gpt-driver-node 1.0.4 → 1.0.6

package/dist/index.cjs

@@ -11,29 +11,6 @@ var winston = require('winston');
 var zod = require('zod');
 var crypto = require('node:crypto');
 
-const delay = async (milliseconds) => {
-  await new Promise((resolve) => setTimeout(resolve, milliseconds));
-};
-function buildUrl(base, extraPath) {
-  let baseUrl = base.toString();
-  if (baseUrl.endsWith("/")) {
-    baseUrl = baseUrl.slice(0, -1);
-  }
-  if (!extraPath.startsWith("/")) {
-    extraPath = "/" + extraPath;
-  }
-  return `${baseUrl}${extraPath}`;
-}
-const getImageDimensions = async (base64) => {
-  const base64Data = base64.replace(/^data:image\/\w+;base64,/, "");
-  const buffer = Buffer.from(base64Data, "base64");
-  const metadata = await sharp(buffer).metadata();
-  if (!metadata.width || !metadata.height) {
-    throw new Error("Unable to get image dimensions");
-  }
-  return { width: metadata.width, height: metadata.height };
-};
-
 const colors = {
   reset: "\x1B[0m",
   bold: "\x1B[1m",
@@ -95,6 +72,75 @@ ${logStyles.gray(stack)}` : logMessage;
   ]
 });
 
+const delay = async (milliseconds) => {
+  await new Promise((resolve) => setTimeout(resolve, milliseconds));
+};
+function isRetryableNetworkError(error) {
+  const retryableCodes = [
+    "ECONNRESET",
+    "ETIMEDOUT",
+    "ECONNREFUSED",
+    "ENOTFOUND",
+    "ENETUNREACH",
+    "EAI_AGAIN",
+    "EPIPE",
+    "ECONNABORTED"
+  ];
+  if (error.code && retryableCodes.includes(error.code)) {
+    return true;
+  }
+  if (error.message?.includes("socket disconnected") || error.message?.includes("TLS") || error.message?.includes("SSL")) {
+    return true;
+  }
+  if (error.code === "ECONNABORTED" && error.message?.includes("timeout")) {
+    return true;
+  }
+  if (error.response?.status >= 500) {
+    return true;
+  }
+  return false;
+}
+async function requestWithRetry(config, maxRetries = 3, baseDelayMs = 1e3) {
+  let lastError;
+  for (let attempt = 1; attempt <= maxRetries; attempt++) {
+    try {
+      const response = await axios.request(config);
+      return response.data;
+    } catch (error) {
+      lastError = error;
+      const isRetryable = isRetryableNetworkError(error);
+      if (!isRetryable || attempt === maxRetries) {
+        throw error;
+      }
+      const delayMs = baseDelayMs * Math.pow(2, attempt - 1);
+      globalLogger.warn(
+        `Request failed (attempt ${attempt}/${maxRetries}): ${error.message}. Retrying in ${delayMs}ms...`
+      );
+      await delay(delayMs);
+    }
+  }
+  throw lastError;
+}
+function buildUrl(base, extraPath) {
+  let baseUrl = base.toString();
+  if (baseUrl.endsWith("/")) {
+    baseUrl = baseUrl.slice(0, -1);
+  }
+  if (!extraPath.startsWith("/")) {
+    extraPath = "/" + extraPath;
+  }
+  return `${baseUrl}${extraPath}`;
+}
+const getImageDimensions = async (base64) => {
+  const base64Data = base64.replace(/^data:image\/\w+;base64,/, "");
+  const buffer = Buffer.from(base64Data, "base64");
+  const metadata = await sharp(buffer).metadata();
+  if (!metadata.width || !metadata.height) {
+    throw new Error("Unable to get image dimensions");
+  }
+  return { width: metadata.width, height: metadata.height };
+};
+
 const waitForStableScreen = async (getScreenshot, options = {}) => {
   const {
     maxTimeoutSec = 5,
@@ -1436,14 +1482,43 @@ ${"=".repeat(50)}`);
   }
   async getScreenshot(appiumSessionConfig, shouldScale = true) {
     globalLogger.debug("Capturing screenshot...");
-    const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${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;
   }
@@ -2248,7 +2323,6 @@ ${"=".repeat(50)}`);
       let conditionSucceeded = false;
       while (!conditionSucceeded) {
         let screenshot;
-        let originalScreenshotBase64 = null;
         if (!this.useGptDriverCloud) {
           const stabilityResult = await waitForStableScreen(
             () => this.getScreenshot(this.appiumSessionConfig)
@@ -2259,8 +2333,9 @@ ${"=".repeat(50)}`);
           }
         }
         globalLogger.info("Requesting next action from GPT Driver...");
-        const response = await axios.request(
-          {
+        let responseData;
+        try {
+          responseData = await requestWithRetry({
             url: `${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/execute`,
             method: "POST",
             data: {
@@ -2270,18 +2345,21 @@ ${"=".repeat(50)}`);
               caching_mode: cachingMode ?? this.cachingMode,
               step_number: this.step_number
             }
-          }
-        );
-        const executeStatus = response.data.status;
+          });
+        } catch (error) {
+          globalLogger.error(`GPT Driver request failed: ${error.message}`);
+          globalLogger.error(`Stack trace: ${error.stack}`);
+          throw error;
+        }
+        const executeStatus = responseData.status;
         if (executeStatus === "failed") {
-          const errorMessage = response.data?.commands?.at(0)?.data;
+          const errorMessage = responseData?.commands?.at(0)?.data;
           globalLogger.error(`Execution failed: ${errorMessage ?? "Unknown error"}`);
           throw new Error(errorMessage ?? "Execution failed");
         }
         conditionSucceeded = executeStatus !== "inProgress";
-        const executeResponse = response.data;
-        globalLogger.debug(`Received ${executeResponse.commands.length} command(s) to execute`);
-        for (const appiumCommand of executeResponse.commands) {
+        globalLogger.debug(`Received ${responseData.commands.length} command(s) to execute`);
+        for (const appiumCommand of responseData.commands) {
           await this.executeCommand(appiumCommand);
         }
       }

package/dist/index.mjs

@@ -9,29 +9,6 @@ import winston from 'winston';
 import { z } from 'zod';
 import crypto from 'node:crypto';
 
-const delay = async (milliseconds) => {
-  await new Promise((resolve) => setTimeout(resolve, milliseconds));
-};
-function buildUrl(base, extraPath) {
-  let baseUrl = base.toString();
-  if (baseUrl.endsWith("/")) {
-    baseUrl = baseUrl.slice(0, -1);
-  }
-  if (!extraPath.startsWith("/")) {
-    extraPath = "/" + extraPath;
-  }
-  return `${baseUrl}${extraPath}`;
-}
-const getImageDimensions = async (base64) => {
-  const base64Data = base64.replace(/^data:image\/\w+;base64,/, "");
-  const buffer = Buffer.from(base64Data, "base64");
-  const metadata = await sharp(buffer).metadata();
-  if (!metadata.width || !metadata.height) {
-    throw new Error("Unable to get image dimensions");
-  }
-  return { width: metadata.width, height: metadata.height };
-};
-
 const colors = {
   reset: "\x1B[0m",
   bold: "\x1B[1m",
@@ -93,6 +70,75 @@ ${logStyles.gray(stack)}` : logMessage;
   ]
 });
 
+const delay = async (milliseconds) => {
+  await new Promise((resolve) => setTimeout(resolve, milliseconds));
+};
+function isRetryableNetworkError(error) {
+  const retryableCodes = [
+    "ECONNRESET",
+    "ETIMEDOUT",
+    "ECONNREFUSED",
+    "ENOTFOUND",
+    "ENETUNREACH",
+    "EAI_AGAIN",
+    "EPIPE",
+    "ECONNABORTED"
+  ];
+  if (error.code && retryableCodes.includes(error.code)) {
+    return true;
+  }
+  if (error.message?.includes("socket disconnected") || error.message?.includes("TLS") || error.message?.includes("SSL")) {
+    return true;
+  }
+  if (error.code === "ECONNABORTED" && error.message?.includes("timeout")) {
+    return true;
+  }
+  if (error.response?.status >= 500) {
+    return true;
+  }
+  return false;
+}
+async function requestWithRetry(config, maxRetries = 3, baseDelayMs = 1e3) {
+  let lastError;
+  for (let attempt = 1; attempt <= maxRetries; attempt++) {
+    try {
+      const response = await axios.request(config);
+      return response.data;
+    } catch (error) {
+      lastError = error;
+      const isRetryable = isRetryableNetworkError(error);
+      if (!isRetryable || attempt === maxRetries) {
+        throw error;
+      }
+      const delayMs = baseDelayMs * Math.pow(2, attempt - 1);
+      globalLogger.warn(
+        `Request failed (attempt ${attempt}/${maxRetries}): ${error.message}. Retrying in ${delayMs}ms...`
+      );
+      await delay(delayMs);
+    }
+  }
+  throw lastError;
+}
+function buildUrl(base, extraPath) {
+  let baseUrl = base.toString();
+  if (baseUrl.endsWith("/")) {
+    baseUrl = baseUrl.slice(0, -1);
+  }
+  if (!extraPath.startsWith("/")) {
+    extraPath = "/" + extraPath;
+  }
+  return `${baseUrl}${extraPath}`;
+}
+const getImageDimensions = async (base64) => {
+  const base64Data = base64.replace(/^data:image\/\w+;base64,/, "");
+  const buffer = Buffer.from(base64Data, "base64");
+  const metadata = await sharp(buffer).metadata();
+  if (!metadata.width || !metadata.height) {
+    throw new Error("Unable to get image dimensions");
+  }
+  return { width: metadata.width, height: metadata.height };
+};
+
 const waitForStableScreen = async (getScreenshot, options = {}) => {
   const {
     maxTimeoutSec = 5,
@@ -1434,14 +1480,43 @@ ${"=".repeat(50)}`);
   }
   async getScreenshot(appiumSessionConfig, shouldScale = true) {
     globalLogger.debug("Capturing screenshot...");
-    const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${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;
   }
@@ -2246,7 +2321,6 @@ ${"=".repeat(50)}`);
       let conditionSucceeded = false;
       while (!conditionSucceeded) {
         let screenshot;
-        let originalScreenshotBase64 = null;
         if (!this.useGptDriverCloud) {
           const stabilityResult = await waitForStableScreen(
             () => this.getScreenshot(this.appiumSessionConfig)
@@ -2257,8 +2331,9 @@ ${"=".repeat(50)}`);
           }
         }
         globalLogger.info("Requesting next action from GPT Driver...");
-        const response = await axios.request(
-          {
+        let responseData;
+        try {
+          responseData = await requestWithRetry({
             url: `${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/execute`,
             method: "POST",
             data: {
@@ -2268,18 +2343,21 @@ ${"=".repeat(50)}`);
               caching_mode: cachingMode ?? this.cachingMode,
               step_number: this.step_number
             }
-          }
-        );
-        const executeStatus = response.data.status;
+          });
+        } catch (error) {
+          globalLogger.error(`GPT Driver request failed: ${error.message}`);
+          globalLogger.error(`Stack trace: ${error.stack}`);
+          throw error;
+        }
+        const executeStatus = responseData.status;
         if (executeStatus === "failed") {
-          const errorMessage = response.data?.commands?.at(0)?.data;
+          const errorMessage = responseData?.commands?.at(0)?.data;
           globalLogger.error(`Execution failed: ${errorMessage ?? "Unknown error"}`);
           throw new Error(errorMessage ?? "Execution failed");
         }
         conditionSucceeded = executeStatus !== "inProgress";
-        const executeResponse = response.data;
-        globalLogger.debug(`Received ${executeResponse.commands.length} command(s) to execute`);
-        for (const appiumCommand of executeResponse.commands) {
+        globalLogger.debug(`Received ${responseData.commands.length} command(s) to execute`);
+        for (const appiumCommand of responseData.commands) {
           await this.executeCommand(appiumCommand);
         }
       }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "gpt-driver-node",
-  "version": "1.0.4",
+  "version": "1.0.6",
   "main": "./dist/index.cjs",
   "module": "./dist/index.mjs",
   "types": "./dist/index.d.cts",