AUTO QA

An agent-driven end-to-end QA system built for Sistava that validates real customer journeys in a browser, then verifies the outcome through live logs, database reads, and APIs. Instead of relying on brittle, element-by-element scripts, each capability is defined by a shared plain-English specification with acceptance criteria and focused scenarios for individual paths.

Every run creates a durable report with a clear verdict and the evidence behind it: screenshots, trace IDs, logs, and the relevant system state. A local dashboard makes the work visible by showing current test status, run history, and specifications that still lack coverage.

The framework combines capability specifications, focused QA cases, and legacy scripted browser drivers. It defaults to local execution with fresh accounts for state-sensitive journeys, uses the paid product configuration for AI-behavior testing, blocks production runs unless explicitly enabled, and requires approval before expensive runs.

When the system confirms a regression, a separate fixer can make a local-only repair. A fresh QA run verifies the result afterward, keeping implementation and validation independent.