gpt-driver-node 1.0.0-alpha.6 → 1.0.0-alpha.8

package/dist/index.cjs

@@ -1,21 +1,22 @@
 'use strict';
 
 var axios = require('axios');
-var sharp = require('sharp');
 var seleniumWebdriver = require('selenium-webdriver');
+var sharp = require('sharp');
 
 const delay = async (milliseconds) => {
   await new Promise((resolve) => setTimeout(resolve, milliseconds));
 };
-const getScreenshot = async (appiumSessionConfig, driver) => {
-  let screenshot = await driver.takeScreenshot();
-  if (appiumSessionConfig.platform === "iOS") {
-    const imageBuffer = Buffer.from(screenshot, "base64");
-    const transformedImage = await sharp(imageBuffer).resize(appiumSessionConfig.size.width, appiumSessionConfig.size.height).toBuffer();
-    screenshot = transformedImage.toString("base64");
+function buildUrl(base, extraPath) {
+  let baseUrl = base.toString();
+  if (baseUrl.endsWith("/")) {
+    baseUrl = baseUrl.slice(0, -1);
   }
-  return screenshot;
-};
+  if (!extraPath.startsWith("/")) {
+    extraPath = "/" + extraPath;
+  }
+  return `${baseUrl}${extraPath}`;
+}
 
 class GptDriver {
   apiKey;
@@ -61,9 +62,9 @@ class GptDriver {
     }
   }
   initializeAppiumConfig(config) {
-    const defaultPort = process.env.APPIUM_PORT ? parseInt(process.env.APPIUM_PORT, 10) : 4723;
+    const defaultPort = parseInt(process.env.APPIUM_PORT ?? "4723", 10);
     const defaultHost = process.env.APPIUM_HOST ?? "127.0.0.1";
-    const serverUrl = config.severConfig?.url != null ? typeof config.severConfig?.url === "string" ? new URL(config.severConfig.url) : config.severConfig?.url : new URL(`http://${defaultPort}:${defaultHost}`);
+    let serverUrl = config.severConfig?.url instanceof URL ? config.severConfig.url : new URL(config.severConfig?.url ?? `http://${defaultHost}:${defaultPort}`);
     this.appiumSessionConfig = {
       serverUrl,
       ...config.severConfig?.device
@@ -78,37 +79,61 @@ class GptDriver {
   async startSession() {
     console.log(">> Starting session...");
     if (this.driver) {
-      const capabilities = await this.driver.getCapabilities();
-      const platform = capabilities.get("platformName");
-      const platformVersion = capabilities.get("platformVersion") ?? this.appiumSessionConfig?.platformVersion;
-      const deviceName = this.appiumSessionConfig?.deviceName ?? capabilities.get("deviceName");
+      let platform;
+      let platformVersion;
+      let deviceName;
+      let sessionId;
+      if (this.driver instanceof seleniumWebdriver.WebDriver) {
+        const capabilities = await this.driver.getCapabilities();
+        platform = capabilities.get("platformName");
+        platformVersion = capabilities.get("platformVersion") ?? this.appiumSessionConfig?.platformVersion;
+        deviceName = this.appiumSessionConfig?.deviceName ?? capabilities.get("deviceName");
+        const session = await this.driver.getSession();
+        sessionId = session.getId();
+      } else {
+        platform = this.driver.capabilities["appium:platformName"];
+        platformVersion = this.driver.capabilities["appium:platformVersion"];
+        deviceName = this.appiumSessionConfig?.deviceName ?? this.driver.capabilities["appium:deviceName"] ?? "";
+        sessionId = this.driver.sessionId;
+      }
       this.appiumSessionConfig = {
         ...this.appiumSessionConfig,
+        id: sessionId,
         platform,
         platformVersion,
         deviceName
       };
     } else {
-      this.driver = await this.buildDriver();
+      this.appiumSessionConfig.id = await this.createSession();
     }
-    const session = await this.driver.getSession();
-    this.appiumSessionConfig.id = session.getId();
     await this.createGptDriverSession();
-    const rectResponse = await this.driver.manage().window().getRect();
+    const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/window/rect`);
+    const rectResponse = await axios.get(
+      url
+    );
     this.appiumSessionConfig.size = {
-      width: rectResponse.width,
-      height: rectResponse.height
+      width: rectResponse.data.value.width,
+      height: rectResponse.data.value.height
     };
     console.log(`>> Session created. Monitor execution at: ${this.getSessionLink()}`);
   }
-  async buildDriver() {
-    const { platform, deviceName, platformVersion } = this.appiumSessionConfig;
-    return new seleniumWebdriver.Builder().withCapabilities({
-      platformName: platform,
-      automationName: platform === "iOS" ? "XCUITest" : "UiAutomator2",
-      deviceName,
-      platformVersion
-    }).usingServer(this.appiumSessionConfig.serverUrl.toString()).build();
+  async createSession() {
+    const { platform, deviceName, platformVersion, serverUrl } = this.appiumSessionConfig;
+    const url = buildUrl(serverUrl, `/session`);
+    const response = await axios.post(
+      url,
+      {
+        capabilities: {
+          alwaysMatch: {
+            platformName: platform,
+            "appium:automationName": platform === "iOS" ? "XCUITest" : "UiAutomator2",
+            "appium:deviceName": deviceName,
+            "appium:platformVersion": platformVersion
+          }
+        }
+      }
+    );
+    return response.data.value.sessionId;
   }
   async createGptDriverSession() {
     const response = await axios.post(
@@ -148,6 +173,8 @@ class GptDriver {
         status
       }
     );
+    const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}`);
+    await axios.delete(url);
     console.log(">> Session stopped.");
     this.gptDriverSessionId = void 0;
   }
@@ -170,7 +197,7 @@ class GptDriver {
     if (appiumHandler != null) {
       try {
         await appiumHandler(driver);
-        const screenshot = await getScreenshot(this.appiumSessionConfig, this.driver);
+        const screenshot = await this.getScreenshot(this.appiumSessionConfig);
         await axios.post(`${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/log_code_execution`, {
           api_key: this.apiKey,
           base64_screenshot: screenshot,
@@ -232,7 +259,7 @@ class GptDriver {
    */
   async checkBulk(conditions) {
     console.log(">> Checking:", conditions);
-    const screenshot = await getScreenshot(this.appiumSessionConfig, this.driver);
+    const screenshot = await this.getScreenshot(this.appiumSessionConfig);
     const response = await axios.post(
       `${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/assert`,
       {
@@ -258,7 +285,7 @@ class GptDriver {
    */
   async extract(extractions) {
     console.log(">> Extracting:", extractions);
-    const screenshot = await getScreenshot(this.appiumSessionConfig, this.driver);
+    const screenshot = await this.getScreenshot(this.appiumSessionConfig);
     const response = await axios.post(
       `${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/extract`,
       {
@@ -274,7 +301,7 @@ class GptDriver {
     try {
       let conditionSucceeded = false;
       while (!conditionSucceeded) {
-        const screenshot = await getScreenshot(this.appiumSessionConfig, this.driver);
+        const screenshot = await this.getScreenshot(this.appiumSessionConfig);
         console.log(">> Asking GTP Driver for next action...");
         const response = await axios.request(
           {
@@ -323,6 +350,17 @@ class GptDriver {
       });
     }
   }
+  async getScreenshot(appiumSessionConfig) {
+    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") {
+      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;
+  }
 }
 
 module.exports = GptDriver;

package/dist/index.d.cts

@@ -1,8 +1,9 @@
 import { WebDriver } from 'selenium-webdriver';
+import { Browser } from 'webdriverio';
 
 type Platform = "iOS" | "Android";
 interface AppiumHandler {
-    (driver: WebDriver): Promise<any>;
+    (driver: WebDriver | Browser): Promise<any>;
 }
 interface ServerSessionInitConfig {
     platform?: Platform;
@@ -11,7 +12,7 @@ interface ServerSessionInitConfig {
 }
 interface GptDriverConfig {
     apiKey: string;
-    driver?: WebDriver;
+    driver?: WebDriver | Browser;
     severConfig: {
         device?: ServerSessionInitConfig;
         url?: URL | string;
@@ -52,7 +53,7 @@ declare class GptDriver {
      * @throws {Error} If the session cannot be started or the driver is not properly initialized.
      */
     startSession(): Promise<void>;
-    private buildDriver;
+    private createSession;
     private createGptDriverSession;
     private getSessionLink;
     /**
@@ -126,6 +127,7 @@ declare class GptDriver {
     extract(extractions: string[]): Promise<any>;
     private gptHandler;
     private executeCommand;
+    private getScreenshot;
 }
 
 export { GptDriver as default };

package/dist/index.mjs

@@ -1,19 +1,20 @@
 import axios from 'axios';
+import { WebDriver } from 'selenium-webdriver';
 import sharp from 'sharp';
-import { Builder } from 'selenium-webdriver';
 
 const delay = async (milliseconds) => {
   await new Promise((resolve) => setTimeout(resolve, milliseconds));
 };
-const getScreenshot = async (appiumSessionConfig, driver) => {
-  let screenshot = await driver.takeScreenshot();
-  if (appiumSessionConfig.platform === "iOS") {
-    const imageBuffer = Buffer.from(screenshot, "base64");
-    const transformedImage = await sharp(imageBuffer).resize(appiumSessionConfig.size.width, appiumSessionConfig.size.height).toBuffer();
-    screenshot = transformedImage.toString("base64");
+function buildUrl(base, extraPath) {
+  let baseUrl = base.toString();
+  if (baseUrl.endsWith("/")) {
+    baseUrl = baseUrl.slice(0, -1);
   }
-  return screenshot;
-};
+  if (!extraPath.startsWith("/")) {
+    extraPath = "/" + extraPath;
+  }
+  return `${baseUrl}${extraPath}`;
+}
 
 class GptDriver {
   apiKey;
@@ -59,9 +60,9 @@ class GptDriver {
     }
   }
   initializeAppiumConfig(config) {
-    const defaultPort = process.env.APPIUM_PORT ? parseInt(process.env.APPIUM_PORT, 10) : 4723;
+    const defaultPort = parseInt(process.env.APPIUM_PORT ?? "4723", 10);
     const defaultHost = process.env.APPIUM_HOST ?? "127.0.0.1";
-    const serverUrl = config.severConfig?.url != null ? typeof config.severConfig?.url === "string" ? new URL(config.severConfig.url) : config.severConfig?.url : new URL(`http://${defaultPort}:${defaultHost}`);
+    let serverUrl = config.severConfig?.url instanceof URL ? config.severConfig.url : new URL(config.severConfig?.url ?? `http://${defaultHost}:${defaultPort}`);
     this.appiumSessionConfig = {
       serverUrl,
       ...config.severConfig?.device
@@ -76,37 +77,61 @@ class GptDriver {
   async startSession() {
     console.log(">> Starting session...");
     if (this.driver) {
-      const capabilities = await this.driver.getCapabilities();
-      const platform = capabilities.get("platformName");
-      const platformVersion = capabilities.get("platformVersion") ?? this.appiumSessionConfig?.platformVersion;
-      const deviceName = this.appiumSessionConfig?.deviceName ?? capabilities.get("deviceName");
+      let platform;
+      let platformVersion;
+      let deviceName;
+      let sessionId;
+      if (this.driver instanceof WebDriver) {
+        const capabilities = await this.driver.getCapabilities();
+        platform = capabilities.get("platformName");
+        platformVersion = capabilities.get("platformVersion") ?? this.appiumSessionConfig?.platformVersion;
+        deviceName = this.appiumSessionConfig?.deviceName ?? capabilities.get("deviceName");
+        const session = await this.driver.getSession();
+        sessionId = session.getId();
+      } else {
+        platform = this.driver.capabilities["appium:platformName"];
+        platformVersion = this.driver.capabilities["appium:platformVersion"];
+        deviceName = this.appiumSessionConfig?.deviceName ?? this.driver.capabilities["appium:deviceName"] ?? "";
+        sessionId = this.driver.sessionId;
+      }
       this.appiumSessionConfig = {
         ...this.appiumSessionConfig,
+        id: sessionId,
         platform,
         platformVersion,
         deviceName
       };
     } else {
-      this.driver = await this.buildDriver();
+      this.appiumSessionConfig.id = await this.createSession();
     }
-    const session = await this.driver.getSession();
-    this.appiumSessionConfig.id = session.getId();
     await this.createGptDriverSession();
-    const rectResponse = await this.driver.manage().window().getRect();
+    const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/window/rect`);
+    const rectResponse = await axios.get(
+      url
+    );
     this.appiumSessionConfig.size = {
-      width: rectResponse.width,
-      height: rectResponse.height
+      width: rectResponse.data.value.width,
+      height: rectResponse.data.value.height
     };
     console.log(`>> Session created. Monitor execution at: ${this.getSessionLink()}`);
   }
-  async buildDriver() {
-    const { platform, deviceName, platformVersion } = this.appiumSessionConfig;
-    return new Builder().withCapabilities({
-      platformName: platform,
-      automationName: platform === "iOS" ? "XCUITest" : "UiAutomator2",
-      deviceName,
-      platformVersion
-    }).usingServer(this.appiumSessionConfig.serverUrl.toString()).build();
+  async createSession() {
+    const { platform, deviceName, platformVersion, serverUrl } = this.appiumSessionConfig;
+    const url = buildUrl(serverUrl, `/session`);
+    const response = await axios.post(
+      url,
+      {
+        capabilities: {
+          alwaysMatch: {
+            platformName: platform,
+            "appium:automationName": platform === "iOS" ? "XCUITest" : "UiAutomator2",
+            "appium:deviceName": deviceName,
+            "appium:platformVersion": platformVersion
+          }
+        }
+      }
+    );
+    return response.data.value.sessionId;
   }
   async createGptDriverSession() {
     const response = await axios.post(
@@ -146,6 +171,8 @@ class GptDriver {
         status
       }
     );
+    const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}`);
+    await axios.delete(url);
     console.log(">> Session stopped.");
     this.gptDriverSessionId = void 0;
   }
@@ -168,7 +195,7 @@ class GptDriver {
     if (appiumHandler != null) {
       try {
         await appiumHandler(driver);
-        const screenshot = await getScreenshot(this.appiumSessionConfig, this.driver);
+        const screenshot = await this.getScreenshot(this.appiumSessionConfig);
         await axios.post(`${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/log_code_execution`, {
           api_key: this.apiKey,
           base64_screenshot: screenshot,
@@ -230,7 +257,7 @@ class GptDriver {
    */
   async checkBulk(conditions) {
     console.log(">> Checking:", conditions);
-    const screenshot = await getScreenshot(this.appiumSessionConfig, this.driver);
+    const screenshot = await this.getScreenshot(this.appiumSessionConfig);
     const response = await axios.post(
       `${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/assert`,
       {
@@ -256,7 +283,7 @@ class GptDriver {
    */
   async extract(extractions) {
     console.log(">> Extracting:", extractions);
-    const screenshot = await getScreenshot(this.appiumSessionConfig, this.driver);
+    const screenshot = await this.getScreenshot(this.appiumSessionConfig);
     const response = await axios.post(
       `${this.gptDriverBaseUrl}/sessions/${this.gptDriverSessionId}/extract`,
       {
@@ -272,7 +299,7 @@ class GptDriver {
     try {
       let conditionSucceeded = false;
       while (!conditionSucceeded) {
-        const screenshot = await getScreenshot(this.appiumSessionConfig, this.driver);
+        const screenshot = await this.getScreenshot(this.appiumSessionConfig);
         console.log(">> Asking GTP Driver for next action...");
         const response = await axios.request(
           {
@@ -321,6 +348,17 @@ class GptDriver {
       });
     }
   }
+  async getScreenshot(appiumSessionConfig) {
+    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") {
+      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;
+  }
 }
 
 export { GptDriver as default };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "gpt-driver-node",
-  "version": "1.0.0-alpha.6",
+  "version": "1.0.0-alpha.8",
   "main": "./dist/index.cjs",
   "module": "./dist/index.mjs",
   "types": "./dist/index.d.cts",
@@ -10,15 +10,18 @@
     "watch": "tsx watch src/index.ts",
     "test": "jest"
   },
+  "type": "module",
   "keywords": [],
   "author": "MobileBoost",
   "license": "CC BY-NC-SA 4.0",
   "description": "Test your mobile apps with the AI native GPT Driver (docs.mobileboost.io).",
   "dependencies": {
     "@types/selenium-webdriver": "^4.1.25",
+    "@wdio/types": "^9.0.4",
     "axios": "^1.7.3",
     "selenium-webdriver": "^4.23.0",
-    "sharp": "^0.33.4"
+    "sharp": "^0.33.4",
+    "webdriverio": "^9.0.7"
   },
   "devDependencies": {
     "tsx": "^4.16.5",