@midscene/ios 1.0.1-beta-20251104075048.0 → 1.0.1-beta-20251106111345.0

package/dist/es/bin.mjs

@@ -12082,8 +12082,17 @@ class IOSWebDriverClient extends WebDriverClient {
             });
             debugIOS(`Tapped at coordinates (${x}, ${y})`);
         } catch (error) {
-            debugIOS(`Failed to tap at (${x}, ${y}): ${error}`);
-            throw new Error(`Failed to tap at coordinates: ${error}`);
+            debugIOS(`New tap endpoint failed, trying legacy endpoint: ${error}`);
+            try {
+                await this.makeRequest('POST', `/session/${this.sessionId}/wda/tap/0`, {
+                    x,
+                    y
+                });
+                debugIOS(`Tapped at coordinates (${x}, ${y}) using legacy endpoint`);
+            } catch (fallbackError) {
+                debugIOS(`Failed to tap at (${x}, ${y}): ${fallbackError}`);
+                throw new Error(`Failed to tap at coordinates: ${fallbackError}`);
+            }
         }
     }
     async swipe(fromX, fromY, toX, toY, duration = 500) {
@@ -12156,13 +12165,35 @@ class IOSWebDriverClient extends WebDriverClient {
         debugIOS(`Triple tapped at coordinates (${x}, ${y})`);
     }
     async getScreenScale() {
-        var _screenResponse_value;
-        const screenResponse = await this.makeRequest('GET', '/wda/screen');
-        if (null == screenResponse ? void 0 : null == (_screenResponse_value = screenResponse.value) ? void 0 : _screenResponse_value.scale) {
-            debugIOS(`Got screen scale from WDA screen endpoint: ${screenResponse.value.scale}`);
-            return screenResponse.value.scale;
+        this.ensureSession();
+        try {
+            var _screenResponse_value;
+            const screenResponse = await this.makeRequest('GET', `/session/${this.sessionId}/wda/screen`);
+            if (null == screenResponse ? void 0 : null == (_screenResponse_value = screenResponse.value) ? void 0 : _screenResponse_value.scale) {
+                debugIOS(`Got screen scale from WDA screen endpoint: ${screenResponse.value.scale}`);
+                return screenResponse.value.scale;
+            }
+        } catch (error) {
+            debugIOS(`Failed to get screen scale from /wda/screen: ${error}`);
+        }
+        try {
+            debugIOS('Calculating screen scale from screenshot and window size');
+            const [screenshotBase64, windowSize] = await Promise.all([
+                this.takeScreenshot(),
+                this.getWindowSize()
+            ]);
+            const { jimpFromBase64 } = await import("@midscene/shared/img");
+            const screenshotImg = await jimpFromBase64(screenshotBase64);
+            const screenshotWidth = screenshotImg.bitmap.width;
+            const screenshotHeight = screenshotImg.bitmap.height;
+            const scale = Math.max(screenshotWidth, screenshotHeight) / Math.max(windowSize.width, windowSize.height);
+            const roundedScale = Math.round(scale);
+            debugIOS(`Calculated screen scale: ${roundedScale} (screenshot: ${screenshotWidth}x${screenshotHeight}, window: ${windowSize.width}x${windowSize.height})`);
+            return roundedScale;
+        } catch (error) {
+            debugIOS(`Failed to calculate screen scale: ${error}`);
         }
-        debugIOS('No screen scale found in WDA screen response');
+        debugIOS('No screen scale found');
         return null;
     }
     async createSession(capabilities) {
@@ -12283,22 +12314,6 @@ class device_IOSDevice {
             defineActionKeyboardPress(async (param)=>{
                 await this.pressKey(param.keyName);
             }),
-            defineAction({
-                name: 'IOSHomeButton',
-                description: 'Trigger the system "home" operation on iOS devices',
-                paramSchema: z.object({}),
-                call: async ()=>{
-                    await this.home();
-                }
-            }),
-            defineAction({
-                name: 'IOSAppSwitcher',
-                description: 'Trigger the system "app switcher" operation on iOS devices',
-                paramSchema: z.object({}),
-                call: async ()=>{
-                    await this.appSwitcher();
-                }
-            }),
             defineAction({
                 name: 'IOSLongPress',
                 description: 'Trigger a long press on the screen at specified coordinates on iOS devices',
@@ -12694,6 +12709,7 @@ ScreenSize: ${size.width}x${size.height} (DPR: ${size.scale})
     async openUrlViaSafari(url) {
         try {
             debugDevice(`Opening URL via Safari: ${url}`);
+            await this.wdaBackend.terminateApp('com.apple.mobilesafari');
             await this.wdaBackend.launchApp('com.apple.mobilesafari');
             await sleep(2000);
             await this.typeText(url);
@@ -12771,6 +12787,22 @@ const createPlatformActions = (device)=>({
             call: async (param)=>{
                 await device.launch(param);
             }
+        }),
+        IOSHomeButton: defineAction({
+            name: 'IOSHomeButton',
+            description: 'Trigger the system "home" operation on iOS devices',
+            paramSchema: z["void"]().describe('No parameters required'),
+            call: async ()=>{
+                await device.home();
+            }
+        }),
+        IOSAppSwitcher: defineAction({
+            name: 'IOSAppSwitcher',
+            description: 'Trigger the system "app switcher" operation on iOS devices',
+            paramSchema: z["void"]().describe('No parameters required'),
+            call: async ()=>{
+                await device.appSwitcher();
+            }
         })
     });
 promisify(exec);
@@ -12787,10 +12819,16 @@ function agent_define_property(obj, key, value) {
 }
 getDebug('ios:agent');
 class IOSAgent extends Agent {
+    createActionWrapper(name) {
+        const action = this.wrapActionInActionSpace(name);
+        return (...args)=>action(args[0]);
+    }
     constructor(device, opts){
-        super(device, opts), agent_define_property(this, "launch", void 0), agent_define_property(this, "runWdaRequest", void 0);
-        this.launch = this.wrapActionInActionSpace('Launch');
-        this.runWdaRequest = this.wrapActionInActionSpace('RunWdaRequest');
+        super(device, opts), agent_define_property(this, "launch", void 0), agent_define_property(this, "runWdaRequest", void 0), agent_define_property(this, "home", void 0), agent_define_property(this, "appSwitcher", void 0);
+        this.launch = this.createActionWrapper('Launch');
+        this.runWdaRequest = this.createActionWrapper('RunWdaRequest');
+        this.home = this.createActionWrapper('IOSHomeButton');
+        this.appSwitcher = this.createActionWrapper('IOSAppSwitcher');
     }
 }
 async function isPortAvailable(port) {

package/dist/es/index.mjs

@@ -233,8 +233,17 @@ class IOSWebDriverClient extends WebDriverClient {
             });
             debugIOS(`Tapped at coordinates (${x}, ${y})`);
         } catch (error) {
-            debugIOS(`Failed to tap at (${x}, ${y}): ${error}`);
-            throw new Error(`Failed to tap at coordinates: ${error}`);
+            debugIOS(`New tap endpoint failed, trying legacy endpoint: ${error}`);
+            try {
+                await this.makeRequest('POST', `/session/${this.sessionId}/wda/tap/0`, {
+                    x,
+                    y
+                });
+                debugIOS(`Tapped at coordinates (${x}, ${y}) using legacy endpoint`);
+            } catch (fallbackError) {
+                debugIOS(`Failed to tap at (${x}, ${y}): ${fallbackError}`);
+                throw new Error(`Failed to tap at coordinates: ${fallbackError}`);
+            }
         }
     }
     async swipe(fromX, fromY, toX, toY, duration = 500) {
@@ -307,13 +316,35 @@ class IOSWebDriverClient extends WebDriverClient {
         debugIOS(`Triple tapped at coordinates (${x}, ${y})`);
     }
     async getScreenScale() {
-        var _screenResponse_value;
-        const screenResponse = await this.makeRequest('GET', '/wda/screen');
-        if (null == screenResponse ? void 0 : null == (_screenResponse_value = screenResponse.value) ? void 0 : _screenResponse_value.scale) {
-            debugIOS(`Got screen scale from WDA screen endpoint: ${screenResponse.value.scale}`);
-            return screenResponse.value.scale;
+        this.ensureSession();
+        try {
+            var _screenResponse_value;
+            const screenResponse = await this.makeRequest('GET', `/session/${this.sessionId}/wda/screen`);
+            if (null == screenResponse ? void 0 : null == (_screenResponse_value = screenResponse.value) ? void 0 : _screenResponse_value.scale) {
+                debugIOS(`Got screen scale from WDA screen endpoint: ${screenResponse.value.scale}`);
+                return screenResponse.value.scale;
+            }
+        } catch (error) {
+            debugIOS(`Failed to get screen scale from /wda/screen: ${error}`);
         }
-        debugIOS('No screen scale found in WDA screen response');
+        try {
+            debugIOS('Calculating screen scale from screenshot and window size');
+            const [screenshotBase64, windowSize] = await Promise.all([
+                this.takeScreenshot(),
+                this.getWindowSize()
+            ]);
+            const { jimpFromBase64 } = await import("@midscene/shared/img");
+            const screenshotImg = await jimpFromBase64(screenshotBase64);
+            const screenshotWidth = screenshotImg.bitmap.width;
+            const screenshotHeight = screenshotImg.bitmap.height;
+            const scale = Math.max(screenshotWidth, screenshotHeight) / Math.max(windowSize.width, windowSize.height);
+            const roundedScale = Math.round(scale);
+            debugIOS(`Calculated screen scale: ${roundedScale} (screenshot: ${screenshotWidth}x${screenshotHeight}, window: ${windowSize.width}x${windowSize.height})`);
+            return roundedScale;
+        } catch (error) {
+            debugIOS(`Failed to calculate screen scale: ${error}`);
+        }
+        debugIOS('No screen scale found');
         return null;
     }
     async createSession(capabilities) {
@@ -434,22 +465,6 @@ class IOSDevice {
             defineActionKeyboardPress(async (param)=>{
                 await this.pressKey(param.keyName);
             }),
-            defineAction({
-                name: 'IOSHomeButton',
-                description: 'Trigger the system "home" operation on iOS devices',
-                paramSchema: z.object({}),
-                call: async ()=>{
-                    await this.home();
-                }
-            }),
-            defineAction({
-                name: 'IOSAppSwitcher',
-                description: 'Trigger the system "app switcher" operation on iOS devices',
-                paramSchema: z.object({}),
-                call: async ()=>{
-                    await this.appSwitcher();
-                }
-            }),
             defineAction({
                 name: 'IOSLongPress',
                 description: 'Trigger a long press on the screen at specified coordinates on iOS devices',
@@ -845,6 +860,7 @@ ScreenSize: ${size.width}x${size.height} (DPR: ${size.scale})
     async openUrlViaSafari(url) {
         try {
             debugDevice(`Opening URL via Safari: ${url}`);
+            await this.wdaBackend.terminateApp('com.apple.mobilesafari');
             await this.wdaBackend.launchApp('com.apple.mobilesafari');
             await sleep(2000);
             await this.typeText(url);
@@ -922,6 +938,22 @@ const createPlatformActions = (device)=>({
             call: async (param)=>{
                 await device.launch(param);
             }
+        }),
+        IOSHomeButton: defineAction({
+            name: 'IOSHomeButton',
+            description: 'Trigger the system "home" operation on iOS devices',
+            paramSchema: z["void"]().describe('No parameters required'),
+            call: async ()=>{
+                await device.home();
+            }
+        }),
+        IOSAppSwitcher: defineAction({
+            name: 'IOSAppSwitcher',
+            description: 'Trigger the system "app switcher" operation on iOS devices',
+            paramSchema: z["void"]().describe('No parameters required'),
+            call: async ()=>{
+                await device.appSwitcher();
+            }
         })
     });
 const execAsync = promisify(exec);
@@ -982,10 +1014,16 @@ function agent_define_property(obj, key, value) {
 }
 const debugAgent = getDebug('ios:agent');
 class IOSAgent extends Agent {
+    createActionWrapper(name) {
+        const action = this.wrapActionInActionSpace(name);
+        return (...args)=>action(args[0]);
+    }
     constructor(device, opts){
-        super(device, opts), agent_define_property(this, "launch", void 0), agent_define_property(this, "runWdaRequest", void 0);
-        this.launch = this.wrapActionInActionSpace('Launch');
-        this.runWdaRequest = this.wrapActionInActionSpace('RunWdaRequest');
+        super(device, opts), agent_define_property(this, "launch", void 0), agent_define_property(this, "runWdaRequest", void 0), agent_define_property(this, "home", void 0), agent_define_property(this, "appSwitcher", void 0);
+        this.launch = this.createActionWrapper('Launch');
+        this.runWdaRequest = this.createActionWrapper('RunWdaRequest');
+        this.home = this.createActionWrapper('IOSHomeButton');
+        this.appSwitcher = this.createActionWrapper('IOSAppSwitcher');
     }
 }
 async function agentFromWebDriverAgent(opts) {

package/dist/lib/bin.js

@@ -11421,6 +11421,12 @@ object-assign
         "use strict";
         module.exports = require("zlib");
     },
+    "@midscene/shared/img": function(module) {
+        "use strict";
+        module.exports = import("@midscene/shared/img").then(function(module) {
+            return module;
+        });
+    },
     open: function(module) {
         "use strict";
         module.exports = import("open").then(function(module) {
@@ -12220,8 +12226,17 @@ var __webpack_exports__ = {};
                 });
                 debugIOS(`Tapped at coordinates (${x}, ${y})`);
             } catch (error) {
-                debugIOS(`Failed to tap at (${x}, ${y}): ${error}`);
-                throw new Error(`Failed to tap at coordinates: ${error}`);
+                debugIOS(`New tap endpoint failed, trying legacy endpoint: ${error}`);
+                try {
+                    await this.makeRequest('POST', `/session/${this.sessionId}/wda/tap/0`, {
+                        x,
+                        y
+                    });
+                    debugIOS(`Tapped at coordinates (${x}, ${y}) using legacy endpoint`);
+                } catch (fallbackError) {
+                    debugIOS(`Failed to tap at (${x}, ${y}): ${fallbackError}`);
+                    throw new Error(`Failed to tap at coordinates: ${fallbackError}`);
+                }
             }
         }
         async swipe(fromX, fromY, toX, toY, duration = 500) {
@@ -12294,13 +12309,35 @@ var __webpack_exports__ = {};
             debugIOS(`Triple tapped at coordinates (${x}, ${y})`);
         }
         async getScreenScale() {
-            var _screenResponse_value;
-            const screenResponse = await this.makeRequest('GET', '/wda/screen');
-            if (null == screenResponse ? void 0 : null == (_screenResponse_value = screenResponse.value) ? void 0 : _screenResponse_value.scale) {
-                debugIOS(`Got screen scale from WDA screen endpoint: ${screenResponse.value.scale}`);
-                return screenResponse.value.scale;
+            this.ensureSession();
+            try {
+                var _screenResponse_value;
+                const screenResponse = await this.makeRequest('GET', `/session/${this.sessionId}/wda/screen`);
+                if (null == screenResponse ? void 0 : null == (_screenResponse_value = screenResponse.value) ? void 0 : _screenResponse_value.scale) {
+                    debugIOS(`Got screen scale from WDA screen endpoint: ${screenResponse.value.scale}`);
+                    return screenResponse.value.scale;
+                }
+            } catch (error) {
+                debugIOS(`Failed to get screen scale from /wda/screen: ${error}`);
+            }
+            try {
+                debugIOS('Calculating screen scale from screenshot and window size');
+                const [screenshotBase64, windowSize] = await Promise.all([
+                    this.takeScreenshot(),
+                    this.getWindowSize()
+                ]);
+                const { jimpFromBase64 } = await Promise.resolve().then(__webpack_require__.bind(__webpack_require__, "@midscene/shared/img"));
+                const screenshotImg = await jimpFromBase64(screenshotBase64);
+                const screenshotWidth = screenshotImg.bitmap.width;
+                const screenshotHeight = screenshotImg.bitmap.height;
+                const scale = Math.max(screenshotWidth, screenshotHeight) / Math.max(windowSize.width, windowSize.height);
+                const roundedScale = Math.round(scale);
+                debugIOS(`Calculated screen scale: ${roundedScale} (screenshot: ${screenshotWidth}x${screenshotHeight}, window: ${windowSize.width}x${windowSize.height})`);
+                return roundedScale;
+            } catch (error) {
+                debugIOS(`Failed to calculate screen scale: ${error}`);
             }
-            debugIOS('No screen scale found in WDA screen response');
+            debugIOS('No screen scale found');
             return null;
         }
         async createSession(capabilities) {
@@ -12421,22 +12458,6 @@ var __webpack_exports__ = {};
                 (0, device_namespaceObject.defineActionKeyboardPress)(async (param)=>{
                     await this.pressKey(param.keyName);
                 }),
-                (0, device_namespaceObject.defineAction)({
-                    name: 'IOSHomeButton',
-                    description: 'Trigger the system "home" operation on iOS devices',
-                    paramSchema: core_namespaceObject.z.object({}),
-                    call: async ()=>{
-                        await this.home();
-                    }
-                }),
-                (0, device_namespaceObject.defineAction)({
-                    name: 'IOSAppSwitcher',
-                    description: 'Trigger the system "app switcher" operation on iOS devices',
-                    paramSchema: core_namespaceObject.z.object({}),
-                    call: async ()=>{
-                        await this.appSwitcher();
-                    }
-                }),
                 (0, device_namespaceObject.defineAction)({
                     name: 'IOSLongPress',
                     description: 'Trigger a long press on the screen at specified coordinates on iOS devices',
@@ -12832,6 +12853,7 @@ ScreenSize: ${size.width}x${size.height} (DPR: ${size.scale})
         async openUrlViaSafari(url) {
             try {
                 debugDevice(`Opening URL via Safari: ${url}`);
+                await this.wdaBackend.terminateApp('com.apple.mobilesafari');
                 await this.wdaBackend.launchApp('com.apple.mobilesafari');
                 await (0, utils_namespaceObject.sleep)(2000);
                 await this.typeText(url);
@@ -12909,6 +12931,22 @@ ScreenSize: ${size.width}x${size.height} (DPR: ${size.scale})
                 call: async (param)=>{
                     await device.launch(param);
                 }
+            }),
+            IOSHomeButton: (0, device_namespaceObject.defineAction)({
+                name: 'IOSHomeButton',
+                description: 'Trigger the system "home" operation on iOS devices',
+                paramSchema: core_namespaceObject.z["void"]().describe('No parameters required'),
+                call: async ()=>{
+                    await device.home();
+                }
+            }),
+            IOSAppSwitcher: (0, device_namespaceObject.defineAction)({
+                name: 'IOSAppSwitcher',
+                description: 'Trigger the system "app switcher" operation on iOS devices',
+                paramSchema: core_namespaceObject.z["void"]().describe('No parameters required'),
+                call: async ()=>{
+                    await device.appSwitcher();
+                }
             })
         });
     require("node:os");
@@ -12927,10 +12965,16 @@ ScreenSize: ${size.width}x${size.height} (DPR: ${size.scale})
     }
     (0, logger_namespaceObject.getDebug)('ios:agent');
     class IOSAgent extends agent_namespaceObject.Agent {
+        createActionWrapper(name) {
+            const action = this.wrapActionInActionSpace(name);
+            return (...args)=>action(args[0]);
+        }
         constructor(device, opts){
-            super(device, opts), agent_define_property(this, "launch", void 0), agent_define_property(this, "runWdaRequest", void 0);
-            this.launch = this.wrapActionInActionSpace('Launch');
-            this.runWdaRequest = this.wrapActionInActionSpace('RunWdaRequest');
+            super(device, opts), agent_define_property(this, "launch", void 0), agent_define_property(this, "runWdaRequest", void 0), agent_define_property(this, "home", void 0), agent_define_property(this, "appSwitcher", void 0);
+            this.launch = this.createActionWrapper('Launch');
+            this.runWdaRequest = this.createActionWrapper('RunWdaRequest');
+            this.home = this.createActionWrapper('IOSHomeButton');
+            this.appSwitcher = this.createActionWrapper('IOSAppSwitcher');
         }
     }
     async function isPortAvailable(port) {