AI Engine Overview

AI Diagnosis Process

The AI Engine supports two execution modes. Choose based on your use case:

Standard Analysis

Executes the full 6-step pipeline and returns a structured JSON response. Use this when you need complete, parsed analysis results.

@RestController
@RequestMapping("/api/my-analysis")
@RequiredArgsConstructor
public class MyAnalysisController {

    private final AICoreOperations<MyContext> aiProcessor;
    private final StandardStreamingService streamingService;

    @PostMapping("/analyze")
    public Mono<ResponseEntity<MyAnalysisResponse>> analyze(
            @RequestBody MyAnalysisItem request) {

        MyAnalysisRequest aiRequest = createRequest(request);

        return streamingService.process(
                aiRequest, aiProcessor, MyAnalysisResponse.class
        ).map(ResponseEntity::ok);
    }
}

Streaming Analysis

Runs context retrieval, preprocessing, and prompt generation, then streams the LLM response in real-time via Server-Sent Events. Use this for chat-like interfaces or when users need immediate feedback.

@PostMapping(value = "/analyze/stream",
             produces = MediaType.TEXT_EVENT_STREAM_VALUE)
public Flux<ServerSentEvent<String>> analyzeStream(
        @RequestBody MyAnalysisItem request) {

    MyAnalysisRequest aiRequest = createRequest(request);

    return streamingService.stream(aiRequest, aiProcessor);
}

Customization Points

To build a custom AI feature, you implement these extension points. Each one plugs into the AI Engine at a specific layer:

Platform Integration

The AI Engine integrates with Contexa's Identity and IAM modules to deliver AI Native Security. Each module contributes a distinct capability:

When integrated together, the platform enables:

• Zero Trust security analysis — AI-powered risk analysis can feed runtime security decisions when the core engine is integrated with Identity and IAM.
• Policy generation workflows — the IAM module provides AI-assisted policy generation templates and retrieval paths for structured XACML authoring.
• Resource naming recommendations — the IAM module includes AI-assisted resource naming templates to help normalize business-facing resource labels.
• Shared prompt and retrieval infrastructure — the same pipeline, model selection, advisor, and RAG primitives can be reused across security and non-security AI features.

Configuration Overview

AI Engine behavior is controlled through application.yml properties. Key configuration areas:

Related