Voice-Driven Test & Debugging Swarm

Resolving UI and API regressions requires immense context switching. When a CI/CD pipeline or local build fails, developers must manually parse extensive RestAssured or Selenium execution logs, trace the failure back through backend Spring Boot logic, and cross-reference database schemas just to begin writing a fix. This manual debugging and root-cause analysis cycle drastically inflates Mean Time to Resolution (MTTR) and reduces developer productivity. The Voice-Driven Test & Debugging Swarm eliminates this context switching by unifying automated testing and AI-powered remediation into a single voice-activated workflow across four phases: Phase 1 — Target App: A Spring Boot 3.2 application with a deliberately broken CheckoutController (flat subtraction instead of percentage-based discount math) and a disabled HTML button. Two hybrid tests — a RestAssured API test and a Selenium UI test — are configured to fail intentionally and generate Maven Surefire XML reports. Phase 2 — Voice Orchestrator: A Python agent recognises the spoken command "Run API regression tests" and triggers the CI/CD pipeline. Due to Python 3.14 MSVC constraints, the microphone trigger is simulated for this prototype, demonstrating the full downstream automation pipeline. Phase 3 — Connector: The test runner executes mvn clean test using absolute paths, then parses the generated target/surefire-reports/*.xml files to extract the failure element, stack trace, and expected vs actual results. Phase 4 — IBM Bob Handoff and Audio: The extracted failure context is passed to the IBM Bob CLI. Bob analyzes the Spring Boot repository, identifies the broken math logic in CheckoutController.java, and returns a two-sentence root-cause summary. The ElevenLabs v2 SDK then converts this summary to speech using the Sarah voice and plays it aloud to the developer, closing the feedback loop without any keyboard interaction. The result is a workflow where a developer speaks one command, watches the pipeline run autonomously, and hears the fix explained out loud — reducing a multi-hour debugging cycle to under 60 seconds.