Foundry and API management stack, exploring text, image, and multimodal moderation APIs, thresholds, and a code-first workflow.

Learn to call Azure Content Safety Studio APIs from Python, insert a safety check between user input and the GPT-4 engine, and validate hate, violence, self-harm, and sexual content.

Integrate azure content safety resource with azure api management gateway to enforce content moderation on calls to ai foundry models, testing with gpt-4 and deep sea carbon model.

Demonstrate content safety filters in a Python notebook, routing requests via API management to Azure AI Foundry with GPT-4 and Deep Sea Carbon, and observe policy blocks.

Master circuit breaker and load balancing patterns to build fault-tolerant, highly available AI systems across multi-region deployments using Azure AI Foundry and API Management.

Deploy two Azure OpenAI GPT-4 instances in Sweden Central and East US, configure per-minute token rate limits, and enable cross-region load balancing with a circuit-breaker design in this hands-on lab.

Create and consolidate two backends for multi-region Azure OpenAI deployments into a single load balanced backend pool. Configure circuit breaker rules and api management policy-based access to the backends.

Explore multi-region load balancing with azure api management and a circuit breaker pattern. Use http calls and the openai python sdk, and configure environment variables.