gpt-driver-node 1.0.4 → 1.0.5

package/dist/index.cjs

@@ -1436,14 +1436,43 @@ ${"=".repeat(50)}`);
   }
   async getScreenshot(appiumSessionConfig, shouldScale = true) {
     globalLogger.debug("Capturing screenshot...");
-    const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/screenshot`);
-    const screenshotResponse = await axios.get(url);
-    let screenshot = await screenshotResponse.data.value;
-    if (appiumSessionConfig.platform === "iOS" && shouldScale) {
-      globalLogger.debug(`Resizing iOS screenshot to ${appiumSessionConfig.size.width}x${appiumSessionConfig.size.height}`);
-      const imageBuffer = Buffer.from(screenshot, "base64");
-      const transformedImage = await sharp(imageBuffer).resize(appiumSessionConfig.size.width, appiumSessionConfig.size.height).toBuffer();
-      screenshot = transformedImage.toString("base64");
+    let screenshot;
+    if (appiumSessionConfig.platform === "Android") {
+      try {
+        const { execSync } = await import('node:child_process');
+        let udid;
+        if (this.driver) {
+          if (this.driver.sessionId != null) {
+            const caps = this.driver.capabilities;
+            udid = caps["appium:udid"] || caps["udid"];
+          } else {
+            const driver = this.driver;
+            const capabilities = await driver.getCapabilities();
+            udid = capabilities.get("appium:udid") || capabilities.get("udid");
+          }
+        }
+        const deviceArg = udid ? `-s ${udid}` : "";
+        const buffer = execSync(`adb ${deviceArg} exec-out screencap -p`, {
+          encoding: "buffer",
+          maxBuffer: 50 * 1024 * 1024
+        });
+        screenshot = buffer.toString("base64");
+      } catch (e) {
+        globalLogger.warn("ADB screenshot failed, falling back to Appium screenshot");
+        const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/screenshot`);
+        const screenshotResponse = await axios.get(url);
+        screenshot = screenshotResponse.data.value;
+      }
+    } else {
+      const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/screenshot`);
+      const screenshotResponse = await axios.get(url);
+      screenshot = screenshotResponse.data.value;
+      if (shouldScale) {
+        globalLogger.debug(`Resizing iOS screenshot to ${appiumSessionConfig.size.width}x${appiumSessionConfig.size.height}`);
+        const imageBuffer = Buffer.from(screenshot, "base64");
+        const transformedImage = await sharp(imageBuffer).resize(appiumSessionConfig.size.width, appiumSessionConfig.size.height).toBuffer();
+        screenshot = transformedImage.toString("base64");
+      }
     }
     return screenshot;
   }

package/dist/index.mjs

@@ -1434,14 +1434,43 @@ ${"=".repeat(50)}`);
   }
   async getScreenshot(appiumSessionConfig, shouldScale = true) {
     globalLogger.debug("Capturing screenshot...");
-    const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/screenshot`);
-    const screenshotResponse = await axios.get(url);
-    let screenshot = await screenshotResponse.data.value;
-    if (appiumSessionConfig.platform === "iOS" && shouldScale) {
-      globalLogger.debug(`Resizing iOS screenshot to ${appiumSessionConfig.size.width}x${appiumSessionConfig.size.height}`);
-      const imageBuffer = Buffer.from(screenshot, "base64");
-      const transformedImage = await sharp(imageBuffer).resize(appiumSessionConfig.size.width, appiumSessionConfig.size.height).toBuffer();
-      screenshot = transformedImage.toString("base64");
+    let screenshot;
+    if (appiumSessionConfig.platform === "Android") {
+      try {
+        const { execSync } = await import('node:child_process');
+        let udid;
+        if (this.driver) {
+          if (this.driver.sessionId != null) {
+            const caps = this.driver.capabilities;
+            udid = caps["appium:udid"] || caps["udid"];
+          } else {
+            const driver = this.driver;
+            const capabilities = await driver.getCapabilities();
+            udid = capabilities.get("appium:udid") || capabilities.get("udid");
+          }
+        }
+        const deviceArg = udid ? `-s ${udid}` : "";
+        const buffer = execSync(`adb ${deviceArg} exec-out screencap -p`, {
+          encoding: "buffer",
+          maxBuffer: 50 * 1024 * 1024
+        });
+        screenshot = buffer.toString("base64");
+      } catch (e) {
+        globalLogger.warn("ADB screenshot failed, falling back to Appium screenshot");
+        const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/screenshot`);
+        const screenshotResponse = await axios.get(url);
+        screenshot = screenshotResponse.data.value;
+      }
+    } else {
+      const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/screenshot`);
+      const screenshotResponse = await axios.get(url);
+      screenshot = screenshotResponse.data.value;
+      if (shouldScale) {
+        globalLogger.debug(`Resizing iOS screenshot to ${appiumSessionConfig.size.width}x${appiumSessionConfig.size.height}`);
+        const imageBuffer = Buffer.from(screenshot, "base64");
+        const transformedImage = await sharp(imageBuffer).resize(appiumSessionConfig.size.width, appiumSessionConfig.size.height).toBuffer();
+        screenshot = transformedImage.toString("base64");
+      }
     }
     return screenshot;
   }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "gpt-driver-node",
-  "version": "1.0.4",
+  "version": "1.0.5",
   "main": "./dist/index.cjs",
   "module": "./dist/index.mjs",
   "types": "./dist/index.d.cts",