@aarwitz/tapp 0.17.0-rc.1 → 0.17.0-rc.10

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/AGENTS.md CHANGED
@@ -21,6 +21,15 @@ npx -y @aarwitz/tapp explore app.apk --platform android --app-id com.acme.app
21
21
  npx -y @aarwitz/tapp flow run .tapp/flows/smoke.yml # committed, keyless E2E replay
22
22
  ```
23
23
 
24
+ If repository onboarding detects multiple application targets, target detection is deterministic but
25
+ the choice is the user's. In a human TTY, Tapp displays a numbered selector and continues in the same
26
+ command. A non-interactive CLI prints the exact choices and exits before building. MCP returns
27
+ `reason: "target-selection-required"` with structured `choices[]` (`platform`, `name`, `sourcePath`,
28
+ `selector`, and exact `command`). **Do not pick one yourself.** Present those choices to the user
29
+ with the client's native multiple-choice question UI when available, then rerun using the selected
30
+ `--platform` and `--target`. Plain chat can list the same choices when the client has no question
31
+ widget.
32
+
24
33
  For focused web evidence, `open` and `tree` accept one semantic interaction plus an async content
25
34
  wait: `tapp open https://example.com --tap "Not now" --wait-for "Dashboard"`. Tapp waits for the
26
35
  page to stabilize before capturing it and warns honestly if the bounded wait ends while it is still
@@ -946,6 +946,7 @@ class ExplorerTests: XCTestCase {
946
946
  var elementScreenPresence: [String: Set<String>] = [:]
947
947
  var totalDistinctStates = 0
948
948
  var recentStateHashes: [String] = []
949
+ var recentScreenTitles: [String] = []
949
950
  var actionsSinceNewState = 0
950
951
  var sameScreenStreak = 0
951
952
  var screenTextEntryCount: [String: Int] = [:]
@@ -1238,6 +1239,15 @@ class ExplorerTests: XCTestCase {
1238
1239
  let initialControlsJson = mapControlsJSON(initialElements)
1239
1240
  print("OCQA_STATE:{\"screen\":\"\(escapeJSON(initialTitle))\",\"hash\":\"\(computeHash(initialElements))\",\"elements\":\(initialElements.count),\"action\":0,\"role\":\"\(escapeJSON(initialRole))\",\"summary\":\"\(escapeJSON(initialSummary))\",\"settled\":\(isScreenSettled() ? "true" : "false"),\"atext\":[\(initialAtext)],\"inputs\":[\(initialInputJson)],\"controls\":[\(initialControlsJson)]}")
1240
1241
 
1242
+ // The launch surface is real evidence even when root normalization immediately dismisses
1243
+ // it. Attach it before the first action so the HTML report's "every screen explored" claim
1244
+ // includes onboarding/login sheets rather than beginning at the post-dismiss destination.
1245
+ let initialScreenshot = app.screenshot()
1246
+ let initialAttachment = XCTAttachment(screenshot: initialScreenshot)
1247
+ initialAttachment.name = "state_0_\(initialTitle.replacingOccurrences(of: " ", with: "_"))"
1248
+ initialAttachment.lifetime = .keepAlways
1249
+ add(initialAttachment)
1250
+
1241
1251
  navigateToRootScreen(actionCount: &actionCount)
1242
1252
 
1243
1253
  // The first state is the true customer launch surface. Directed replay begins after
@@ -1386,8 +1396,10 @@ class ExplorerTests: XCTestCase {
1386
1396
  }
1387
1397
 
1388
1398
  recentStateHashes.append(stateHash)
1399
+ recentScreenTitles.append(titleStr)
1389
1400
  if recentStateHashes.count > 12 {
1390
1401
  recentStateHashes.removeFirst(recentStateHashes.count - 12)
1402
+ recentScreenTitles.removeFirst(recentScreenTitles.count - 12)
1391
1403
  }
1392
1404
 
1393
1405
  if previousStateHash == stateHash {
@@ -1474,10 +1486,19 @@ class ExplorerTests: XCTestCase {
1474
1486
  // is the fuller a11y text inventory that grounds the vision reviewer (see visionTextInventory).
1475
1487
  let escapedTitle = escapeJSON(titleStr)
1476
1488
  let settled = isScreenSettled()
1477
- let atextJson = visionTextInventory(elements).map { "\"\(escapeJSON($0))\"" }.joined(separator: ",")
1489
+ let visibleTextInventory = visionTextInventory(elements)
1490
+ let atextJson = visibleTextInventory.map { "\"\(escapeJSON($0))\"" }.joined(separator: ",")
1478
1491
  let controlsJson = mapControlsJSON(elements)
1479
1492
  print("OCQA_STATE:{\"screen\":\"\(escapedTitle)\",\"hash\":\"\(stateHash)\",\"elements\":\(elements.count),\"action\":\(actionCount),\"role\":\"\(escapeJSON(screenRole))\",\"summary\":\"\(escapeJSON(screenSummary))\",\"settled\":\(settled ? "true" : "false"),\"atext\":[\(atextJson)],\"inputs\":[\(inputJsonArray)],\"controls\":[\(controlsJson)]}")
1480
1493
 
1494
+ // Signing out after a successful login is a completed auth cycle, not a navigation
1495
+ // trap and not lost-form-state. Stop cleanly instead of probing the root login screen
1496
+ // for an impossible back path or expecting credentials to persist after logout.
1497
+ if authSucceeded && detectedInputs.contains(where: { $0.secure }) {
1498
+ print("OCQA_STATE:auth_cycle_complete screen=\(escapedTitle) step=\(actionCount)")
1499
+ break
1500
+ }
1501
+
1481
1502
  // ---- Persistence probe: on a fresh RE-ARRIVAL at a screen, fields we previously
1482
1503
  // typed into (and verified visible in the a11y value) should still hold their value.
1483
1504
  // An empty field here means the entered state was silently lost on navigation —
@@ -1497,6 +1518,8 @@ class ExplorerTests: XCTestCase {
1497
1518
  if current.isEmpty || current == ph {
1498
1519
  reportedPersistenceKeys.insert(memKey)
1499
1520
  let t = "Entered value did not persist: '\(fieldKey)' on \(titleStr)"
1521
+ issues.append((type: "state_persistence", severity: "medium", title: t,
1522
+ desc: "Typed '\(typed)' into this field earlier in the run; after navigating away and returning, the field is empty — entered state was silently lost."))
1500
1523
  print("OCQA_ISSUE:{\"type\":\"state_persistence\",\"severity\":\"medium\",\"title\":\"\(escapeJSON(t))\",\"screen\":\"\(escapedTitle)\",\"control\":\"\(escapeJSON(fieldKey))\",\"step\":\(actionCount),\"desc\":\"Typed '\(escapeJSON(typed))' into this field earlier in the run; after navigating away and returning, the field is empty — entered state was silently lost.\"}")
1501
1524
  }
1502
1525
  }
@@ -1591,14 +1614,14 @@ class ExplorerTests: XCTestCase {
1591
1614
  // content-feed app: post detail + replies-loading spinner flagged app_hang HIGH).
1592
1615
  let visibleTextCount = elements.filter { isStaticTextType($0.type) && normalizeVisibleText($0.label).count >= 3 }.count
1593
1616
  if screenVisitCount[titleStr] ?? 0 <= 1, visibleTextCount <= 4,
1594
- app.activityIndicators.firstMatch.exists || app.progressIndicators.firstMatch.exists {
1617
+ hasIndeterminateLoadingIndicator() {
1595
1618
  let loadingKey = "loading:\(titleStr)"
1596
1619
  if !reportedIssueKeys.contains(loadingKey) {
1597
1620
  var resolved = false
1598
1621
  let deadline = Date().addingTimeInterval(8.0)
1599
1622
  while Date() < deadline {
1600
1623
  Thread.sleep(forTimeInterval: 1.0)
1601
- if !(app.activityIndicators.firstMatch.exists || app.progressIndicators.firstMatch.exists) {
1624
+ if !hasIndeterminateLoadingIndicator() {
1602
1625
  resolved = true
1603
1626
  break
1604
1627
  }
@@ -1658,7 +1681,13 @@ class ExplorerTests: XCTestCase {
1658
1681
 
1659
1682
  // ---- Blank-screen detection ----
1660
1683
  // Distinguish between "no a11y labels / custom UI" vs genuinely empty.
1661
- if elements.count < 5 && interactable.count == 0 {
1684
+ // A system Back/Close affordance is navigation chrome, not screen content. Treat a
1685
+ // destination whose only usable control is that chrome as blank too; otherwise an
1686
+ // EmptyView pushed by NavigationStack looks like a healthy one-control screen.
1687
+ let contentInteractables = interactable.filter {
1688
+ !isNavBackButton($0) && !isLikelyGlobalNavigation($0, screenBounds: screenBounds)
1689
+ }
1690
+ if visibleTextInventory.isEmpty && contentInteractables.isEmpty {
1662
1691
  let blankKey = "blank:\(titleStr)"
1663
1692
  let blankCount = (actionCounts[blankKey] ?? 0) + 1
1664
1693
  actionCounts[blankKey] = blankCount
@@ -1690,16 +1719,26 @@ class ExplorerTests: XCTestCase {
1690
1719
  }
1691
1720
 
1692
1721
  // ---- Navigation-loop detection ----
1693
- // Check if recentStateHashes has a repeating cycle of length 2 or 3
1722
+ // A cycle must actually move across distinct states. Four identical reads satisfy the
1723
+ // arithmetic shape A,A,A,A of the old period-2 check, which mislabeled ordinary
1724
+ // scroll/probe recovery on a stable screen as a navigation loop.
1694
1725
  if recentStateHashes.count >= 6 {
1695
1726
  let recent = recentStateHashes
1727
+ // Distinct structural hashes are not enough: a list and its detail rows can share
1728
+ // one navigation title and alternate A/B while the explorer intentionally samples
1729
+ // different rows. Calling that a navigation loop is a false positive. Require the
1730
+ // cycle to cross distinct user-visible screen titles as well.
1696
1731
  let hasLoop2 = recent.count >= 4 &&
1697
1732
  recent[recent.count - 1] == recent[recent.count - 3] &&
1698
- recent[recent.count - 2] == recent[recent.count - 4]
1733
+ recent[recent.count - 2] == recent[recent.count - 4] &&
1734
+ Set(recent.suffix(2)).count == 2 &&
1735
+ Set(recentScreenTitles.suffix(2)).count == 2
1699
1736
  let hasLoop3 = recent.count >= 6 &&
1700
1737
  recent[recent.count - 1] == recent[recent.count - 4] &&
1701
1738
  recent[recent.count - 2] == recent[recent.count - 5] &&
1702
- recent[recent.count - 3] == recent[recent.count - 6]
1739
+ recent[recent.count - 3] == recent[recent.count - 6] &&
1740
+ Set(recent.suffix(3)).count == 3 &&
1741
+ Set(recentScreenTitles.suffix(3)).count == 3
1703
1742
  if (hasLoop2 || hasLoop3) && !(authSucceeded && detectedInputs.contains { $0.secure }) {
1704
1743
  let loopKey = "nav_loop:\(titleStr)"
1705
1744
  if actionCounts[loopKey] == nil {
@@ -1711,17 +1750,10 @@ class ExplorerTests: XCTestCase {
1711
1750
  }
1712
1751
  }
1713
1752
 
1714
- // ---- Unresponsive-element detection ----
1715
- // Skip when we're merely re-poking a login screen we've already passed (Sign Out → re-login
1716
- // churn) that's an exploration artifact, not a frozen/broken screen.
1717
- if repeatedStateCount >= 5 && !(authSucceeded && detectedInputs.contains { $0.secure }) {
1718
- let unrespKey = "unresponsive:\(titleStr)"
1719
- if actionCounts[unrespKey] == nil {
1720
- issues.append((type: "unresponsive_element", severity: "medium", title: "Unresponsive UI on \(titleStr)", desc: "Actions are not changing app state — possible frozen or broken screen"))
1721
- print("OCQA_ISSUE:{\"type\":\"unresponsive_element\",\"severity\":\"medium\",\"title\":\"Unresponsive UI\",\"screen\":\"\(escapedTitle)\",\"repeated_state_count\":\(repeatedStateCount),\"step\":\(actionCount)}")
1722
- actionCounts[unrespKey] = 1
1723
- }
1724
- }
1753
+ // Do not infer an unresponsive app merely from an unchanged state streak: recovery
1754
+ // gestures (scroll, carousel probe, center probe) are expected to be no-ops on many
1755
+ // healthy screens. Labeled controls have a stronger detector below: a direct tap plus
1756
+ // two delayed, content-signature reads. Hangs have their own time-based detector.
1725
1757
 
1726
1758
  if interactable.count < 3 {
1727
1759
  print("OCQA_STATE:low_interactable screen=\(escapedTitle) total=\(elements.count) interactable=\(interactable.count) global=\(globalNavElements.count) nonGlobal=\(nonGlobalCandidates.count)")
@@ -1757,10 +1789,10 @@ class ExplorerTests: XCTestCase {
1757
1789
  break
1758
1790
  }
1759
1791
 
1760
- let issueTitle = "Dead end: \(titleStr)"
1761
- issues.append((type: "dead_end", severity: "medium", title: issueTitle, desc: "No interactable elements found"))
1762
- print("OCQA_ISSUE:{\"type\":\"dead_end\",\"severity\":\"medium\",\"title\":\"\(escapeJSON(issueTitle))\",\"screen\":\"\(escapedTitle)\",\"step\":\(actionCount)}")
1763
-
1792
+ // Exhausting Tapp's untried candidate pool is not itself a user-visible dead end:
1793
+ // leaf screens commonly have only a working Back control that was already mapped.
1794
+ // Recover first; only the stronger navigation-trap path below emits a finding when
1795
+ // every real back/dismiss route fails.
1764
1796
  // tryGoBack does swipe-down as its last resort (sheet dismiss)
1765
1797
  let preBackTitle = titleStr
1766
1798
  let backWorked = tryGoBack()
@@ -1772,6 +1804,7 @@ class ExplorerTests: XCTestCase {
1772
1804
  print("OCQA_ACTION:{\"type\":\"back\",\"reason\":\"dead_end_escape\",\"from\":\"\(escapedTitle)\",\"to\":\"\(escapeJSON(postTitle))\",\"step\":\(actionCount),\"screen\":\"\(escapedTitle)\",\"narrative\":\"\(escapeJSON(recoveryNarrative("back_dead_end", screen: titleStr, to: postTitle)))\"}")
1773
1805
  continue
1774
1806
  }
1807
+ if actionCount >= maxActions { break }
1775
1808
  // Swipe right (back gesture) as another option
1776
1809
  let swipeStart = app.coordinate(withNormalizedOffset: CGVector(dx: 0.02, dy: 0.5))
1777
1810
  let swipeEnd = app.coordinate(withNormalizedOffset: CGVector(dx: 0.8, dy: 0.5))
@@ -2047,6 +2080,7 @@ class ExplorerTests: XCTestCase {
2047
2080
  }
2048
2081
  // Back didn't change screens — fall through to global nav
2049
2082
  print("OCQA_ACTION:{\"type\":\"back\",\"reason\":\"screen_exhausted_failed\",\"screen\":\"\(escapedTitle)\",\"step\":\(actionCount),\"narrative\":\"\(escapeJSON(recoveryNarrative("back_failed", screen: titleStr)))\"}")
2083
+ if actionCount >= maxActions { break }
2050
2084
  // Stuck on this screen — use global navigation (tab bar) to reach unexplored areas
2051
2085
  let globalNav = interactable
2052
2086
  .filter { isLikelyGlobalNavigation($0, screenBounds: screenBounds) }
@@ -2247,12 +2281,21 @@ class ExplorerTests: XCTestCase {
2247
2281
  // the crash itself goes unreported. app.state is non-throwing even when the app is dead.
2248
2282
  // (Found on a real app: a login submit terminated the app; the run limped on but emitted
2249
2283
  // no crash finding.) Try one relaunch to distinguish a hard crash from a transient exit.
2250
- if app.state != .runningForeground {
2251
- print("OCQA_STATE:app_left_foreground step=\(actionCount) state=\(app.state.rawValue)")
2284
+ let stateAfterAction = app.state
2285
+ if stateAfterAction != .runningForeground {
2286
+ print("OCQA_STATE:app_left_foreground step=\(actionCount) state=\(stateAfterAction.rawValue)")
2287
+ let crashKey = "crash:\(titleStr)|\(key)"
2288
+ // A terminated process is already proof of an in-run crash. Relaunching it may
2289
+ // succeed, but that must not erase the user-visible failure that just happened.
2290
+ if stateAfterAction == .notRunning && !reportedIssueKeys.contains(crashKey) {
2291
+ reportedIssueKeys.insert(crashKey)
2292
+ issues.append((type: "crash", severity: "critical", title: "App crashed after \(actionType) on \(titleStr)",
2293
+ desc: "The app process terminated after \(actionType) '\(targetName)' on '\(titleStr)'."))
2294
+ print("OCQA_ISSUE:{\"type\":\"crash\",\"severity\":\"critical\",\"title\":\"\(escapeJSON("App crashed after \(actionType) on \(titleStr)"))\",\"screen\":\"\(escapedTitle)\",\"control\":\"\(escapedTarget)\",\"step\":\(actionCount)}")
2295
+ }
2252
2296
  app.activate()
2253
2297
  Thread.sleep(forTimeInterval: 3.0)
2254
2298
  if app.state != .runningForeground {
2255
- let crashKey = "crash:\(titleStr)|\(key)"
2256
2299
  if !reportedIssueKeys.contains(crashKey) {
2257
2300
  reportedIssueKeys.insert(crashKey)
2258
2301
  issues.append((type: "crash", severity: "critical", title: "App crashed after \(actionType) on \(titleStr)",
@@ -2356,16 +2399,28 @@ class ExplorerTests: XCTestCase {
2356
2399
 
2357
2400
  // ---- App left foreground / crash detection ----
2358
2401
  // Check both .exists and .state — external links may cause either to fail
2359
- let appInForeground = app.state == .runningForeground
2402
+ let stateAfterDelayedChecks = app.state
2403
+ let appInForeground = stateAfterDelayedChecks == .runningForeground
2360
2404
  if !appInForeground || !app.exists {
2361
- print("OCQA_STATE:app_left_foreground step=\(actionCount) state=\(app.state.rawValue)")
2405
+ print("OCQA_STATE:app_left_foreground step=\(actionCount) state=\(stateAfterDelayedChecks.rawValue)")
2406
+ let crashKey = "crash:\(titleStr)|\(key)"
2407
+ if stateAfterDelayedChecks == .notRunning && !reportedIssueKeys.contains(crashKey) {
2408
+ reportedIssueKeys.insert(crashKey)
2409
+ issues.append((type: "crash", severity: "critical",
2410
+ title: "App crashed after \(actionType) on \(titleStr)",
2411
+ desc: "The app process terminated after: \(actionDesc)"))
2412
+ print("OCQA_ISSUE:{\"type\":\"crash\",\"severity\":\"critical\",\"title\":\"\(escapeJSON("App crashed after \(actionType) on \(titleStr)"))\",\"screen\":\"\(escapedTitle)\",\"control\":\"\(escapedTarget)\",\"step\":\(actionCount)}")
2413
+ }
2362
2414
  app.activate()
2363
2415
  Thread.sleep(forTimeInterval: 3.0)
2364
2416
  if app.state != .runningForeground {
2365
- issues.append((type: "crash", severity: "critical",
2366
- title: "App not recoverable",
2367
- desc: "App left foreground after: \(actionDesc)"))
2368
- print("OCQA_ISSUE:{\"type\":\"crash\",\"severity\":\"critical\",\"title\":\"App not recoverable\",\"action\":\"\(escapedTarget)\",\"step\":\(actionCount)}")
2417
+ if !reportedIssueKeys.contains(crashKey) {
2418
+ reportedIssueKeys.insert(crashKey)
2419
+ issues.append((type: "crash", severity: "critical",
2420
+ title: "App not recoverable",
2421
+ desc: "App left foreground after: \(actionDesc)"))
2422
+ print("OCQA_ISSUE:{\"type\":\"crash\",\"severity\":\"critical\",\"title\":\"App not recoverable\",\"action\":\"\(escapedTarget)\",\"step\":\(actionCount)}")
2423
+ }
2369
2424
  break
2370
2425
  }
2371
2426
  print("OCQA_STATE:app_reactivated step=\(actionCount)")
@@ -3679,6 +3734,20 @@ class ExplorerTests: XCTestCase {
3679
3734
  return kb.exists ? kb.frame : .zero
3680
3735
  }
3681
3736
 
3737
+ /// Activity indicators are inherently indeterminate. `ProgressIndicator`, however, is also the
3738
+ /// XCTest type for legitimate determinate progress bars (loyalty points, upload percentage,
3739
+ /// onboarding completion). Only value-less/loading-valued progress indicators are hang signals.
3740
+ private func hasIndeterminateLoadingIndicator() -> Bool {
3741
+ if app.activityIndicators.allElementsBoundByIndex.contains(where: { $0.exists && $0.frame.width > 0 && $0.frame.height > 0 }) {
3742
+ return true
3743
+ }
3744
+ return app.progressIndicators.allElementsBoundByIndex.contains { indicator in
3745
+ guard indicator.exists, indicator.frame.width > 0, indicator.frame.height > 0 else { return false }
3746
+ let value = (indicator.value as? String ?? "").trimmingCharacters(in: .whitespacesAndNewlines).lowercased()
3747
+ return value.isEmpty || value == "in progress" || value == "loading"
3748
+ }
3749
+ }
3750
+
3682
3751
  /// True when the screen is in a "settled" resting state — no on-screen keyboard and no open
3683
3752
  /// transient overlay (menu / dropdown / popover / sheet / picker wheel). A screenshot taken while
3684
3753
  /// one of these is up is inherently ambiguous to a visual reviewer (the keyboard "covers" the
package/README.md CHANGED
@@ -80,6 +80,12 @@ npx -y @aarwitz/tapp baseline create . --platform web
80
80
  npx -y @aarwitz/tapp ci install .
81
81
  ```
82
82
 
83
+ On a fresh repository containing multiple apps (for example, iOS plus web), bare
84
+ `tapp init . --explore` does not guess from detection order. A human terminal gets a numbered
85
+ selector; a non-interactive CLI prints exact target-selection commands, while MCP also returns
86
+ structured choices. Neither builds or writes before the choice. After you choose one, the model
87
+ retains every detected target and records the choice as the default for the next bare `tapp explore`.
88
+
83
89
  The baseline command writes only after exploration and every selected deterministic suite pass
84
90
  conclusively. It stores `.tapp/baselines/<platform>/<target-id>.json`; the generated workflow
85
91
  uses that exact target identity so two apps on the same platform never share a baseline. `ci
package/bin/tapp.js CHANGED
@@ -28,8 +28,8 @@ const __dirname = path.dirname(fileURLToPath(import.meta.url));
28
28
  const packageRoot = path.resolve(__dirname, "..");
29
29
  const pkg = JSON.parse(fs.readFileSync(path.join(packageRoot, "package.json"), "utf8"));
30
30
 
31
- // Redirect all writable output away from the (possibly read-only) package dir.
32
- // The old environment alias remains a read-only fallback for older integrations.
31
+ // Redirect all writable output away from the (possibly read-only) package dir. TAPP_HOME and
32
+ // ~/.tapp are the only current runtime locations; retired environment/path aliases are ignored.
33
33
  const tappHome = (process.env.TAPP_HOME || path.join(os.homedir(), ".tapp")).trim();
34
34
  process.env.TAPP_HOME = tappHome;
35
35
  // TAPP_HOME is created lazily (just before the switch) so `--help`, `help`, and `version` never
@@ -130,6 +130,27 @@ function repeatedFlagValues(argv, name) {
130
130
  return values;
131
131
  }
132
132
 
133
+ function iosLaunchOptions(flags, argv) {
134
+ const appLaunchArgs = repeatedFlagValues(argv, "launch-arg");
135
+ let appLaunchEnv;
136
+ if (typeof flags["launch-env"] === "string") {
137
+ try {
138
+ const parsed = JSON.parse(flags["launch-env"]);
139
+ if (!parsed || Array.isArray(parsed) || typeof parsed !== "object" || Object.values(parsed).some((value) => typeof value !== "string")) {
140
+ throw new Error("expected a JSON object with string values");
141
+ }
142
+ appLaunchEnv = parsed;
143
+ } catch (error) {
144
+ console.error(`❌ --launch-env must be a JSON object with string values: ${error.message}`);
145
+ process.exit(2);
146
+ }
147
+ }
148
+ return {
149
+ ...(appLaunchArgs.length ? { appLaunchArgs } : {}),
150
+ ...(appLaunchEnv ? { appLaunchEnv } : {}),
151
+ };
152
+ }
153
+
133
154
  const engineImport = () => import(path.join(packageRoot, "mcp-server", "src", "index.js"));
134
155
 
135
156
  function requireMacFor(what) {
@@ -185,6 +206,26 @@ function printEngineError(r) {
185
206
  }
186
207
  }
187
208
 
209
+ async function promptForInitTarget(details) {
210
+ const choices = Array.isArray(details?.choices) ? details.choices : [];
211
+ if (!choices.length || !process.stdin.isTTY || !process.stderr.isTTY || process.env.CI) return null;
212
+ const { createInterface } = await import("node:readline/promises");
213
+ const terminal = createInterface({ input: process.stdin, output: process.stderr });
214
+ console.error("\nTapp found multiple application targets. Which one should it explore?");
215
+ choices.forEach((choice, index) => console.error(` ${index + 1}) ${choice.platform} · ${choice.name} (${choice.sourcePath})`));
216
+ try {
217
+ while (true) {
218
+ const answer = String(await terminal.question(`Select 1-${choices.length} (or q to cancel): `)).trim();
219
+ if (/^(?:q|quit|cancel)$/i.test(answer)) return null;
220
+ const selected = Number(answer);
221
+ if (Number.isInteger(selected) && selected >= 1 && selected <= choices.length) return choices[selected - 1];
222
+ console.error(`Enter a number from 1 to ${choices.length}, or q to cancel.`);
223
+ }
224
+ } finally {
225
+ terminal.close();
226
+ }
227
+ }
228
+
188
229
  // Turn whatever the user gave us (nothing / repo dir / .app / bundle id) into an installed
189
230
  // bundle id, narrating build/install progress on stderr.
190
231
  async function resolveTargetOrExit(engine, input) {
@@ -197,11 +238,48 @@ async function resolveTargetOrExit(engine, input) {
197
238
  return resolved.bundleId;
198
239
  }
199
240
 
241
+ function safeCommandUsage(verb) {
242
+ const usage = {
243
+ explore: "tapp explore [target] [--platform ios|android|web] [--actions N] [--timeout SEC] [--email VALUE] [--password VALUE] [--baseline FILE] [--json FILE]\n iOS launch configuration: [--launch-arg VALUE ...] [--launch-env '{\"KEY\":\"VALUE\"}']\n Android: [--app-id ID] [--apk FILE] [--serial ID] [--keep-data]",
244
+ init: "tapp init [repo] [--explore] [--refresh] [--platform PLATFORM] [--target NAME] [--url URL] [--dry-run]",
245
+ open: "tapp open [target] [--platform ios|android|web] [--out FILE] [--tap TEXT] [--wait-for TEXT]",
246
+ tree: "tapp tree [target] [--platform ios|android|web] [--json] [--tap TEXT] [--wait-for TEXT]",
247
+ shot: "tapp shot [--out FILE]",
248
+ apps: "tapp apps",
249
+ build: "tapp build [repo] [--scheme NAME] [--configuration NAME]",
250
+ flow: "tapp flow validate FILE [--platform PLATFORM] [--map FILE]\ntapp flow run FILE [--email VALUE] [--password VALUE]",
251
+ task: "tapp task validate FILE [--platform PLATFORM] [--map FILE]\ntapp task compile FILE --platform PLATFORM [--inputs JSON] [--out FILE]\ntapp task run FILE --platform PLATFORM [--url URL|--bundle-id ID|--app-id ID] [--inputs JSON]",
252
+ contract: "tapp contract validate FILE [--platform PLATFORM] [--map FILE]\ntapp contract compile FILE --platform PLATFORM [--out FILE]\ntapp contract run FILE --platform PLATFORM [--url URL|--bundle-id ID|--app-id ID]",
253
+ scenario: "tapp scenario validate FILE [--project-dir DIR]\ntapp scenario run FILE --platform web --url URL [--project-dir DIR]",
254
+ map: "tapp map build MARKERS [--platform PLATFORM] [--out FILE] [--replace]\ntapp map inspect [FILE]\ntapp map diff BEFORE AFTER [--comparable]",
255
+ pr: "tapp pr plan [--base REF|--changed-files FILE] [--head REF] [--platform PLATFORM] [--out FILE]\ntapp pr gate PLAN [gate target/options]\ntapp pr adopt PLAN --item ID [--project-dir DIR]",
256
+ plan: "tapp plan show [FILE]\ntapp plan review [FILE] --approve NAME[,NAME] --reject NAME[,NAME] --defer NAME[,NAME]\ntapp plan generate|validate|promote [FILE] [options]",
257
+ baseline: "tapp baseline create [repo] [--platform PLATFORM] [--target NAME] [--from GATE.json] [--replace]",
258
+ ci: "tapp ci ...\ntapp ci install [repo] [--out FILE] [--manifest FILE] [--dry-run] [--replace]",
259
+ actor: "tapp actor set NAME --email-env ENV --password-env ENV [--project-dir DIR]\ntapp actor list [repo]",
260
+ app: "tapp app [repo] [--no-open] [--port PORT]",
261
+ report: "tapp report [captureId|latest]",
262
+ doctor: "tapp doctor",
263
+ install: "tapp install",
264
+ mcp: "tapp mcp",
265
+ };
266
+ return usage[verb] || `tapp ${verb}`;
267
+ }
268
+
200
269
  // Safe help: `--help`/`-h` on ANY verb prints the command reference and does NOTHING else — never
201
270
  // builds, launches, writes, or opens (ADR-0005 manual-testing requirement). `ci` keeps its own
202
271
  // richer `--help` (a safe usage print in ci-gate.sh); help/version don't need interception.
203
- if ((rest.includes("--help") || rest.includes("-h")) && !["help", "version", "--version", "-v", "ci"].includes(command)) {
204
- console.log(`ℹ️ '${command} --help' showing the command reference (--help never builds, launches, writes, or opens):\n`);
272
+ // A bare `tapp --help` puts the flag in `command`, not `rest`; normalize it before the
273
+ // per-verb interception so the root help path receives the same no-write guarantee.
274
+ if (["--help", "-h"].includes(command)) {
275
+ command = "help";
276
+ rest = [];
277
+ }
278
+ const safeHelpRequested = (rest.includes("--help") || rest.includes("-h"))
279
+ && !["help", "version", "--version", "-v"].includes(command)
280
+ && (command !== "ci" || rest[0] === "install");
281
+ if (safeHelpRequested) {
282
+ console.log(`Usage:\n ${safeCommandUsage(command).replaceAll("\n", "\n ")}\n\nℹ️ --help never builds, launches, writes, or opens. Full command reference:\n`);
205
283
  command = "help";
206
284
  rest = [];
207
285
  }
@@ -244,7 +322,7 @@ switch (command) {
244
322
  }
245
323
  const platform = typeof flags.platform === "string" ? flags.platform.toLowerCase()
246
324
  : typeof flags.url === "string" ? "web"
247
- : typeof flags["app-id"] === "string" || typeof flags.apk === "string" ? "android" : "ios";
325
+ : typeof flags["app-id"] === "string" || typeof flags.apk === "string" ? "android" : "";
248
326
  if (explore && platform === "ios") requireMacFor("iOS init exploration");
249
327
  const actions = flags.actions === undefined ? 40 : Number(flags.actions);
250
328
  const timeout = flags.timeout === undefined ? 600 : Number(flags.timeout);
@@ -254,31 +332,52 @@ switch (command) {
254
332
  }
255
333
  const engine = explore ? await engineImport() : null;
256
334
  const { initializeProductProject } = await import(path.join(packageRoot, "mcp-server", "src", "product-operations.js"));
335
+ const initOptions = {
336
+ projectDir,
337
+ mode: flags["dry-run"] === true ? "inspect" : explore ? "explore" : flags.refresh === true ? "refresh" : "write",
338
+ ownedUrl: typeof flags.url === "string" ? flags.url : "",
339
+ platform: typeof flags.platform === "string" ? flags.platform.toLowerCase() : explore ? platform : "",
340
+ target: typeof flags.target === "string" ? flags.target : "",
341
+ bundleId: typeof flags["bundle-id"] === "string" ? flags["bundle-id"] : "",
342
+ appId: typeof flags["app-id"] === "string" ? flags["app-id"] : "",
343
+ apkPath: typeof flags.apk === "string" ? path.resolve(flags.apk) : undefined,
344
+ serial: typeof flags.serial === "string" ? flags.serial : undefined,
345
+ maxActions: actions,
346
+ timeout,
347
+ testEmail: typeof flags.email === "string" ? flags.email : undefined,
348
+ testPassword: typeof flags.password === "string" ? flags.password : undefined,
349
+ runExploration: engine?.runInitExploration,
350
+ onProgress: (progress) => {
351
+ const activePlatform = progress.platform || platform;
352
+ process.stderr.write(`\r🔍 Import exploration… ${progress.action}/${progress.max || actions} actions · ${progress.states} ${activePlatform === "web" ? "pages reached" : activePlatform === "ios" ? "structural states observed" : "screens reached"} `);
353
+ },
354
+ onStatus: (status) => console.error(`⏳ ${status}`),
355
+ outDir,
356
+ maxContracts,
357
+ };
257
358
  let result;
359
+ let failure = null;
258
360
  try {
259
- result = await initializeProductProject({
260
- projectDir,
261
- mode: flags["dry-run"] === true ? "inspect" : explore ? "explore" : flags.refresh === true ? "refresh" : "write",
262
- ownedUrl: typeof flags.url === "string" ? flags.url : "",
263
- platform: typeof flags.platform === "string" ? flags.platform.toLowerCase() : explore ? platform : "",
264
- target: typeof flags.target === "string" ? flags.target : projectDir,
265
- bundleId: typeof flags["bundle-id"] === "string" ? flags["bundle-id"] : "",
266
- appId: typeof flags["app-id"] === "string" ? flags["app-id"] : "",
267
- apkPath: typeof flags.apk === "string" ? path.resolve(flags.apk) : undefined,
268
- serial: typeof flags.serial === "string" ? flags.serial : undefined,
269
- maxActions: actions,
270
- timeout,
271
- testEmail: typeof flags.email === "string" ? flags.email : undefined,
272
- testPassword: typeof flags.password === "string" ? flags.password : undefined,
273
- runExploration: engine?.runInitExploration,
274
- onProgress: (progress) => process.stderr.write(`\r🔍 Import exploration… ${progress.action}/${progress.max || actions} actions · ${progress.states} ${platform === "web" ? "pages" : "screens"} reached `),
275
- onStatus: (status) => console.error(`⏳ ${status}`),
276
- outDir,
277
- maxContracts,
278
- });
361
+ result = await initializeProductProject(initOptions);
279
362
  } catch (error) {
363
+ failure = error;
364
+ const choice = explore && error.details?.reason === "target-selection-required"
365
+ ? await promptForInitTarget(error.details)
366
+ : null;
367
+ if (choice) {
368
+ if (choice.platform === "ios") requireMacFor("iOS init exploration");
369
+ console.error(`🎯 Exploring ${choice.platform}:${choice.name} (${choice.sourcePath})`);
370
+ try {
371
+ result = await initializeProductProject({ ...initOptions, platform: choice.platform, target: choice.selector });
372
+ failure = null;
373
+ } catch (retryError) {
374
+ failure = retryError;
375
+ }
376
+ }
377
+ }
378
+ if (failure) {
280
379
  if (explore) process.stderr.write("\n");
281
- console.error(`❌ Could not initialize repository: ${error.message || String(error)}`);
380
+ printEngineError({ error: `Could not initialize repository: ${failure.message || String(failure)}`, details: failure.details || {} });
282
381
  process.exit(2);
283
382
  }
284
383
  if (explore) process.stderr.write("\n");
@@ -353,7 +452,10 @@ switch (command) {
353
452
  testPassword: typeof flags.password === "string" ? flags.password : undefined,
354
453
  startWebTarget: engine.startManagedWebTarget,
355
454
  stopWebTarget: engine.stopManagedWebTarget,
356
- onProgress: (entry) => { if (entry.text) console.error(`⏳ ${entry.text}`); },
455
+ // Contract execution is emitted in full on stdout below. Keep build/runtime/replay
456
+ // status live on stderr, but do not echo the execution transcript there as well — an
457
+ // interactive terminal merges the streams and would otherwise show every result twice.
458
+ onProgress: (entry) => { if (entry.text && entry.phase !== "execute") console.error(`⏳ ${entry.text}`); },
357
459
  });
358
460
  } catch (error) {
359
461
  console.error(`❌ Could not validate contract drafts: ${error.message || String(error)}`);
@@ -408,6 +510,7 @@ switch (command) {
408
510
  // hidden deprecated alias.
409
511
  if (command === "qa") console.error("note: 'qa' is now 'explore' — 'qa' still works for now.\n");
410
512
  const { flags, positionals } = parseVerbArgs(rest);
513
+ const launchOptions = iosLaunchOptions(flags, rest);
411
514
  let target = positionals[0] || "";
412
515
  let baselineFindings;
413
516
  if (flags.baseline) {
@@ -431,7 +534,7 @@ switch (command) {
431
534
  const modelPlatform = typeof flags.platform === "string" ? flags.platform.toLowerCase() : "";
432
535
  if (modelPlatform === "ios") requireMacFor("iOS testing");
433
536
  const onProgress = (p) =>
434
- process.stderr.write(`\r🔍 Exploring… ${p.action}/${p.max || flags.actions || 60} actions · ${p.states} reached `);
537
+ process.stderr.write(`\r🔍 Exploring… ${p.action}/${p.max || flags.actions || 60} actions · ${p.states} states observed `);
435
538
  const r = await engine.runExploreTarget({
436
539
  projectDir: process.cwd(),
437
540
  platform: modelPlatform,
@@ -440,6 +543,7 @@ switch (command) {
440
543
  timeout: flags.timeout,
441
544
  testEmail: flags.email,
442
545
  testPassword: flags.password,
546
+ ...launchOptions,
443
547
  baselineFindings,
444
548
  surface: "cli",
445
549
  onProgress,
@@ -461,15 +565,19 @@ switch (command) {
461
565
  process.exit(2);
462
566
  }
463
567
  if (platform === "ios") requireMacFor("iOS testing");
568
+ if (platform !== "ios" && Object.keys(launchOptions).length) {
569
+ console.error("❌ --launch-arg and --launch-env apply only to iOS targets");
570
+ process.exit(2);
571
+ }
464
572
  if (platform === "web" && !/^https?:\/\//i.test(target)) {
465
573
  console.error("❌ Web QA needs an http(s) URL");
466
574
  process.exit(2);
467
575
  }
468
576
  const bundleId = platform === "ios" ? await resolveTargetOrExit(engine, target) : null;
469
577
  const android = platform === "android" ? androidTarget(flags, target) : null;
470
- const unit = platform === "web" ? "pages" : "screens";
578
+ const progressMetric = platform === "web" ? "pages reached" : platform === "ios" ? "structural states observed" : "screens reached";
471
579
  const onProgress = (p) =>
472
- process.stderr.write(`\r🔍 Exploring… ${p.action}/${p.max || flags.actions || 60} actions · ${p.states} ${unit} reached `);
580
+ process.stderr.write(`\r🔍 Exploring… ${p.action}/${p.max || flags.actions || 60} actions · ${p.states} ${progressMetric} `);
473
581
  const r = platform === "web"
474
582
  ? await engine.runQaWeb({
475
583
  url: target,
@@ -497,7 +605,7 @@ switch (command) {
497
605
  bundleId,
498
606
  maxActions: flags.actions,
499
607
  timeout: flags.timeout,
500
- args: { testEmail: flags.email, testPassword: flags.password, baselineFindings },
608
+ args: { testEmail: flags.email, testPassword: flags.password, baselineFindings, ...launchOptions },
501
609
  surface: "cli",
502
610
  onProgress,
503
611
  });
@@ -555,7 +663,6 @@ switch (command) {
555
663
  if (target.apkPath) await driver.install(target.apkPath);
556
664
  const snap = await driver.launch({ clearData: flags["clear-data"] === true });
557
665
  const data = await driver.screenshot();
558
- await driver.forceStop();
559
666
  console.log(`🚀 Launched \`${target.appId}\` (Android)\n`);
560
667
  console.log(engine.formatScreen(snap.screenTitle, snap.elements));
561
668
  const out = typeof flags.out === "string" ? flags.out : path.join(tappHome, "shots", `${target.appId}-${Date.now()}.png`);
@@ -615,12 +722,17 @@ switch (command) {
615
722
  break;
616
723
  }
617
724
  if (platform === "android") {
618
- const target = androidTarget(flags, positionals[0] || "");
725
+ const input = positionals[0] || "";
726
+ const hasTarget = !!(input || flags["app-id"] || flags.apk);
727
+ const target = hasTarget
728
+ ? androidTarget(flags, input)
729
+ : { serial: typeof flags.serial === "string" ? flags.serial : undefined };
619
730
  const { AndroidDriver } = await import(path.join(packageRoot, "mcp-server", "src", "android-driver.js"));
620
731
  const driver = new AndroidDriver(target);
621
732
  await driver.ensureDevice();
622
- const snap = await driver.snapshot();
623
- if (flags.json) console.log(JSON.stringify({ screenTitle: snap.screenTitle, elements: snap.elements }, null, 2));
733
+ if (target.apkPath) await driver.install(target.apkPath);
734
+ const snap = target.appId ? await driver.launch() : await driver.snapshot();
735
+ if (flags.json) console.log(JSON.stringify({ platform: "android", appId: target.appId || null, activity: snap.activity, screenTitle: snap.screenTitle, elements: snap.elements }, null, 2));
624
736
  else console.log(engine.formatScreen(snap.screenTitle, snap.elements));
625
737
  break;
626
738
  }
@@ -1111,8 +1223,14 @@ switch (command) {
1111
1223
  }
1112
1224
 
1113
1225
  try {
1114
- await import("playwright");
1115
- ok("Web", "Playwright installed");
1226
+ const { chromium } = await import("playwright");
1227
+ let executable = "";
1228
+ try { executable = chromium.executablePath(); } catch { /* report the missing browser below */ }
1229
+ if (executable && fs.existsSync(executable)) {
1230
+ ok("Web", `Playwright + Chromium (${executable})`);
1231
+ } else {
1232
+ console.log(" ⬜ Web — Playwright installed; Chromium browser missing (run: npx playwright install chromium)");
1233
+ }
1116
1234
  } catch {
1117
1235
  console.log(" ⬜ Web — install Playwright in the app workspace: npm install -D playwright && npx playwright install chromium");
1118
1236
  }
@@ -1402,9 +1520,15 @@ switch (command) {
1402
1520
  const runs = roots
1403
1521
  .flatMap((root) => fs.readdirSync(root, { withFileTypes: true }).filter((e) => e.isDirectory()).map((e) => path.join(root, e.name)))
1404
1522
  .sort((a, b) => fs.statSync(b).mtimeMs - fs.statSync(a).mtimeMs);
1405
- const wanted = rest[0] && rest[0] !== "latest" ? runs.find((r) => path.basename(r) === rest[0]) : runs[0];
1523
+ const hasMarkers = (dir) => fs.existsSync(path.join(dir, "ocqa-markers.txt"));
1524
+ // `latest` (default) resolves to the newest *exploration* capture — the captures directory is
1525
+ // also full of flow-*/scenario-* evidence dirs with no ocqa-markers.txt, and picking the newest
1526
+ // of those made `tapp report` fail even though valid exploration captures existed. An explicitly
1527
+ // named capture is honored as-is so a non-exploration capture still gets a clear "no markers".
1528
+ const explicit = rest[0] && rest[0] !== "latest";
1529
+ const wanted = explicit ? runs.find((r) => path.basename(r) === rest[0]) : runs.find(hasMarkers);
1406
1530
  if (!wanted) {
1407
- bad("No captures found", rest[0] ? `no capture named "${rest[0]}"` : "run a QA exploration first");
1531
+ bad("No captures found", explicit ? `no capture named "${rest[0]}"` : "run an exploration first (no capture with exploration markers was found)");
1408
1532
  process.exit(1);
1409
1533
  }
1410
1534
  const { writeHtmlReport } = await import(path.join(packageRoot, "mcp-server", "src", "html-report.js"));
@@ -1414,7 +1538,9 @@ switch (command) {
1414
1538
  process.exit(1);
1415
1539
  }
1416
1540
  ok("Evidence report", out);
1417
- spawnSync("open", [out], { stdio: "ignore" });
1541
+ // Only launch a browser from an interactive terminal — agents, scripts, and tests that invoke
1542
+ // `tapp report` non-interactively get the path without a surprise GUI window.
1543
+ if (process.stdout.isTTY) spawnSync("open", [out], { stdio: "ignore" });
1418
1544
  break;
1419
1545
  }
1420
1546