praman-cli

Praman CLI — SAP UI5 Testing Patterns

Token-efficient CLI alternative to MCP. Bridge object: window.__praman_bridge.

Bridge Readiness Check

Always verify the bridge is ready before interacting with controls:

npx playwright-cli run-code -s=sap "async page => {
  const ready = await page.evaluate(() => window.__praman_bridge?.ready);
  return ready;
}"

Never proceed if ready is not true. The bridge is injected via browser.initScript in .playwright/praman-cli.config.json.

Control Discovery via run-code

Use getById() to locate UI5 controls. Always return values — console.log() output is invisible to the agent.

npx playwright-cli run-code -s=sap "async page => {
  const ctrl = await page.evaluate(() =>
    window.__praman_bridge.getById('container-app---main--myInput')
  );
  return ctrl;
}"

WARNING: console.log() Is Invisible

console.log() inside run-code produces NO output for the agent. Always use return to surface results.

// WRONG — agent sees nothing
async page => { console.log(await page.evaluate(() => ...)); }

// CORRECT — agent receives the value
async page => { return await page.evaluate(() => ...); }

setValue + fireChange + waitForUI5 Pattern

Every input interaction requires all three steps — never skip fireChange() or waitForUI5():

npx playwright-cli run-code -s=sap "async page => {
  await page.evaluate(() => {
    const ctrl = window.__praman_bridge.getById('container-app---main--inputField');
    ctrl.setValue('test value');
    ctrl.fireChange({ value: 'test value' });
  });
  await page.evaluate(() => window.__praman_bridge.waitForUI5());
  return 'input set';
}"

Snapshot: Always Use --filename

Agents MUST use --filename so the snapshot is written to a deterministic path they can read back:

npx playwright-cli snapshot -s=sap --filename=current-state.txt

Without --filename, the snapshot goes to stdout and may be truncated or lost in large outputs.