Architecture

NATL separates what to test (YAML scenarios) from how the browser is driven (adapters). Typical UI flows stay portable when you change engine:.

Packages

flowchart LR YAML["YAML scenarios"] --> CLI["@natl/cli"] CLI --> Core["@natl/core"] Core --> Adapters["UI adapters"] Core --> HTTP["built-in http"] Adapters --> PW["Playwright"] Adapters --> SE["Selenium"] Adapters --> CY["Cypress"]

@natl/cli — finds files, loads config, filters tags, loads the adapter
@natl/core — parse, interpolate, interpret; no browser dependency
Adapters — Playwright / Selenium / Cypress as separate packages
http — built into core for API-only or with: http blocks

Run path

flowchart TB Run["natl run"] --> Config["Config + YAML"] Config --> Parse["Parse + interpolate"] Parse --> Interp["Interpreter"] Interp --> Choice{"engine"} Choice -->|"playwright / selenium / cypress"| UI["UI adapter"] Choice -->|"http"| Fetch["HTTP fetch"]

Browsers and driver details stay in the adapter. Core passes opaque options like browser and viewport from natl.config.yaml. More design rules: Canon.