WatchMyWork — Demonstrate Once, Test Everywhere

WatchMyWork is a demonstration-driven developer testing platform designed to reduce the repetitive effort involved in writing and maintaining browser automation. Instead of asking a developer to manually describe every step or write Playwright code from scratch, WatchMyWork allows them to demonstrate one browser test. A custom Chrome extension records semantic actions such as filling fields, clicking buttons, and reading results. Nemotron then generalizes that demonstration into a reusable workflow, which the developer reviews and confirms before execution. Once confirmed, WatchMyWork generates data-driven Playwright automation and runs the workflow across multiple spreadsheet-based test cases. Each case compares the expected result with the actual result and is automatically classified as PASS, FAIL, or ERROR. AI is used during workflow understanding rather than once per test row. After confirmation, execution is deterministic through Playwright. A semantic fallback compiler also keeps the workflow usable if the external model is temporarily unavailable. When a test fails, WatchMyWork creates a downloadable Bob Debug Package containing the generated test, failed cases, workflow JSON, failure summary, and relevant debugging context. The developer can open this package in IBM Bob IDE to investigate the failure and work toward the smallest safe fix. The system uses React and TypeScript for the frontend, FastAPI for the backend, a custom Manifest V3 Chrome extension for semantic recording, Playwright for execution, SQLite for state and checkpoints, Nemotron through OpenRouter for workflow inference, and Railway for deployment. The main value of WatchMyWork is simple: developers already know how to perform many repetitive tests. Instead of forcing them to re-describe or re-code those workflows, WatchMyWork lets them show the process once and turns that demonstration into reusable developer automation.