gpt-driver-node 1.0.0-alpha.12 → 1.0.0-alpha.14

package/dist/index.cjs

@@ -1,6 +1,7 @@
 'use strict';
 
 var axios = require('axios');
+var webdriverio = require('webdriverio');
 var sharp = require('sharp');
 
 const delay = async (milliseconds) => {
@@ -24,6 +25,9 @@ class GptDriver {
   appiumSessionConfig;
   driver;
   appiumSessionStarted;
+  useGptDriverCloud;
+  gptDriverCloudPlatform;
+  buildId;
   /**
    * Creates an instance of the GptDriver class.
    *
@@ -44,9 +48,18 @@ class GptDriver {
    */
   constructor(config) {
     this.apiKey = config.apiKey;
+    this.buildId = config.buildId;
+    this.useGptDriverCloud = config.useGptDriverCloud;
     this.gptDriverBaseUrl = "https://api.mobileboost.io";
-    this.initializeDriver(config);
-    this.initializeAppiumConfig(config);
+    if (config.useGptDriverCloud) {
+      if (config.serverConfig?.device?.platform == null) {
+        throw new Error("Platform is missing. Please specify the platform when using GPTDriver Cloud.");
+      }
+      this.gptDriverCloudPlatform = config.serverConfig.device?.platform;
+    } else {
+      this.initializeDriver(config);
+      this.initializeAppiumConfig(config);
+    }
   }
   initializeDriver(config) {
     if (config.driver) {
@@ -78,46 +91,48 @@ class GptDriver {
    */
   async startSession() {
     console.log(">> Starting session...");
-    if (this.driver) {
-      let platform;
-      let platformVersion;
-      let deviceName;
-      let sessionId;
-      if (this.driver.sessionId == null) {
-        const driver = this.driver;
-        const capabilities = await driver.getCapabilities();
-        platform = capabilities.get("platformName");
-        platformVersion = capabilities.get("platformVersion") ?? this.appiumSessionConfig?.platformVersion;
-        deviceName = this.appiumSessionConfig?.deviceName ?? capabilities.get("deviceName");
-        const session = await driver.getSession();
-        sessionId = session.getId();
+    if (!this.useGptDriverCloud) {
+      if (this.driver) {
+        let platform;
+        let platformVersion;
+        let deviceName;
+        let sessionId;
+        if (this.driver.sessionId == null) {
+          const driver = this.driver;
+          const capabilities = await driver.getCapabilities();
+          platform = capabilities.get("platformName");
+          platformVersion = capabilities.get("platformVersion") ?? this.appiumSessionConfig?.platformVersion;
+          deviceName = this.appiumSessionConfig?.deviceName ?? capabilities.get("deviceName");
+          const session = await driver.getSession();
+          sessionId = session.getId();
+        } else {
+          const driver = this.driver;
+          platform = driver.capabilities["appium:platformName"] ?? driver.capabilities["platformName"];
+          platformVersion = driver.capabilities["appium:platformVersion"] ?? driver.capabilities["platformVersion"];
+          deviceName = this.appiumSessionConfig?.deviceName ?? driver.capabilities["appium:deviceName"] ?? driver.capabilities["deviceName"];
+          sessionId = driver.sessionId;
+        }
+        this.appiumSessionConfig = {
+          ...this.appiumSessionConfig,
+          id: sessionId,
+          platform,
+          platformVersion,
+          deviceName
+        };
       } else {
-        const driver = this.driver;
-        platform = driver.capabilities["appium:platformName"] ?? driver.capabilities["platformName"];
-        platformVersion = driver.capabilities["appium:platformVersion"] ?? driver.capabilities["platformVersion"];
-        deviceName = this.appiumSessionConfig?.deviceName ?? driver.capabilities["appium:deviceName"] ?? driver.capabilities["deviceName"];
-        sessionId = driver.sessionId;
+        this.appiumSessionConfig.id = await this.createSession();
       }
-      this.appiumSessionConfig = {
-        ...this.appiumSessionConfig,
-        id: sessionId,
-        platform,
-        platformVersion,
-        deviceName
+      const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/window/rect`);
+      const rectResponse = await axios.get(
+        url
+      );
+      this.appiumSessionConfig.size = {
+        width: rectResponse.data.value.width,
+        height: rectResponse.data.value.height
       };
-    } else {
-      this.appiumSessionConfig.id = await this.createSession();
+      this.appiumSessionStarted = true;
     }
     await this.createGptDriverSession();
-    const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/window/rect`);
-    const rectResponse = await axios.get(
-      url
-    );
-    this.appiumSessionConfig.size = {
-      width: rectResponse.data.value.width,
-      height: rectResponse.data.value.height
-    };
-    this.appiumSessionStarted = true;
     console.log(`>> Session created. Monitor execution at: ${this.getSessionLink()}`);
   }
   async createSession() {
@@ -143,15 +158,28 @@ class GptDriver {
       `${this.gptDriverBaseUrl}/sessions/create`,
       {
         api_key: this.apiKey,
-        appium_session_id: this.appiumSessionConfig.id,
+        appium_session_id: this.appiumSessionConfig?.id,
         device_config: {
-          platform: this.appiumSessionConfig.platform,
-          device: this.appiumSessionConfig.deviceName,
-          os: this.appiumSessionConfig.platformVersion
-        }
+          platform: this.appiumSessionConfig?.platform ?? this.gptDriverCloudPlatform,
+          device: this.appiumSessionConfig?.deviceName,
+          os: this.appiumSessionConfig?.platformVersion
+        },
+        use_internal_virtual_device: this.useGptDriverCloud,
+        build_id: this.buildId
       }
     );
     this.gptDriverSessionId = response.data.sessionId;
+    if (this.useGptDriverCloud) {
+      const parsedUrl = new URL(response.data.appiumServerUrl);
+      this.driver = await webdriverio.attach({
+        options: {
+          hostname: parsedUrl.hostname,
+          path: parsedUrl.pathname
+        },
+        sessionId: response.data.appiumSessionId
+      });
+      this.appiumSessionStarted = true;
+    }
   }
   getSessionLink() {
     return `https://app.mobileboost.io/gpt-driver/sessions/${this.gptDriverSessionId}`;
@@ -168,17 +196,19 @@ class GptDriver {
    * @throws {Error} If the request to stop the session fails.
    */
   async setSessionStatus(status) {
-    console.log(">> Stopping session...");
-    await axios.post(
-      `${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/stop`,
-      {
-        api_key: this.apiKey,
-        status
-      }
-    );
-    console.log(">> Session stopped.");
-    this.appiumSessionStarted = false;
-    this.gptDriverSessionId = void 0;
+    if (this.gptDriverSessionId) {
+      console.log(">> Stopping session...");
+      await axios.post(
+        `${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/stop`,
+        {
+          api_key: this.apiKey,
+          status
+        }
+      );
+      console.log(">> Session stopped.");
+      this.appiumSessionStarted = false;
+      this.gptDriverSessionId = void 0;
+    }
   }
   /**
    * Executes a specified command within the WebDriver session, optionally using an Appium handler.
@@ -228,10 +258,15 @@ class GptDriver {
     if (!this.appiumSessionStarted) {
       await this.startSession();
     }
-    console.log(">> Asserting:", assertion);
-    const results = await this.checkBulk([assertion]);
-    if (!Object.values(results).at(0)) {
-      throw new Error(`Failed assertion: ${assertion}`);
+    try {
+      const results = await this.checkBulk([assertion]);
+      if (!Object.values(results).at(0)) {
+        await this.setSessionStatus("failed");
+        throw new Error(`Failed assertion: ${assertion}`);
+      }
+    } catch (e) {
+      await this.setSessionStatus("failed");
+      throw e;
     }
   }
   /**
@@ -247,16 +282,21 @@ class GptDriver {
     if (!this.appiumSessionStarted) {
       await this.startSession();
     }
-    console.log(">> Asserting:", assertions);
-    const results = await this.checkBulk(assertions);
-    const failedAssertions = Object.values(results).reduce((prev, current, currentIndex) => {
-      if (!current) {
-        return [...prev, assertions.at(currentIndex)];
+    try {
+      const results = await this.checkBulk(assertions);
+      const failedAssertions = Object.values(results).reduce((prev, current, currentIndex) => {
+        if (!current) {
+          return [...prev, assertions.at(currentIndex)];
+        }
+        return prev;
+      }, []);
+      if (failedAssertions.length > 0) {
+        await this.setSessionStatus("failed");
+        throw new Error(`Failed assertions: ${failedAssertions.join(", ")}`);
       }
-      return prev;
-    }, []);
-    if (failedAssertions.length > 0) {
-      throw new Error(`Failed assertions: ${failedAssertions.join(", ")}`);
+    } catch (e) {
+      await this.setSessionStatus("failed");
+      throw e;
     }
   }
   /**
@@ -273,17 +313,25 @@ class GptDriver {
       await this.startSession();
     }
     console.log(">> Checking:", conditions);
-    const screenshot = await this.getScreenshot(this.appiumSessionConfig);
-    const response = await axios.post(
-      `${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/assert`,
-      {
-        api_key: this.apiKey,
-        base64_screenshot: screenshot,
-        assertions: conditions,
-        command: `Assert: ${JSON.stringify(conditions)}`
+    try {
+      let screenshot;
+      if (!this.useGptDriverCloud) {
+        screenshot = await this.getScreenshot(this.appiumSessionConfig);
       }
-    );
-    return response.data.results;
+      const response = await axios.post(
+        `${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/assert`,
+        {
+          api_key: this.apiKey,
+          base64_screenshot: screenshot,
+          assertions: conditions,
+          command: `Assert: ${JSON.stringify(conditions)}`
+        }
+      );
+      return response.data.results;
+    } catch (e) {
+      await this.setSessionStatus("failed");
+      throw e;
+    }
   }
   /**
    * Extracts specified information using the GPTDriver.
@@ -302,7 +350,10 @@ class GptDriver {
       await this.startSession();
     }
     console.log(">> Extracting:", extractions);
-    const screenshot = await this.getScreenshot(this.appiumSessionConfig);
+    let screenshot;
+    if (!this.useGptDriverCloud) {
+      screenshot = await this.getScreenshot(this.appiumSessionConfig);
+    }
     const response = await axios.post(
       `${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/extract`,
       {
@@ -318,7 +369,10 @@ class GptDriver {
     try {
       let conditionSucceeded = false;
       while (!conditionSucceeded) {
-        const screenshot = await this.getScreenshot(this.appiumSessionConfig);
+        let screenshot;
+        if (!this.useGptDriverCloud) {
+          screenshot = await this.getScreenshot(this.appiumSessionConfig);
+        }
         console.log(">> Asking GTP Driver for next action...");
         const response = await axios.request(
           {
@@ -351,10 +405,10 @@ class GptDriver {
     }
   }
   async executeCommand(command) {
-    const firstAction = command.data.actions?.at(0);
+    const firstAction = command?.data?.actions?.at(0);
     if (firstAction?.type === "pause" && firstAction.duration != null) {
       await delay(firstAction * 1e3);
-    } else {
+    } else if (!this.useGptDriverCloud) {
       const parsedUrl = new URL(command.url);
       parsedUrl.protocol = this.appiumSessionConfig.serverUrl.protocol;
       parsedUrl.host = this.appiumSessionConfig.serverUrl.host;

package/dist/index.d.cts

@@ -17,6 +17,8 @@ interface GptDriverConfig {
         device?: ServerSessionInitConfig;
         url?: URL | string;
     };
+    useGptDriverCloud?: boolean;
+    buildId?: string;
 }
 
 declare class GptDriver {
@@ -26,6 +28,9 @@ declare class GptDriver {
     private appiumSessionConfig?;
     private driver?;
     private appiumSessionStarted?;
+    private useGptDriverCloud?;
+    private gptDriverCloudPlatform?;
+    private buildId?;
     /**
      * Creates an instance of the GptDriver class.
      *

package/dist/index.mjs

@@ -1,4 +1,5 @@
 import axios from 'axios';
+import { attach } from 'webdriverio';
 import sharp from 'sharp';
 
 const delay = async (milliseconds) => {
@@ -22,6 +23,9 @@ class GptDriver {
   appiumSessionConfig;
   driver;
   appiumSessionStarted;
+  useGptDriverCloud;
+  gptDriverCloudPlatform;
+  buildId;
   /**
    * Creates an instance of the GptDriver class.
    *
@@ -42,9 +46,18 @@ class GptDriver {
    */
   constructor(config) {
     this.apiKey = config.apiKey;
+    this.buildId = config.buildId;
+    this.useGptDriverCloud = config.useGptDriverCloud;
     this.gptDriverBaseUrl = "https://api.mobileboost.io";
-    this.initializeDriver(config);
-    this.initializeAppiumConfig(config);
+    if (config.useGptDriverCloud) {
+      if (config.serverConfig?.device?.platform == null) {
+        throw new Error("Platform is missing. Please specify the platform when using GPTDriver Cloud.");
+      }
+      this.gptDriverCloudPlatform = config.serverConfig.device?.platform;
+    } else {
+      this.initializeDriver(config);
+      this.initializeAppiumConfig(config);
+    }
   }
   initializeDriver(config) {
     if (config.driver) {
@@ -76,46 +89,48 @@ class GptDriver {
    */
   async startSession() {
     console.log(">> Starting session...");
-    if (this.driver) {
-      let platform;
-      let platformVersion;
-      let deviceName;
-      let sessionId;
-      if (this.driver.sessionId == null) {
-        const driver = this.driver;
-        const capabilities = await driver.getCapabilities();
-        platform = capabilities.get("platformName");
-        platformVersion = capabilities.get("platformVersion") ?? this.appiumSessionConfig?.platformVersion;
-        deviceName = this.appiumSessionConfig?.deviceName ?? capabilities.get("deviceName");
-        const session = await driver.getSession();
-        sessionId = session.getId();
+    if (!this.useGptDriverCloud) {
+      if (this.driver) {
+        let platform;
+        let platformVersion;
+        let deviceName;
+        let sessionId;
+        if (this.driver.sessionId == null) {
+          const driver = this.driver;
+          const capabilities = await driver.getCapabilities();
+          platform = capabilities.get("platformName");
+          platformVersion = capabilities.get("platformVersion") ?? this.appiumSessionConfig?.platformVersion;
+          deviceName = this.appiumSessionConfig?.deviceName ?? capabilities.get("deviceName");
+          const session = await driver.getSession();
+          sessionId = session.getId();
+        } else {
+          const driver = this.driver;
+          platform = driver.capabilities["appium:platformName"] ?? driver.capabilities["platformName"];
+          platformVersion = driver.capabilities["appium:platformVersion"] ?? driver.capabilities["platformVersion"];
+          deviceName = this.appiumSessionConfig?.deviceName ?? driver.capabilities["appium:deviceName"] ?? driver.capabilities["deviceName"];
+          sessionId = driver.sessionId;
+        }
+        this.appiumSessionConfig = {
+          ...this.appiumSessionConfig,
+          id: sessionId,
+          platform,
+          platformVersion,
+          deviceName
+        };
       } else {
-        const driver = this.driver;
-        platform = driver.capabilities["appium:platformName"] ?? driver.capabilities["platformName"];
-        platformVersion = driver.capabilities["appium:platformVersion"] ?? driver.capabilities["platformVersion"];
-        deviceName = this.appiumSessionConfig?.deviceName ?? driver.capabilities["appium:deviceName"] ?? driver.capabilities["deviceName"];
-        sessionId = driver.sessionId;
+        this.appiumSessionConfig.id = await this.createSession();
       }
-      this.appiumSessionConfig = {
-        ...this.appiumSessionConfig,
-        id: sessionId,
-        platform,
-        platformVersion,
-        deviceName
+      const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/window/rect`);
+      const rectResponse = await axios.get(
+        url
+      );
+      this.appiumSessionConfig.size = {
+        width: rectResponse.data.value.width,
+        height: rectResponse.data.value.height
       };
-    } else {
-      this.appiumSessionConfig.id = await this.createSession();
+      this.appiumSessionStarted = true;
     }
     await this.createGptDriverSession();
-    const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/window/rect`);
-    const rectResponse = await axios.get(
-      url
-    );
-    this.appiumSessionConfig.size = {
-      width: rectResponse.data.value.width,
-      height: rectResponse.data.value.height
-    };
-    this.appiumSessionStarted = true;
     console.log(`>> Session created. Monitor execution at: ${this.getSessionLink()}`);
   }
   async createSession() {
@@ -141,15 +156,28 @@ class GptDriver {
       `${this.gptDriverBaseUrl}/sessions/create`,
       {
         api_key: this.apiKey,
-        appium_session_id: this.appiumSessionConfig.id,
+        appium_session_id: this.appiumSessionConfig?.id,
         device_config: {
-          platform: this.appiumSessionConfig.platform,
-          device: this.appiumSessionConfig.deviceName,
-          os: this.appiumSessionConfig.platformVersion
-        }
+          platform: this.appiumSessionConfig?.platform ?? this.gptDriverCloudPlatform,
+          device: this.appiumSessionConfig?.deviceName,
+          os: this.appiumSessionConfig?.platformVersion
+        },
+        use_internal_virtual_device: this.useGptDriverCloud,
+        build_id: this.buildId
       }
     );
     this.gptDriverSessionId = response.data.sessionId;
+    if (this.useGptDriverCloud) {
+      const parsedUrl = new URL(response.data.appiumServerUrl);
+      this.driver = await attach({
+        options: {
+          hostname: parsedUrl.hostname,
+          path: parsedUrl.pathname
+        },
+        sessionId: response.data.appiumSessionId
+      });
+      this.appiumSessionStarted = true;
+    }
   }
   getSessionLink() {
     return `https://app.mobileboost.io/gpt-driver/sessions/${this.gptDriverSessionId}`;
@@ -166,17 +194,19 @@ class GptDriver {
    * @throws {Error} If the request to stop the session fails.
    */
   async setSessionStatus(status) {
-    console.log(">> Stopping session...");
-    await axios.post(
-      `${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/stop`,
-      {
-        api_key: this.apiKey,
-        status
-      }
-    );
-    console.log(">> Session stopped.");
-    this.appiumSessionStarted = false;
-    this.gptDriverSessionId = void 0;
+    if (this.gptDriverSessionId) {
+      console.log(">> Stopping session...");
+      await axios.post(
+        `${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/stop`,
+        {
+          api_key: this.apiKey,
+          status
+        }
+      );
+      console.log(">> Session stopped.");
+      this.appiumSessionStarted = false;
+      this.gptDriverSessionId = void 0;
+    }
   }
   /**
    * Executes a specified command within the WebDriver session, optionally using an Appium handler.
@@ -226,10 +256,15 @@ class GptDriver {
     if (!this.appiumSessionStarted) {
       await this.startSession();
     }
-    console.log(">> Asserting:", assertion);
-    const results = await this.checkBulk([assertion]);
-    if (!Object.values(results).at(0)) {
-      throw new Error(`Failed assertion: ${assertion}`);
+    try {
+      const results = await this.checkBulk([assertion]);
+      if (!Object.values(results).at(0)) {
+        await this.setSessionStatus("failed");
+        throw new Error(`Failed assertion: ${assertion}`);
+      }
+    } catch (e) {
+      await this.setSessionStatus("failed");
+      throw e;
     }
   }
   /**
@@ -245,16 +280,21 @@ class GptDriver {
     if (!this.appiumSessionStarted) {
       await this.startSession();
     }
-    console.log(">> Asserting:", assertions);
-    const results = await this.checkBulk(assertions);
-    const failedAssertions = Object.values(results).reduce((prev, current, currentIndex) => {
-      if (!current) {
-        return [...prev, assertions.at(currentIndex)];
+    try {
+      const results = await this.checkBulk(assertions);
+      const failedAssertions = Object.values(results).reduce((prev, current, currentIndex) => {
+        if (!current) {
+          return [...prev, assertions.at(currentIndex)];
+        }
+        return prev;
+      }, []);
+      if (failedAssertions.length > 0) {
+        await this.setSessionStatus("failed");
+        throw new Error(`Failed assertions: ${failedAssertions.join(", ")}`);
       }
-      return prev;
-    }, []);
-    if (failedAssertions.length > 0) {
-      throw new Error(`Failed assertions: ${failedAssertions.join(", ")}`);
+    } catch (e) {
+      await this.setSessionStatus("failed");
+      throw e;
     }
   }
   /**
@@ -271,17 +311,25 @@ class GptDriver {
       await this.startSession();
     }
     console.log(">> Checking:", conditions);
-    const screenshot = await this.getScreenshot(this.appiumSessionConfig);
-    const response = await axios.post(
-      `${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/assert`,
-      {
-        api_key: this.apiKey,
-        base64_screenshot: screenshot,
-        assertions: conditions,
-        command: `Assert: ${JSON.stringify(conditions)}`
+    try {
+      let screenshot;
+      if (!this.useGptDriverCloud) {
+        screenshot = await this.getScreenshot(this.appiumSessionConfig);
       }
-    );
-    return response.data.results;
+      const response = await axios.post(
+        `${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/assert`,
+        {
+          api_key: this.apiKey,
+          base64_screenshot: screenshot,
+          assertions: conditions,
+          command: `Assert: ${JSON.stringify(conditions)}`
+        }
+      );
+      return response.data.results;
+    } catch (e) {
+      await this.setSessionStatus("failed");
+      throw e;
+    }
   }
   /**
    * Extracts specified information using the GPTDriver.
@@ -300,7 +348,10 @@ class GptDriver {
       await this.startSession();
     }
     console.log(">> Extracting:", extractions);
-    const screenshot = await this.getScreenshot(this.appiumSessionConfig);
+    let screenshot;
+    if (!this.useGptDriverCloud) {
+      screenshot = await this.getScreenshot(this.appiumSessionConfig);
+    }
     const response = await axios.post(
       `${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/extract`,
       {
@@ -316,7 +367,10 @@ class GptDriver {
     try {
       let conditionSucceeded = false;
       while (!conditionSucceeded) {
-        const screenshot = await this.getScreenshot(this.appiumSessionConfig);
+        let screenshot;
+        if (!this.useGptDriverCloud) {
+          screenshot = await this.getScreenshot(this.appiumSessionConfig);
+        }
         console.log(">> Asking GTP Driver for next action...");
         const response = await axios.request(
           {
@@ -349,10 +403,10 @@ class GptDriver {
     }
   }
   async executeCommand(command) {
-    const firstAction = command.data.actions?.at(0);
+    const firstAction = command?.data?.actions?.at(0);
     if (firstAction?.type === "pause" && firstAction.duration != null) {
       await delay(firstAction * 1e3);
-    } else {
+    } else if (!this.useGptDriverCloud) {
       const parsedUrl = new URL(command.url);
       parsedUrl.protocol = this.appiumSessionConfig.serverUrl.protocol;
       parsedUrl.host = this.appiumSessionConfig.serverUrl.host;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "gpt-driver-node",
-  "version": "1.0.0-alpha.12",
+  "version": "1.0.0-alpha.14",
   "main": "./dist/index.cjs",
   "module": "./dist/index.mjs",
   "types": "./dist/index.d.cts",