Package org.egothor.methodatlas.ai

package org.egothor.methodatlas.ai

AI integration layer for MethodAtlas providing automated security classification of JUnit test methods.

This package contains the infrastructure required to obtain AI-assisted suggestions for security tagging of JUnit 5 tests. The subsystem analyzes complete test classes, submits classification prompts to an AI provider, and converts the returned results into structured suggestions that can be consumed by the main application.

Architecture Overview

The AI subsystem follows a layered design:

• Engine layer – AiSuggestionEngine orchestrates provider communication and taxonomy handling.
• Provider layer – implementations of AiProviderClient integrate with specific AI services such as Ollama, OpenAI-compatible APIs, or Anthropic.
• Prompt construction – PromptBuilder builds the prompt that instructs the model how to perform security classification.
• Taxonomy definition – DefaultSecurityTaxonomy and OptimizedSecurityTaxonomy define the controlled vocabulary used for tagging.
• Result normalization – AI responses are converted into the structured domain model (AiClassSuggestion and AiMethodSuggestion).

Security Considerations

Source code analyzed by the AI subsystem may contain sensitive information. For environments where external transmission of code is undesirable, the subsystem supports local inference through OllamaClient.

Deterministic Output

The subsystem is designed to obtain deterministic, machine-readable output from AI models. Prompts enforce strict JSON responses and classification decisions are constrained by a controlled taxonomy.

Extensibility

Additional AI providers can be integrated by implementing AiProviderClient and registering the implementation in AiProviderFactory.

Since:

1.0.1

Related Packages

Package	Description
org.egothor.methodatlas	Provides the core command-line utility for analyzing Java test sources and producing structured metadata about JUnit test methods.
org.egothor.methodatlas.command	CLI command handler implementations for MethodAtlas.
org.egothor.methodatlas.emit	Output emitters for MethodAtlas scan results.

Class	Description
AiClassSuggestion	Immutable AI-generated classification result for a single parsed test class.
AiMethodSuggestion	Immutable AI-generated security classification result for a single test method.
AiOptions	Immutable configuration describing how AI-based enrichment should be performed during a MethodAtlasApp execution.
AiOptions.Builder	Mutable builder used to construct validated AiOptions instances.
AiOptions.TaxonomyMode	Built-in taxonomy modes used for security classification.
AiProvider	Enumeration of supported AI provider implementations used by the AiSuggestionEngine.
AiProviderClient	Provider-specific client abstraction used to communicate with external AI inference services.
AiProviderFactory	Factory responsible for creating provider-specific AI client implementations.
AiSuggestionEngine	High-level AI orchestration contract for security classification of parsed test classes.
AiSuggestionEngineImpl	Default implementation of AiSuggestionEngine that coordinates provider selection and taxonomy loading for AI-based security classification.
AiSuggestionException	Checked exception indicating failure during AI-based suggestion generation or related AI subsystem operations.
AnthropicClient	AiProviderClient implementation for the Anthropic API.
AzureOpenAiClient	AiProviderClient implementation for Azure OpenAI Service deployments.
DefaultSecurityTaxonomy	Provides the default built-in taxonomy used to guide AI-based security classification of JUnit test methods.
HttpSupport	Small HTTP utility component used by AI provider clients for outbound network communication and JSON processing support.
JsonText	Utility methods for extracting JSON fragments from free-form text produced by AI model responses.
ManualConsumeEngine	Handles the consume phase of the manual AI workflow.
ManualPrepareEngine	Handles the prepare phase of the manual AI workflow.
OllamaClient	AiProviderClient implementation for a locally running Ollama inference service.
OpenAiCompatibleClient	AiProviderClient implementation for AI providers that expose an OpenAI-compatible chat completion API.
OptimizedSecurityTaxonomy	Provides the optimized built-in taxonomy used to guide AI-based security classification when prompt compactness and model reliability are prioritized.
PromptBuilder	Utility responsible for constructing the prompt supplied to AI providers for security classification of JUnit test classes.
PromptBuilder.TargetMethod	Deterministically extracted test method descriptor supplied to the prompt.
RateLimitListener	Callback invoked by HttpSupport immediately before each rate-limit pause caused by an HTTP 429 response from an AI provider.
SuggestionLookup	Immutable lookup structure providing efficient access to AI-generated method suggestions by method name.