Feature Implementation
This section contains weekly summaries of feature, task, and bug work completed by the engineering team on the GTP-Playwright repository - the core test automation platform used to record, generate, and execute browser-based tests.
All feature updates listed under this folder refer specifically to work done in the GTP-Playwright repository, not the broader GoTestPro product suite.
Repository Overview
GTP-Playwright is structured as a monorepo with five components. Three form the core record-and-run pipeline, and two are separate, independently deployed/vendored modules:
Component | What it is | Role |
|---|---|---|
gtp_recorder | Chrome browser extension | Lets a user record their actions in the browser (clicks, inputs, screenshots, etc.) to build a test case without writing code. |
execution_engine | Node.js cloud server | Runs recorded tests at scale in the cloud, across different browsers and devices (including third-party cloud grids like LambdaTest), and reports results back. |
shared_module | Shared Playwright execution logic | Common library used by both the recorder and the execution engine, so that a test behaves the same way whether it's being previewed locally or run in the cloud. |
agent_engine | Standalone Node.js/Express service, orchestrated with LangGraph.js | Acts as the "brain" for GoTestPro's autonomous browsing-agent mode. The recorder extension observes the page and executes actions (the "hands and eyes"); agent_engine decides what action to take next at each step of a run, with progress checkpointed in Redis so a run can be tracked, resumed, or cancelled. Deployed independently of the core pipeline. |
mcp | Internal vendored copy of the Playwright browser automation engine | Provides the underlying cross-browser (Chromium, Firefox, WebKit) automation capabilities - auto-waiting, resilient element interaction, and tracing/screenshots - used internally by the platform's execution layer. Maintained/updated on its own cadence, separate from core pipeline releases. |
Why this matters
Because the recorder and the execution engine share the same underlying logic (shared_module), a fix or improvement made in one place - such as more reliable element detection or consistent browser sizing - automatically benefits both recording and cloud execution. This is why several of the weekly updates below touch multiple parts of the system at once. agent_engine and mcp are tracked in the same weekly updates but called out separately when relevant, since they are independently deployed/vendored components with their own release cadence.
Weekly Updates
The pages nested under this one track feature, task, and bug-fix activity across all five components on a week-by-week basis.