Package org.egothor.methodatlas.ai

Class AzureOpenAiClient

java.lang.Object

org.egothor.methodatlas.ai.AzureOpenAiClient

All Implemented Interfaces:

AiProviderClient

public final class AzureOpenAiClient
extends Object
implements AiProviderClient

AiProviderClient implementation for Azure OpenAI Service deployments.

Azure OpenAI exposes a chat completions API that is structurally similar to the public OpenAI API but differs in three important ways:

• Endpoint structure — the deployment name is embedded in the path rather than supplied as a JSON field: {baseUrl}/openai/deployments/{deployment}/chat/completions?api-version={version}
• Authentication header — requests carry an api-key header instead of the standard Authorization: Bearer form used by the public OpenAI API
• Model identifier — AiOptions.modelName() is interpreted as the Azure deployment name, not the underlying model family name; the deployment name is chosen when the resource is configured in the Azure portal

These differences are fully encapsulated within this class. The request and response JSON structures are identical to those used by OpenAiCompatibleClient, allowing the same prompt builder and response normalization logic to be reused.

Data Residency

Requests are sent to a resource endpoint within the organization's own Azure tenant. Data does not leave the tenant boundary, making this provider suitable for regulated environments where source code must not be transmitted to third-party cloud services.

Operational Responsibilities

• constructing the Azure-specific deployment endpoint URL
• injecting the api-key authentication header
• constructing and submitting chat completion requests
• extracting JSON content from the model response
• normalizing the result into AiClassSuggestion

Instances are typically created through AiProviderFactory.create(AiOptions).

See Also:

• AiProvider.AZURE_OPENAI
• AiProviderClient
• AiProviderFactory
• OpenAiCompatibleClient

Constructor Summary

Constructors

Constructor	Description
AzureOpenAiClient(AiOptions options)	Creates a new Azure OpenAI client with no rate-limit notification.
AzureOpenAiClient(AiOptions options, RateLimitListener rateLimitListener)	Creates a new Azure OpenAI client that notifies rateLimitListener before each rate-limit sleep.

Method Summary

Modifier and Type	Method	Description
boolean	isAvailable()	Determines whether this client can be used in the current runtime environment.
AiClassSuggestion	suggestForClass(String fqcn, String classSource, String taxonomyText, List<PromptBuilder.TargetMethod> targetMethods)	Submits a classification request to the configured Azure OpenAI deployment.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

AzureOpenAiClient

public AzureOpenAiClient(AiOptions options)

Creates a new Azure OpenAI client with no rate-limit notification.

Rate-limit pauses are handled transparently. Use AzureOpenAiClient(AiOptions, RateLimitListener) when callers need to be notified of such pauses.

The supplied configuration must provide:

• AiOptions.baseUrl() — resource endpoint, e.g. https://contoso.openai.azure.com
• AiOptions.modelName() — deployment name as configured in the Azure portal
• AiOptions.apiVersion() — REST API version, e.g. 2024-02-01
• AiOptions.resolvedApiKey() — resource-scoped API key

Parameters:

options - AI runtime configuration

AzureOpenAiClient

public AzureOpenAiClient(AiOptions options,
 RateLimitListener rateLimitListener)

Creates a new Azure OpenAI client that notifies rateLimitListener before each rate-limit sleep.

Parameters:

options - AI runtime configuration

rateLimitListener - callback invoked before each HTTP 429 pause; must not be null

See Also:

• RateLimitListener

Method Details

isAvailable

public boolean isAvailable()

Determines whether this client can be used in the current runtime environment.

Availability requires a non-blank API key resolved through AiOptions.resolvedApiKey().

Specified by:

isAvailable in interface AiProviderClient

Returns:

true if a usable API key is available

suggestForClass

public AiClassSuggestion suggestForClass(String fqcn,
 String classSource,
 String taxonomyText,
 List<PromptBuilder.TargetMethod> targetMethods)
                                  throws AiSuggestionException

Submits a classification request to the configured Azure OpenAI deployment.

The request is sent to the deployment-specific endpoint:

 {baseUrl}/openai/deployments/{modelName}/chat/completions?api-version={apiVersion}
 

The request payload includes:

• the deployment name as the model field
• a system prompt defining classification rules
• a user prompt containing the test class source and taxonomy
• a deterministic temperature setting of 0.0

Authentication uses the api-key HTTP header carrying the value returned by AiOptions.resolvedApiKey().

Specified by:

suggestForClass in interface AiProviderClient

Parameters:

fqcn - fully qualified class name being analyzed

classSource - complete source code of the class

taxonomyText - taxonomy definition guiding classification

targetMethods - deterministically extracted JUnit test methods that must be classified

Returns:

normalized classification result

Throws:

AiSuggestionException - if the provider request fails, the model response is invalid, or JSON deserialization fails

See Also:

• AiClassSuggestion
• AiMethodSuggestion