@yoqa/runner 0.3.21 → 0.3.23

package/dist/index.js

@@ -16801,7 +16801,7 @@ function getCatalogDb() {
 // package.json
 var package_default = {
   name: "@yoqa/runner",
-  version: "0.3.21",
+  version: "0.3.23",
   description: "Yoqa local runner \u2014 Appium + catalog HTTP API (Bun). Start with yoqa serve or the yoqa-runner bin.",
   license: "UNLICENSED",
   type: "module",
@@ -16899,6 +16899,229 @@ import { homedir as homedir6 } from "os";
 import { join as join7 } from "path";
 import { remote } from "webdriverio";
 
+// src/domains/devices/android-gestures.ts
+function isAndroidDriver(browser) {
+  return String(browser.capabilities?.platformName ?? "").toLowerCase() === "android";
+}
+function toPx(norm, size) {
+  return Math.round(Math.min(1000, Math.max(0, norm)) / 1000 * size);
+}
+function pngSizeFromBase64(base64) {
+  if (!base64)
+    return null;
+  try {
+    const buf = Buffer.from(base64.slice(0, 48), "base64");
+    if (buf.length < 24)
+      return null;
+    if (buf[0] !== 137 || buf[1] !== 80 || buf[2] !== 78 || buf[3] !== 71)
+      return null;
+    const width = buf.readUInt32BE(16);
+    const height = buf.readUInt32BE(20);
+    if (width < 1 || height < 1 || width > 16384 || height > 16384)
+      return null;
+    return { width, height };
+  } catch {
+    return null;
+  }
+}
+function preferPointerSize(window2, screenshot) {
+  if (!screenshot || screenshot.width < 1 || screenshot.height < 1)
+    return window2;
+  if (window2.width < 1 || window2.height < 1)
+    return screenshot;
+  const dw = Math.abs(screenshot.width - window2.width) / window2.width;
+  const dh = Math.abs(screenshot.height - window2.height) / window2.height;
+  if (dw < 0.02 && dh < 0.02)
+    return screenshot;
+  return window2;
+}
+function w3cTapActions(x, y, holdMs) {
+  return [
+    {
+      type: "pointer",
+      id: "finger1",
+      parameters: { pointerType: "touch" },
+      actions: [
+        { type: "pointerMove", duration: 0, x, y },
+        { type: "pointerDown", button: 0 },
+        { type: "pause", duration: holdMs },
+        { type: "pointerUp", button: 0 }
+      ]
+    }
+  ];
+}
+function w3cSwipeActions(x1, y1, x2, y2, durationMs) {
+  return [
+    {
+      type: "pointer",
+      id: "finger1",
+      parameters: { pointerType: "touch" },
+      actions: [
+        { type: "pointerMove", duration: 0, x: x1, y: y1 },
+        { type: "pointerDown", button: 0 },
+        { type: "pointerMove", duration: durationMs, x: x2, y: y2 },
+        { type: "pointerUp", button: 0 }
+      ]
+    }
+  ];
+}
+async function w3cPerform(browser, actions) {
+  await browser.performActions(actions);
+  try {
+    await browser.releaseActions();
+  } catch {}
+}
+async function injectTap(browser, x, y, holdMs) {
+  const duration = Math.max(50, holdMs);
+  if (isAndroidDriver(browser)) {
+    try {
+      await browser.execute("mobile: clickGesture", { x, y, duration });
+      return "clickGesture";
+    } catch {}
+  }
+  await w3cPerform(browser, w3cTapActions(x, y, duration));
+  return "w3c";
+}
+async function injectSwipe(browser, x1, y1, x2, y2, durationMs) {
+  const duration = Math.max(50, durationMs);
+  if (isAndroidDriver(browser)) {
+    try {
+      const distance = Math.hypot(x2 - x1, y2 - y1);
+      const speed = Math.max(200, Math.round(distance / (duration / 1000)));
+      await browser.execute("mobile: dragGesture", {
+        startX: x1,
+        startY: y1,
+        endX: x2,
+        endY: y2,
+        speed
+      });
+      return "dragGesture";
+    } catch {}
+  }
+  await w3cPerform(browser, w3cSwipeActions(x1, y1, x2, y2, duration));
+  return "w3c";
+}
+
+// src/domains/devices/android-alerts.ts
+var ANDROID_ACCEPT_BUTTON_LABELS = [
+  "Allow",
+  "ALLOW",
+  "While using the app",
+  "While using this app",
+  "Allow all the time",
+  "Allow only while using the app",
+  "OK"
+];
+var ANDROID_DISMISS_BUTTON_LABELS = [
+  "Don't allow",
+  "Don\u2019t allow",
+  "Don't Allow",
+  "Deny",
+  "DENY",
+  "No thanks"
+];
+var ANDROID_ACCEPT_RESOURCE_IDS = [
+  "com.android.permissioncontroller:id/permission_allow_button",
+  "com.android.permissioncontroller:id/permission_allow_foreground_only_button",
+  "com.android.permissioncontroller:id/permission_allow_one_time_button",
+  "com.android.packageinstaller:id/permission_allow_button",
+  "com.samsung.android.permissioncontroller:id/permission_allow_button"
+];
+var ANDROID_DISMISS_RESOURCE_IDS = [
+  "com.android.permissioncontroller:id/permission_deny_button",
+  "com.android.permissioncontroller:id/permission_deny_and_dont_ask_again_button",
+  "com.android.packageinstaller:id/permission_deny_button",
+  "com.samsung.android.permissioncontroller:id/permission_deny_button"
+];
+function androidAlertAttempts(action) {
+  const labels = action === "accept" ? ANDROID_ACCEPT_BUTTON_LABELS : ANDROID_DISMISS_BUTTON_LABELS;
+  const ids = action === "accept" ? ANDROID_ACCEPT_RESOURCE_IDS : ANDROID_DISMISS_RESOURCE_IDS;
+  const mobile = action === "accept" ? "mobile-acceptAlert" : "mobile-dismissAlert";
+  const attempts = [];
+  for (const buttonLabel of labels) {
+    attempts.push({ method: mobile, buttonLabel });
+  }
+  attempts.push({ method: mobile });
+  for (const id of ids) {
+    attempts.push({ method: "resource-id", id });
+  }
+  for (const text2 of labels) {
+    attempts.push({ method: "text", text: text2 });
+  }
+  return attempts;
+}
+function iosAlertAttempts(action) {
+  return [
+    action === "accept" ? { method: "w3c-acceptAlert" } : { method: "w3c-dismissAlert" },
+    { method: "mobile-alert", action }
+  ];
+}
+function alertAttemptsFor(browser, action) {
+  return isAndroidDriver(browser) ? androidAlertAttempts(action) : iosAlertAttempts(action);
+}
+async function tryClickSelector(browser, selector) {
+  try {
+    const el = browser.$(selector);
+    if (await el.isExisting()) {
+      await el.click();
+      return true;
+    }
+  } catch {
+    return false;
+  }
+  return false;
+}
+function xpathExactText(text2) {
+  if (text2.includes("'") && text2.includes('"')) {
+    const parts = text2.split("'").map((part) => `'${part}'`);
+    return `//*[@text=concat(${parts.join(`, "'", `)})]`;
+  }
+  if (text2.includes("'")) {
+    return `//*[@text="${text2}"]`;
+  }
+  return `//*[@text='${text2}']`;
+}
+async function runAttempt(browser, attempt) {
+  try {
+    switch (attempt.method) {
+      case "mobile-acceptAlert": {
+        const args = attempt.buttonLabel ? { buttonLabel: attempt.buttonLabel } : {};
+        await browser.execute("mobile: acceptAlert", args);
+        return true;
+      }
+      case "mobile-dismissAlert": {
+        const args = attempt.buttonLabel ? { buttonLabel: attempt.buttonLabel } : {};
+        await browser.execute("mobile: dismissAlert", args);
+        return true;
+      }
+      case "resource-id":
+        return await tryClickSelector(browser, `android=new UiSelector().resourceId("${attempt.id}")`);
+      case "text":
+        return await tryClickSelector(browser, xpathExactText(attempt.text));
+      case "w3c-acceptAlert":
+        await browser.acceptAlert();
+        return true;
+      case "w3c-dismissAlert":
+        await browser.dismissAlert();
+        return true;
+      case "mobile-alert":
+        await browser.execute("mobile: alert", { action: attempt.action });
+        return true;
+    }
+  } catch {
+    return false;
+  }
+}
+async function resolveNativeAlert(browser, action) {
+  const attempts = alertAttemptsFor(browser, action);
+  for (const attempt of attempts) {
+    if (await runAttempt(browser, attempt)) {
+      return attempt;
+    }
+  }
+  throw new Error(action === "accept" ? "Could not accept the system alert (no Allow / permission button found)" : "Could not dismiss the system alert (no Don't allow / Deny button found)");
+}
+
 // src/domains/devices/application.ts
 import { mkdtemp as mkdtemp2, readFile, rm as rm2 } from "fs/promises";
 import { tmpdir as tmpdir2 } from "os";
@@ -17557,7 +17780,7 @@ class ActionGate {
       this.locked = false;
     }
   }
-  async handlePointer(browser, toPx, getWindowSize, phase, xNorm, yNorm, seq) {
+  async handlePointer(browser, toPx2, getWindowSize, phase, xNorm, yNorm, seq) {
     if (this.locked && !this.gesture && !this.deferredTap) {
       throw new Error("Device is busy with another action");
     }
@@ -17587,7 +17810,7 @@ class ActionGate {
         await this.flushGesture(pending.gesture, pending.browser, pending.toPx, pending.getWindowSize);
       }
       if (this.gesture) {
-        await this.flushGesture(this.gesture, browser, toPx, getWindowSize);
+        await this.flushGesture(this.gesture, browser, toPx2, getWindowSize);
         this.gesture = null;
       }
       this.locked = true;
@@ -17616,16 +17839,16 @@ class ActionGate {
     this.gesture.endXNorm = x;
     this.gesture.endYNorm = y;
     const size = await getWindowSize();
-    const startX = toPx(this.gesture.startXNorm, size.width);
-    const startY = toPx(this.gesture.startYNorm, size.height);
-    const endX = toPx(this.gesture.endXNorm, size.width);
-    const endY = toPx(this.gesture.endYNorm, size.height);
+    const startX = toPx2(this.gesture.startXNorm, size.width);
+    const startY = toPx2(this.gesture.startYNorm, size.height);
+    const endX = toPx2(this.gesture.endXNorm, size.width);
+    const endY = toPx2(this.gesture.endYNorm, size.height);
     const distance = Math.hypot(endX - startX, endY - startY);
     const tapSlopPx = 12;
     if (distance >= tapSlopPx || this.gesture.double) {
       const gesture2 = this.gesture;
       this.gesture = null;
-      await this.flushGesture(gesture2, browser, toPx, getWindowSize);
+      await this.flushGesture(gesture2, browser, toPx2, getWindowSize);
       return;
     }
     const gesture = this.gesture;
@@ -17640,74 +17863,41 @@ class ActionGate {
         this.locked = false;
       });
     }, doubleTapMs);
-    this.deferredTap = { gesture, browser, toPx, getWindowSize, timer };
+    this.deferredTap = { gesture, browser, toPx: toPx2, getWindowSize, timer };
   }
-  async flushGesture(gesture, browser, toPx, getWindowSize) {
+  async flushGesture(gesture, browser, toPx2, getWindowSize) {
     const size = await getWindowSize();
-    const startX = toPx(gesture.startXNorm, size.width);
-    const startY = toPx(gesture.startYNorm, size.height);
-    const endX = toPx(gesture.endXNorm, size.width);
-    const endY = toPx(gesture.endYNorm, size.height);
+    const startX = toPx2(gesture.startXNorm, size.width);
+    const startY = toPx2(gesture.startYNorm, size.height);
+    const endX = toPx2(gesture.endXNorm, size.width);
+    const endY = toPx2(gesture.endYNorm, size.height);
     const distance = Math.hypot(endX - startX, endY - startY);
     const elapsedMs = Math.max(50, Date.now() - gesture.startedAt);
     const tapSlopPx = 12;
     try {
       if (distance < tapSlopPx || gesture.double) {
         const holdMs = Math.min(200, elapsedMs);
-        const tapOnce = async () => {
-          await browser.performActions([
-            {
-              type: "pointer",
-              id: "finger1",
-              parameters: { pointerType: "touch" },
-              actions: [
-                { type: "pointerMove", duration: 0, x: startX, y: startY },
-                { type: "pointerDown", button: 0 },
-                { type: "pause", duration: holdMs },
-                { type: "pointerUp", button: 0 }
-              ]
-            }
-          ]);
-          try {
-            await browser.releaseActions();
-          } catch {}
-        };
-        await tapOnce();
+        await injectTap(browser, startX, startY, holdMs);
         if (gesture.double) {
           await Bun.sleep(50);
-          await tapOnce();
+          await injectTap(browser, startX, startY, holdMs);
         }
       } else {
         const duration = Math.min(2000, Math.max(80, elapsedMs));
-        await browser.performActions([
-          {
-            type: "pointer",
-            id: "finger1",
-            parameters: { pointerType: "touch" },
-            actions: [
-              { type: "pointerMove", duration: 0, x: startX, y: startY },
-              { type: "pointerDown", button: 0 },
-              { type: "pointerMove", duration, x: endX, y: endY },
-              { type: "pointerUp", button: 0 }
-            ]
-          }
-        ]);
-        try {
-          await browser.releaseActions();
-        } catch {}
+        await injectSwipe(browser, startX, startY, endX, endY, duration);
       }
     } finally {
       this.locked = false;
     }
   }
-  async forceEnd(browser, toPx, getWindowSize) {
+  async forceEnd(browser, toPx2, getWindowSize) {
     if (this.deferredTap) {
       clearTimeout(this.deferredTap.timer);
       const pending = this.deferredTap;
       this.deferredTap = null;
-      if (toPx && getWindowSize) {
+      if (toPx2 && getWindowSize) {
         try {
-          await this.flushGesture(pending.gesture, pending.browser, toPx, getWindowSize);
+          await this.flushGesture(pending.gesture, pending.browser, toPx2, getWindowSize);
         } catch {
           this.locked = false;
         }
@@ -17720,7 +17910,7 @@ class ActionGate {
       this.locked = false;
       return;
     }
-    if (!toPx || !getWindowSize) {
+    if (!toPx2 || !getWindowSize) {
       this.gesture = null;
       this.locked = false;
       return;
@@ -17728,7 +17918,7 @@ class ActionGate {
     const gesture = this.gesture;
     this.gesture = null;
     try {
-      await this.flushGesture(gesture, browser, toPx, getWindowSize);
+      await this.flushGesture(gesture, browser, toPx2, getWindowSize);
     } catch {
       this.locked = false;
     }
@@ -17791,7 +17981,11 @@ async function createDeviceSession(options) {
     }
   };
   const getWindowSize = async () => guard(() => browser.getWindowSize());
-  const toPx = (norm, size) => Math.round(Math.min(1000, Math.max(0, norm)) / 1000 * size);
+  let lastShotSize = null;
+  const getPointerSize = async () => {
+    const window2 = await getWindowSize();
+    return preferPointerSize(window2, lastShotSize);
+  };
   const owned = { current: null };
   const quit = async () => {
     if (openByDeviceId.get(options.deviceId) === owned.current) {
@@ -17799,7 +17993,7 @@ async function createDeviceSession(options) {
     }
     if (gate.isPointerActive() && !sessionDeadNotified) {
       try {
-        await withTimeout(gate.forceEnd(browser, toPx, getWindowSize), 3000, "live gesture flush");
+        await withTimeout(gate.forceEnd(browser, toPx, getPointerSize), 3000, "live gesture flush");
       } catch (error) {
         console.warn("[yoqa-runner] gesture flush on quit:", error instanceof Error ? error.message : error);
       }
@@ -17818,6 +18012,7 @@ async function createDeviceSession(options) {
   };
   const captureFrame = async () => guard(async () => {
     const base64 = await browser.takeScreenshot();
+    lastShotSize = pngSizeFromBase64(base64) ?? lastShotSize;
     return { base64, mime: "image/png" };
   });
   const screenshot = async () => {
@@ -17831,55 +18026,19 @@ async function createDeviceSession(options) {
   const tap = async (xNorm, yNorm, tapOptions) => {
     await gate.withLock(async () => {
       await guard(async () => {
-        const size = await getWindowSize();
+        const size = await getPointerSize();
         const x = toPx(xNorm, size.width);
         const y = toPx(yNorm, size.height);
         const holdMs = Math.max(50, tapOptions?.durationMs ?? 50);
-        await browser.performActions([
-          {
-            type: "pointer",
-            id: "finger1",
-            parameters: { pointerType: "touch" },
-            actions: [
-              { type: "pointerMove", duration: 0, x, y },
-              { type: "pointerDown", button: 0 },
-              { type: "pause", duration: holdMs },
-              { type: "pointerUp", button: 0 }
-            ]
-          }
-        ]);
-        await browser.releaseActions();
+        await injectTap(browser, x, y, holdMs);
       });
     });
   };
   const swipe = async (x1, y1, x2, y2, durationMs = 400) => {
     await gate.withLock(async () => {
       await guard(async () => {
-        const size = await getWindowSize();
-        await browser.performActions([
-          {
-            type: "pointer",
-            id: "finger1",
-            parameters: { pointerType: "touch" },
-            actions: [
-              {
-                type: "pointerMove",
-                duration: 0,
-                x: toPx(x1, size.width),
-                y: toPx(y1, size.height)
-              },
-              { type: "pointerDown", button: 0 },
-              {
-                type: "pointerMove",
-                duration: durationMs,
-                x: toPx(x2, size.width),
-                y: toPx(y2, size.height)
-              },
-              { type: "pointerUp", button: 0 }
-            ]
-          }
-        ]);
-        await browser.releaseActions();
+        const size = await getPointerSize();
+        await injectSwipe(browser, toPx(x1, size.width), toPx(y1, size.height), toPx(x2, size.width), toPx(y2, size.height), durationMs);
       });
     });
   };
@@ -17924,20 +18083,20 @@ async function createDeviceSession(options) {
   const acceptAlert = async () => {
     await gate.withLock(async () => {
       await guard(async () => {
-        await browser.acceptAlert();
+        await resolveNativeAlert(browser, "accept");
       });
     });
   };
   const dismissAlert = async () => {
     await gate.withLock(async () => {
       await guard(async () => {
-        await browser.dismissAlert();
+        await resolveNativeAlert(browser, "dismiss");
       });
     });
   };
   const pointerEvent = async (phase, xNorm, yNorm, seq) => {
     await guard(async () => {
-      await gate.handlePointer(browser, toPx, getWindowSize, phase, xNorm, yNorm, seq);
+      await gate.handlePointer(browser, toPx, getPointerSize, phase, xNorm, yNorm, seq);
     });
   };
   const session = {
@@ -24118,6 +24277,11 @@ var caseScriptActionSchema = exports_external.union([
     text: exports_external.string().min(1),
     timeoutMs: exports_external.number().min(0).max(120000).optional(),
     reason: exports_external.string().optional()
+  }),
+  exports_external.object({
+    type: exports_external.literal("alert"),
+    alertAction: exports_external.union([exports_external.literal("accept"), exports_external.literal("dismiss")]).default("accept"),
+    reason: exports_external.string().optional()
   })
 ]);
 var caseScriptSchema = exports_external.object({
@@ -24789,12 +24953,26 @@ function buildScriptFromDecisions(decisions, sourceRunId) {
   const actions = [];
   for (const decision of decisions) {
     if (decision.type === "tap") {
-      actions.push({
-        type: "tap",
-        x: decision.x ?? 500,
-        y: decision.y ?? 500,
-        reason: decision.reason
-      });
+      if (decision.label) {
+        actions.push({
+          type: "tap",
+          label: decision.label,
+          reason: decision.reason
+        });
+      } else if (decision.id) {
+        actions.push({
+          type: "tap",
+          id: decision.id,
+          reason: decision.reason
+        });
+      } else {
+        actions.push({
+          type: "tap",
+          x: decision.x ?? 500,
+          y: decision.y ?? 500,
+          reason: decision.reason
+        });
+      }
     } else if (decision.type === "type") {
       actions.push({
         type: "type",
@@ -24807,6 +24985,12 @@ function buildScriptFromDecisions(decisions, sourceRunId) {
         ms: Math.min(3000, Math.max(500, decision.ms ?? 1500)),
         reason: decision.reason
       });
+    } else if (decision.type === "alert") {
+      actions.push({
+        type: "alert",
+        alertAction: decision.alertAction === "dismiss" ? "dismiss" : "accept",
+        reason: decision.reason
+      });
     }
   }
   if (actions.length === 0)
@@ -57546,6 +57730,195 @@ var originalGenerateCallId6 = createIdGenerator3({
 });
 var defaultDownload2 = createDownload();
 
+// src/domains/providers/json-salvage.ts
+function charAt(raw2, index) {
+  return raw2[index];
+}
+function coerceLooseJson(raw2) {
+  let out = "";
+  let i = 0;
+  while (i < raw2.length) {
+    const c = charAt(raw2, i);
+    if (c === undefined)
+      break;
+    if (c === '"') {
+      out += c;
+      i += 1;
+      while (i < raw2.length) {
+        const d = charAt(raw2, i);
+        if (d === undefined)
+          break;
+        out += d;
+        i += 1;
+        if (d === "\\") {
+          const escaped = charAt(raw2, i);
+          if (escaped !== undefined) {
+            out += escaped;
+            i += 1;
+          }
+          continue;
+        }
+        if (d === '"')
+          break;
+      }
+      continue;
+    }
+    if (c === "'") {
+      i += 1;
+      let content = "";
+      while (i < raw2.length) {
+        const d = charAt(raw2, i);
+        if (d === undefined)
+          break;
+        if (d === "\\" && i + 1 < raw2.length) {
+          const next = charAt(raw2, i + 1);
+          if (next === undefined)
+            break;
+          if (next === "n")
+            content += `
+`;
+          else if (next === "t")
+            content += "\t";
+          else if (next === "r")
+            content += "\r";
+          else
+            content += next;
+          i += 2;
+          continue;
+        }
+        if (d === "'") {
+          i += 1;
+          break;
+        }
+        content += d;
+        i += 1;
+      }
+      out += JSON.stringify(content);
+      continue;
+    }
+    if (c === ",") {
+      let j = i + 1;
+      while (j < raw2.length) {
+        const ws = charAt(raw2, j);
+        if (ws === undefined || !/\s/.test(ws))
+          break;
+        j += 1;
+      }
+      const closer = charAt(raw2, j);
+      if (closer === "}" || closer === "]") {
+        i += 1;
+        continue;
+      }
+    }
+    out += c;
+    i += 1;
+  }
+  return out;
+}
+function tryParseJson(text3) {
+  try {
+    return JSON.parse(text3);
+  } catch {
+    return;
+  }
+}
+function closeTruncatedJson(raw2) {
+  let inString = false;
+  let escaped = false;
+  const stack = [];
+  for (let i = 0;i < raw2.length; i += 1) {
+    const c = charAt(raw2, i);
+    if (c === undefined)
+      break;
+    if (inString) {
+      if (escaped) {
+        escaped = false;
+        continue;
+      }
+      if (c === "\\") {
+        escaped = true;
+        continue;
+      }
+      if (c === '"')
+        inString = false;
+      continue;
+    }
+    if (c === '"') {
+      inString = true;
+      continue;
+    }
+    if (c === "{")
+      stack.push("{");
+    else if (c === "[")
+      stack.push("[");
+    else if (c === "}" || c === "]")
+      stack.pop();
+  }
+  let out = raw2;
+  if (inString) {
+    if (escaped)
+      out = out.slice(0, -1);
+    out += '"';
+  }
+  while (stack.length > 0) {
+    const open = stack.pop();
+    out += open === "{" ? "}" : "]";
+  }
+  return out;
+}
+function salvageAgentJsonText(text3) {
+  const trimmed = text3.trim();
+  const fenced = trimmed.match(/```(?:json)?\s*([\s\S]*?)```/);
+  const inner = fenced?.[1]?.trim() ?? trimmed;
+  const start = inner.indexOf("{");
+  if (start < 0)
+    return trimmed;
+  const fromBrace = inner.slice(start);
+  const lastBrace = fromBrace.lastIndexOf("}");
+  const candidates = lastBrace > 0 ? [fromBrace.slice(0, lastBrace + 1), fromBrace] : [fromBrace];
+  for (const candidate of candidates) {
+    const variants = [
+      candidate,
+      coerceLooseJson(candidate),
+      closeTruncatedJson(candidate),
+      closeTruncatedJson(coerceLooseJson(candidate))
+    ];
+    for (const variant of variants) {
+      if (tryParseJson(variant) !== undefined)
+        return variant;
+    }
+  }
+  return closeTruncatedJson(coerceLooseJson(fromBrace));
+}
+function clampNormCoord(n) {
+  if (!Number.isFinite(n))
+    return n;
+  return Math.min(1000, Math.max(0, n));
+}
+function normalizeVisionJson(value) {
+  if (!value || typeof value !== "object" || Array.isArray(value))
+    return value;
+  const out = { ...value };
+  if (typeof out.x === "number")
+    out.x = clampNormCoord(out.x);
+  if (typeof out.y === "number")
+    out.y = clampNormCoord(out.y);
+  if (typeof out.ms === "number" && Number.isFinite(out.ms)) {
+    out.ms = Math.min(1e4, Math.max(0, out.ms));
+  }
+  const reason = typeof out.reason === "string" ? out.reason.trim() : "";
+  const thoughts = typeof out.thoughts === "string" ? out.thoughts.trim() : "";
+  if (reason)
+    out.reason = reason;
+  if (thoughts)
+    out.thoughts = thoughts;
+  if (reason && !thoughts)
+    out.thoughts = reason;
+  if (thoughts && !reason)
+    out.reason = thoughts;
+  return out;
+}
+
 // src/domains/providers/vision-model.ts
 class AgentProviderError extends Error {
   constructor(message) {
@@ -57646,20 +58019,27 @@ function mapProviderError(label, error40) {
     }
     throw new AgentProviderError(`Model did not return JSON (got: ${snippet}${text3.length > 160 ? "\u2026" : ""})`);
   }
+  if (error40 instanceof exports_external.ZodError) {
+    const first = error40.issues[0];
+    const detail = first ? `${first.path.length > 0 ? `${first.path.join(".")}: ` : ""}${first.message}` : error40.message;
+    throw new AgentProviderError(`${label} JSON was not a valid action: ${detail}`);
+  }
   if (error40 instanceof Error) {
     throw new AgentProviderError(error40.message);
   }
   throw new AgentProviderError(String(error40));
 }
-function repairModelJsonText(text3) {
-  const trimmed = text3.trim();
-  const fenced = trimmed.match(/```(?:json)?\s*([\s\S]*?)```/);
-  const candidate = fenced?.[1]?.trim() ?? trimmed;
-  const start = candidate.indexOf("{");
-  const end = candidate.lastIndexOf("}");
-  if (start < 0 || end <= start)
-    return trimmed;
-  return candidate.slice(start, end + 1);
+function parseVisionSchema(schema, raw2, label) {
+  try {
+    return schema.parse(normalizeVisionJson(raw2));
+  } catch (error40) {
+    if (error40 instanceof exports_external.ZodError) {
+      const first = error40.issues[0];
+      const detail = first ? `${first.path.length > 0 ? `${first.path.join(".")}: ` : ""}${first.message}` : error40.message;
+      throw new AgentProviderError(`${label} JSON was not a valid action: ${detail}`);
+    }
+    throw error40;
+  }
 }
 async function completeWithAiSdk(input) {
   const run = async (prompt) => {
@@ -57669,7 +58049,7 @@ async function completeWithAiSdk(input) {
         schema: input.schema,
         system: input.system,
         maxOutputTokens: VISION_MAX_TOKENS,
-        experimental_repairText: async ({ text: text3 }) => repairModelJsonText(text3),
+        experimental_repairText: async ({ text: text3 }) => salvageAgentJsonText(text3),
         messages: [
           {
             role: "user",
@@ -57684,12 +58064,18 @@ async function completeWithAiSdk(input) {
           }
         ]
       });
-      return object3;
+      return parseVisionSchema(input.schema, object3, input.label);
     } catch (error40) {
+      if (error40 instanceof AgentProviderError)
+        throw error40;
       if (NoObjectGeneratedError.isInstance(error40) && error40.text?.trim()) {
         try {
-          return input.schema.parse(JSON.parse(repairModelJsonText(error40.text)));
-        } catch {}
+          const salvaged = salvageAgentJsonText(error40.text);
+          return parseVisionSchema(input.schema, JSON.parse(salvaged), input.label);
+        } catch (salvageError) {
+          if (salvageError instanceof AgentProviderError)
+            throw salvageError;
+        }
       }
       mapProviderError(input.label, error40);
     }
@@ -67601,108 +67987,37 @@ var google = createGoogle();
 
 // src/domains/providers/agent-json.ts
 function extractAgentJsonObject(text3, providerLabel) {
-  const trimmed = text3.trim();
-  const fenced = trimmed.match(/```(?:json)?\s*([\s\S]*?)```/);
-  const candidate = fenced?.[1]?.trim() ?? trimmed;
-  const start = candidate.indexOf("{");
-  const end = candidate.lastIndexOf("}");
-  if (start < 0 || end <= start) {
-    throw new AgentProviderError(`${providerLabel} did not return JSON (got: ${trimmed.replace(/\s+/g, " ").slice(0, 160)})`);
-  }
-  const slice = candidate.slice(start, end + 1);
+  const snippet = text3.trim().replace(/\s+/g, " ").slice(0, 160);
+  const fail = () => {
+    throw new AgentProviderError(`${providerLabel} did not return JSON (got: ${snippet})`);
+  };
+  const salvaged = salvageAgentJsonText(text3);
+  let parsed;
   try {
-    return JSON.parse(slice);
-  } catch (strictError) {
+    parsed = JSON.parse(salvaged);
+  } catch {
     try {
-      return JSON.parse(coerceLooseJson(slice));
+      parsed = JSON.parse(coerceLooseJson(salvaged));
     } catch {
-      throw strictError;
+      fail();
     }
   }
+  if (parsed === undefined || parsed === null || typeof parsed !== "object") {
+    fail();
+  }
+  return parsed;
 }
-function charAt(raw2, index) {
-  return raw2[index];
-}
-function coerceLooseJson(raw2) {
-  let out = "";
-  let i = 0;
-  while (i < raw2.length) {
-    const c = charAt(raw2, i);
-    if (c === undefined)
-      break;
-    if (c === '"') {
-      out += c;
-      i += 1;
-      while (i < raw2.length) {
-        const d = charAt(raw2, i);
-        if (d === undefined)
-          break;
-        out += d;
-        i += 1;
-        if (d === "\\") {
-          const escaped = charAt(raw2, i);
-          if (escaped !== undefined) {
-            out += escaped;
-            i += 1;
-          }
-          continue;
-        }
-        if (d === '"')
-          break;
-      }
-      continue;
-    }
-    if (c === "'") {
-      i += 1;
-      let content = "";
-      while (i < raw2.length) {
-        const d = charAt(raw2, i);
-        if (d === undefined)
-          break;
-        if (d === "\\" && i + 1 < raw2.length) {
-          const next = charAt(raw2, i + 1);
-          if (next === undefined)
-            break;
-          if (next === "n")
-            content += `
-`;
-          else if (next === "t")
-            content += "\t";
-          else if (next === "r")
-            content += "\r";
-          else
-            content += next;
-          i += 2;
-          continue;
-        }
-        if (d === "'") {
-          i += 1;
-          break;
-        }
-        content += d;
-        i += 1;
-      }
-      out += JSON.stringify(content);
-      continue;
-    }
-    if (c === ",") {
-      let j = i + 1;
-      while (j < raw2.length) {
-        const ws = charAt(raw2, j);
-        if (ws === undefined || !/\s/.test(ws))
-          break;
-        j += 1;
-      }
-      const closer = charAt(raw2, j);
-      if (closer === "}" || closer === "]") {
-        i += 1;
-        continue;
-      }
+function parseVisionObject(schema, raw2, providerLabel) {
+  try {
+    return schema.parse(normalizeVisionJson(raw2));
+  } catch (error40) {
+    if (error40 instanceof exports_external.ZodError) {
+      const first = error40.issues[0];
+      const detail = first ? `${first.path.length > 0 ? `${first.path.join(".")}: ` : ""}${first.message}` : error40.message;
+      throw new AgentProviderError(`${providerLabel} JSON was not a valid action: ${detail}`);
     }
-    out += c;
-    i += 1;
+    throw error40;
   }
-  return out;
 }
 
 // src/domains/providers/drivers/antigravity-vision.ts
@@ -67748,13 +68063,10 @@ ${result.stderr}`.trim();
     if (result.exitCode !== 0 && !result.stdout.trim()) {
       throw new AgentProviderError(combined ? `Antigravity CLI failed: ${combined.slice(0, 400)}` : `Antigravity CLI exited ${result.exitCode}`);
     }
-    return input.schema.parse(extractAgentJsonObject(result.stdout || combined, "Antigravity CLI"));
+    return parseVisionObject(input.schema, extractAgentJsonObject(result.stdout || combined, "Antigravity CLI"), "Antigravity CLI");
   } catch (error40) {
     if (error40 instanceof AgentProviderError)
       throw error40;
-    if (error40 instanceof exports_external.ZodError) {
-      throw new AgentProviderError(`Antigravity JSON was not a valid action: ${error40.message}`);
-    }
     if (error40 instanceof SyntaxError) {
       throw new AgentProviderError(`Antigravity CLI returned invalid JSON: ${error40.message}`);
     }
@@ -76312,13 +76624,10 @@ Screenshot file (open and look at it): ${shotName}`,
       imageBase64: image.base64,
       mediaType: image.mediaType
     });
-    return input.schema.parse(extractAgentJsonObject(stdout, "Cursor Agent CLI"));
+    return parseVisionObject(input.schema, extractAgentJsonObject(stdout, "Cursor Agent CLI"), "Cursor Agent CLI");
   } catch (error40) {
     if (error40 instanceof AgentProviderError)
       throw error40;
-    if (error40 instanceof exports_external.ZodError) {
-      throw new AgentProviderError(`Cursor JSON was not a valid action: ${error40.message}`);
-    }
     if (error40 instanceof SyntaxError) {
       throw new AgentProviderError(`Cursor Agent CLI returned invalid JSON: ${error40.message}`);
     }
@@ -97455,14 +97764,18 @@ var agentDecisionSchema = exports_external.object({
     exports_external.literal("tap"),
     exports_external.literal("type"),
     exports_external.literal("wait"),
+    exports_external.literal("alert"),
     exports_external.literal("verify"),
     exports_external.literal("done"),
     exports_external.literal("fail")
   ]),
-  x: exports_external.number().min(0).max(1000).optional(),
-  y: exports_external.number().min(0).max(1000).optional(),
+  x: exports_external.number().optional(),
+  y: exports_external.number().optional(),
+  label: exports_external.string().min(1).optional(),
+  id: exports_external.string().min(1).optional(),
   text: exports_external.string().optional(),
-  ms: exports_external.number().min(0).max(1e4).optional(),
+  ms: exports_external.number().optional(),
+  alertAction: exports_external.union([exports_external.literal("accept"), exports_external.literal("dismiss")]).optional(),
   reason: exports_external.string().min(1),
   thoughts: exports_external.string().min(1)
 });
@@ -97475,17 +97788,21 @@ Every action MUST include:
 - "thoughts": 2\u20134 sentences describing what you see on screen and how that led to this action
 
 Valid shapes:
+{"type":"tap","label":"Allow","reason":"...","thoughts":"..."}
+{"type":"tap","id":"permission_allow_button","reason":"...","thoughts":"..."}
 {"type":"tap","x":0-1000,"y":0-1000,"reason":"...","thoughts":"..."}
+{"type":"alert","alertAction":"accept","reason":"...","thoughts":"..."}
 {"type":"type","text":"...","reason":"...","thoughts":"..."}
 {"type":"wait","ms":500-3000,"reason":"...","thoughts":"..."}
 {"type":"verify","reason":"...","thoughts":"..."}
 {"type":"done","reason":"...","thoughts":"..."}
 {"type":"fail","reason":"...","thoughts":"..."}
 
-Coordinates use a 0\u20131000 normalized grid (0,0 top-left).
+Coordinates use a 0\u20131000 normalized grid (0,0 top-left). Prefer label or id over coordinates whenever the control has visible text or a resource-id.
 
 Rules:
 - Look at the attached screenshot before deciding. Base taps on what is visible now.
+- System permission / notification dialogs (Allow, Don't allow, While using the app): use {"type":"tap","label":"Allow"} or {"type":"alert","alertAction":"accept"}. Never guess coordinates for these buttons \u2014 coordinate taps often miss.
 - If the expected result (or the goal of the instructions) is already visible, return verify or done \u2014 do not keep tapping the same control.
 - Splash / loading / blank screens: return wait (do not tap imaginary tabs).
 - Do not repeat the same tap target from Recent actions unless the screenshot still shows you are on the wrong screen.
@@ -97496,8 +97813,17 @@ function formatRecentActions(actions) {
   return actions.slice(-6).map((action, i) => {
     const n = actions.length - Math.min(actions.length, 6) + i + 1;
     if (action.type === "tap") {
+      if (action.label) {
+        return `${n}. tap label "${action.label}" \u2014 ${action.reason ?? ""}`;
+      }
+      if (action.id) {
+        return `${n}. tap id "${action.id}" \u2014 ${action.reason ?? ""}`;
+      }
       return `${n}. tap (${action.x ?? "?"},${action.y ?? "?"}) \u2014 ${action.reason ?? ""}`;
     }
+    if (action.type === "alert") {
+      return `${n}. alert ${action.alertAction ?? "accept"} \u2014 ${action.reason ?? ""}`;
+    }
     if (action.type === "type") {
       return `${n}. type "${action.text ?? ""}" \u2014 ${action.reason ?? ""}`;
     }
@@ -97658,8 +97984,8 @@ function cleanPageSource(xml, window2) {
 
 // src/domains/devices/grounding.ts
 var groundResultSchema = exports_external.object({
-  x: exports_external.number().min(0).max(1000),
-  y: exports_external.number().min(0).max(1000)
+  x: exports_external.number(),
+  y: exports_external.number()
 });
 var GROUND_SYSTEM = `You locate UI elements on a mobile screen.
 Respond with ONLY JSON: {"x":0-1000,"y":0-1000} for the center of the best match.
@@ -97730,7 +98056,7 @@ class ActionNotFoundError extends Error {
 async function performAction(session, body) {
   let x = body.x;
   let y = body.y;
-  if ((body.id || body.label) && (body.kind === "tap" || body.kind === "input") && (x == null || y == null)) {
+  if ((body.id || body.label) && (body.kind === "tap" || body.kind === "input")) {
     const screen = await getScreen(session, { full: false });
     const elements = screen.elements ?? [];
     const match2 = body.id ? findElementById(elements, body.id) : findElementByLabel(elements, body.label ?? "");
@@ -97930,6 +98256,25 @@ async function executeScriptCase(deps) {
           latencyMs,
           detail: action.reason ?? `${assertion}: ${action.text}`
         });
+      } else if (action.type === "alert") {
+        await perform(deps.session, {
+          kind: "alert",
+          alertAction: action.alertAction === "dismiss" ? "dismiss" : "accept"
+        });
+        await clock.sleep(settleMs);
+        await deps.appendStep({
+          idx: stepIdx,
+          action: {
+            type: "alert",
+            alertAction: action.alertAction === "dismiss" ? "dismiss" : "accept",
+            reason: action.reason ?? "Replayed saved script alert",
+            thoughts: "Replaying the saved script without calling the AI agent."
+          },
+          screenshotUri: shot.path,
+          ok: true,
+          latencyMs,
+          detail: action.reason ?? action.alertAction ?? "accept"
+        });
       } else {
         const waitMs = Math.min(3000, Math.max(500, action.ms));
         await clock.sleep(waitMs);
@@ -98054,9 +98399,30 @@ async function executeAgentCase(deps) {
         recentActions.push(decision);
         recordedDecisions.push(decision);
         if (decision.type === "tap") {
-          const x = decision.x ?? 500;
-          const y = decision.y ?? 500;
-          await perform(deps.session, { kind: "tap", x, y });
+          const tapBody = { kind: "tap" };
+          if (decision.label)
+            tapBody.label = decision.label;
+          if (decision.id)
+            tapBody.id = decision.id;
+          if (!tapBody.label && !tapBody.id) {
+            tapBody.x = decision.x ?? 500;
+            tapBody.y = decision.y ?? 500;
+          }
+          await perform(deps.session, tapBody);
+          await clock.sleep(settleMs);
+          await deps.appendStep({
+            idx: stepIdx,
+            action: decision,
+            screenshotUri: shot.path,
+            ok: true,
+            latencyMs,
+            detail: decision.reason ?? null
+          });
+        } else if (decision.type === "alert") {
+          await perform(deps.session, {
+            kind: "alert",
+            alertAction: decision.alertAction === "dismiss" ? "dismiss" : "accept"
+          });
           await clock.sleep(settleMs);
           await deps.appendStep({
             idx: stepIdx,

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@yoqa/runner",
-	"version": "0.3.21",
+	"version": "0.3.23",
 	"description": "Yoqa local runner — Appium + catalog HTTP API (Bun). Start with yoqa serve or the yoqa-runner bin.",
 	"license": "UNLICENSED",
 	"type": "module",