"Code is liable to change. Tests should be resilient to that change."
Sentinel is a Next-Generation Quality Assurance Platform that focuses on Agentic Automation.
This repository serves as the public documentation and architectural overview for the Sentinel engine, which uses Large Language Models (LLMs) to understand test intent, generate dynamic code, and self-heal when the UI changes.
Because Sentinel relies on proprietary machine-learning integrations and secure, ephemeral Docker execution, the core automation engine is kept private.
However, you can explore the live Mission Control Dashboard, watch the AI generated locators in real-time, and read the entire engineering breakdown on my official portfolio:
While the codebase is closed-source, the framework operates on the following principles:
Standard tests break when developers change IDs or CSS Classes. Sentinel intercepts NoSuchElementExceptions, captures the "Deep DOM" (including Shadow DOMs), and queries the Gemini 2.0 Flash multimodal model to dynamically identify the new optimal selector, pausing and resuming execution autonomously.
Sentinel bridges the gap between manual QA and automation engineers by allowing users to compose tests in plain English. The "Cortex" engine translates these requirements into executable BDD feature files on the fly.
A custom-built Next.js 14 dashboard streams live test telemetry, network traffic, and VNC "ghost views" of the headless browsers via WebSockets, allowing teams to watch the AI heal tests in real-time.
- Frontend Observability: Next.js 14, React, TailwindCSS, WebSockets
- Execution Engine: Java 17, Selenium WebDriver 4, Cucumber BDD
- AI Core: Google Gemini 2.0 Flash (Multimodal Vision API)
- Infrastructure: Docker Containerization, Google Kubernetes Engine (GKE), Google Cloud Run
For business inquiries, speaking engagements, or engineering consultations regarding Agentic AI Automation, please connect with Safwan Saba via the official portfolio.