gpt-driver-node 1.0.0-alpha.3 → 1.0.0-alpha.4

package/dist/index.cjs

@@ -0,0 +1,202 @@
+'use strict';
+
+var axios = require('axios');
+var webdriverio = require('webdriverio');
+var sharp = require('sharp');
+
+const delay = async (milliseconds) => {
+  await new Promise((resolve) => setTimeout(resolve, milliseconds));
+};
+const getScreenshot = async (driver, deviceSize) => {
+  let screenshot = await driver.takeScreenshot();
+  if (driver.capabilities.platformName === "iOS") {
+    const imageBuffer = Buffer.from(screenshot, "base64");
+    const transformedImage = await sharp(imageBuffer).resize(deviceSize.width, deviceSize.height).toBuffer();
+    screenshot = transformedImage.toString("base64");
+  }
+  return screenshot;
+};
+const delayIfNeeded = async (lastResponseTime) => {
+  if (lastResponseTime == null) {
+    return;
+  }
+  const currentTime = Date.now();
+  const difference = currentTime - lastResponseTime;
+  if (difference < 1500) {
+    await delay(difference);
+  }
+};
+
+class GptDriver {
+  apiKey;
+  baseUrl;
+  driver;
+  deviceConfig;
+  firestoreSessionId;
+  deviceSize;
+  lastCommandExecutionTime;
+  constructor(config) {
+    this.apiKey = config.apiKey;
+    this.baseUrl = "https://api.mobileboost.io";
+    if (config.driver !== void 0) {
+      this.driver = config.driver;
+    } else if (config.deviceConfig !== void 0) {
+      this.deviceConfig = config.deviceConfig;
+    } else {
+      throw new Error("Either provide an appium driver or a deviceConfig object");
+    }
+  }
+  async startSession() {
+    console.log(">> Starting session...");
+    if (this.driver == void 0 && this.deviceConfig != void 0) {
+      const capabilities = {
+        platformName: this.deviceConfig.platform,
+        "appium:automationName": this.deviceConfig.platform === "Android" ? "UiAutomator2" : "XCUITest",
+        "appium:deviceName": this.deviceConfig.deviceName
+      };
+      const wdOpts = {
+        hostname: process.env.APPIUM_HOST || "localhost",
+        port: process.env.APPIUM_PORT != null ? parseInt(process.env.APPIUM_PORT, 10) : 4723,
+        logLevel: "error",
+        capabilities
+      };
+      this.driver = await webdriverio.remote(wdOpts);
+    }
+    const driver = this.driver;
+    const response = await axios.post(
+      `${this.baseUrl}/sessions/create`,
+      {
+        api_key: this.apiKey,
+        appium_session_id: driver.sessionId,
+        device_config: {
+          // @ts-ignore
+          platform: driver.capabilities.platformName,
+          // @ts-ignore
+          device: driver.capabilities.deviceName,
+          // @ts-ignore
+          os: driver.capabilities.platformVersion
+        }
+      }
+    );
+    const rect = await driver.getWindowRect();
+    this.deviceSize = { width: rect.width, height: rect.height };
+    if (response.data.sessionId) {
+      const sessionLink = `https://app.mobileboost.io/gpt-driver/sessions/${response.data.sessionId}`;
+      console.log(`>> Session created. Monitor execution at: ${sessionLink}`);
+      this.firestoreSessionId = response.data.sessionId;
+    }
+  }
+  async stopSession({ status }) {
+    console.log(">> Stopping session...");
+    await axios.post(
+      `${this.baseUrl}/sessions/${this.firestoreSessionId}/stop`,
+      {
+        api_key: this.apiKey,
+        status
+      }
+    );
+    console.log(">> Session stopped.");
+    this.firestoreSessionId = void 0;
+  }
+  async execute(command, appiumHandler) {
+    console.log(">> Executing command:", command);
+    const driver = this.driver;
+    if (appiumHandler != null) {
+      try {
+        await appiumHandler(driver);
+      } catch (e) {
+        await this.gptHandler(command);
+      }
+    } else {
+      await this.gptHandler(command);
+    }
+  }
+  async assert(assertions) {
+    console.log(">> Asserting:", assertions);
+    const driver = this.driver;
+    const screenshot = await getScreenshot(driver, this.deviceSize);
+    const response = await axios.post(
+      `${this.baseUrl}/sessions/${this.firestoreSessionId}/assert`,
+      {
+        api_key: this.apiKey,
+        base64_screenshot: screenshot,
+        assertions,
+        command: `Assert: ${JSON.stringify(assertions)}`
+      }
+    );
+    const results = response.data.results;
+    const failedAssertions = results.reduce((prev, current, currentIndex) => {
+      if (!current) {
+        return [...prev, assertions.at(currentIndex)];
+      }
+      return prev;
+    }, []);
+    if (failedAssertions.length > 0) {
+      throw new Error(`Failed assertions: ${failedAssertions.join(", ")}`);
+    }
+  }
+  async extract(extractions) {
+    console.log(">> Extracting:", extractions);
+    const driver = this.driver;
+    const screenshot = await getScreenshot(driver, this.deviceSize);
+    const response = await axios.post(
+      `${this.baseUrl}/sessions/${this.firestoreSessionId}/extract`,
+      {
+        api_key: this.apiKey,
+        base64_screenshot: screenshot,
+        extractions,
+        command: `Extract: ${JSON.stringify(extractions)}`
+      }
+    );
+    return response.data.results;
+  }
+  async gptHandler(command) {
+    const driver = this.driver;
+    try {
+      let conditionSucceeded = false;
+      while (!conditionSucceeded) {
+        await delayIfNeeded(this.lastCommandExecutionTime);
+        const screenshot = await getScreenshot(driver, this.deviceSize);
+        console.log(">> Waiting for GTPDriver response...");
+        const response = await axios.request(
+          {
+            url: `${this.baseUrl}/sessions/${this.firestoreSessionId}/execute`,
+            method: "POST",
+            data: {
+              api_key: this.apiKey,
+              command,
+              base64_screenshot: screenshot
+            }
+          }
+        );
+        const executeStatus = response.data.status;
+        if (executeStatus === "failed") {
+          throw new Error("Execution failed");
+        }
+        conditionSucceeded = executeStatus !== "inProgress";
+        const executeResponse = response.data;
+        for (const command2 of executeResponse.commands) {
+          await this.executeCommand(command2);
+        }
+        this.lastCommandExecutionTime = Date.now();
+      }
+    } catch (e) {
+      await this.stopSession({ status: "failed" });
+      throw e;
+    }
+  }
+  async executeCommand(command) {
+    const firstAction = command.data.actions?.at(0);
+    if (firstAction?.type === "pause" && firstAction.duration != null) {
+      await delay(firstAction * 1e3);
+    } else {
+      await axios.request({
+        url: command.url,
+        method: command.method,
+        data: command.data
+      });
+    }
+  }
+}
+
+module.exports = GptDriver;

package/dist/index.d.cts

@@ -0,0 +1,39 @@
+interface AppiumHandler {
+    (driver: WebdriverIO.Browser): Promise<any>;
+}
+interface DeviceConfig {
+    platform: "iOS" | "Android";
+    deviceName?: string;
+    platformVersion?: string;
+}
+interface GptDriverConfig {
+    apiKey: string;
+    driver?: WebdriverIO.Browser;
+    deviceConfig?: DeviceConfig;
+}
+interface DeviceSize {
+    width: number;
+    height: number;
+}
+
+declare class GptDriver {
+    apiKey: string;
+    baseUrl: string;
+    driver?: WebdriverIO.Browser;
+    deviceConfig?: DeviceConfig;
+    firestoreSessionId?: string;
+    deviceSize?: DeviceSize;
+    lastCommandExecutionTime?: number;
+    constructor(config: GptDriverConfig);
+    startSession(): Promise<void>;
+    stopSession({ status }: {
+        status: "failed" | "success";
+    }): Promise<void>;
+    execute(command: string, appiumHandler?: AppiumHandler): Promise<void>;
+    assert(assertions: string[]): Promise<void>;
+    extract(extractions: string[]): Promise<string[]>;
+    private gptHandler;
+    private executeCommand;
+}
+
+export { GptDriver as default };

package/dist/index.mjs

@@ -0,0 +1,200 @@
+import axios from 'axios';
+import { remote } from 'webdriverio';
+import sharp from 'sharp';
+
+const delay = async (milliseconds) => {
+  await new Promise((resolve) => setTimeout(resolve, milliseconds));
+};
+const getScreenshot = async (driver, deviceSize) => {
+  let screenshot = await driver.takeScreenshot();
+  if (driver.capabilities.platformName === "iOS") {
+    const imageBuffer = Buffer.from(screenshot, "base64");
+    const transformedImage = await sharp(imageBuffer).resize(deviceSize.width, deviceSize.height).toBuffer();
+    screenshot = transformedImage.toString("base64");
+  }
+  return screenshot;
+};
+const delayIfNeeded = async (lastResponseTime) => {
+  if (lastResponseTime == null) {
+    return;
+  }
+  const currentTime = Date.now();
+  const difference = currentTime - lastResponseTime;
+  if (difference < 1500) {
+    await delay(difference);
+  }
+};
+
+class GptDriver {
+  apiKey;
+  baseUrl;
+  driver;
+  deviceConfig;
+  firestoreSessionId;
+  deviceSize;
+  lastCommandExecutionTime;
+  constructor(config) {
+    this.apiKey = config.apiKey;
+    this.baseUrl = "https://api.mobileboost.io";
+    if (config.driver !== void 0) {
+      this.driver = config.driver;
+    } else if (config.deviceConfig !== void 0) {
+      this.deviceConfig = config.deviceConfig;
+    } else {
+      throw new Error("Either provide an appium driver or a deviceConfig object");
+    }
+  }
+  async startSession() {
+    console.log(">> Starting session...");
+    if (this.driver == void 0 && this.deviceConfig != void 0) {
+      const capabilities = {
+        platformName: this.deviceConfig.platform,
+        "appium:automationName": this.deviceConfig.platform === "Android" ? "UiAutomator2" : "XCUITest",
+        "appium:deviceName": this.deviceConfig.deviceName
+      };
+      const wdOpts = {
+        hostname: process.env.APPIUM_HOST || "localhost",
+        port: process.env.APPIUM_PORT != null ? parseInt(process.env.APPIUM_PORT, 10) : 4723,
+        logLevel: "error",
+        capabilities
+      };
+      this.driver = await remote(wdOpts);
+    }
+    const driver = this.driver;
+    const response = await axios.post(
+      `${this.baseUrl}/sessions/create`,
+      {
+        api_key: this.apiKey,
+        appium_session_id: driver.sessionId,
+        device_config: {
+          // @ts-ignore
+          platform: driver.capabilities.platformName,
+          // @ts-ignore
+          device: driver.capabilities.deviceName,
+          // @ts-ignore
+          os: driver.capabilities.platformVersion
+        }
+      }
+    );
+    const rect = await driver.getWindowRect();
+    this.deviceSize = { width: rect.width, height: rect.height };
+    if (response.data.sessionId) {
+      const sessionLink = `https://app.mobileboost.io/gpt-driver/sessions/${response.data.sessionId}`;
+      console.log(`>> Session created. Monitor execution at: ${sessionLink}`);
+      this.firestoreSessionId = response.data.sessionId;
+    }
+  }
+  async stopSession({ status }) {
+    console.log(">> Stopping session...");
+    await axios.post(
+      `${this.baseUrl}/sessions/${this.firestoreSessionId}/stop`,
+      {
+        api_key: this.apiKey,
+        status
+      }
+    );
+    console.log(">> Session stopped.");
+    this.firestoreSessionId = void 0;
+  }
+  async execute(command, appiumHandler) {
+    console.log(">> Executing command:", command);
+    const driver = this.driver;
+    if (appiumHandler != null) {
+      try {
+        await appiumHandler(driver);
+      } catch (e) {
+        await this.gptHandler(command);
+      }
+    } else {
+      await this.gptHandler(command);
+    }
+  }
+  async assert(assertions) {
+    console.log(">> Asserting:", assertions);
+    const driver = this.driver;
+    const screenshot = await getScreenshot(driver, this.deviceSize);
+    const response = await axios.post(
+      `${this.baseUrl}/sessions/${this.firestoreSessionId}/assert`,
+      {
+        api_key: this.apiKey,
+        base64_screenshot: screenshot,
+        assertions,
+        command: `Assert: ${JSON.stringify(assertions)}`
+      }
+    );
+    const results = response.data.results;
+    const failedAssertions = results.reduce((prev, current, currentIndex) => {
+      if (!current) {
+        return [...prev, assertions.at(currentIndex)];
+      }
+      return prev;
+    }, []);
+    if (failedAssertions.length > 0) {
+      throw new Error(`Failed assertions: ${failedAssertions.join(", ")}`);
+    }
+  }
+  async extract(extractions) {
+    console.log(">> Extracting:", extractions);
+    const driver = this.driver;
+    const screenshot = await getScreenshot(driver, this.deviceSize);
+    const response = await axios.post(
+      `${this.baseUrl}/sessions/${this.firestoreSessionId}/extract`,
+      {
+        api_key: this.apiKey,
+        base64_screenshot: screenshot,
+        extractions,
+        command: `Extract: ${JSON.stringify(extractions)}`
+      }
+    );
+    return response.data.results;
+  }
+  async gptHandler(command) {
+    const driver = this.driver;
+    try {
+      let conditionSucceeded = false;
+      while (!conditionSucceeded) {
+        await delayIfNeeded(this.lastCommandExecutionTime);
+        const screenshot = await getScreenshot(driver, this.deviceSize);
+        console.log(">> Waiting for GTPDriver response...");
+        const response = await axios.request(
+          {
+            url: `${this.baseUrl}/sessions/${this.firestoreSessionId}/execute`,
+            method: "POST",
+            data: {
+              api_key: this.apiKey,
+              command,
+              base64_screenshot: screenshot
+            }
+          }
+        );
+        const executeStatus = response.data.status;
+        if (executeStatus === "failed") {
+          throw new Error("Execution failed");
+        }
+        conditionSucceeded = executeStatus !== "inProgress";
+        const executeResponse = response.data;
+        for (const command2 of executeResponse.commands) {
+          await this.executeCommand(command2);
+        }
+        this.lastCommandExecutionTime = Date.now();
+      }
+    } catch (e) {
+      await this.stopSession({ status: "failed" });
+      throw e;
+    }
+  }
+  async executeCommand(command) {
+    const firstAction = command.data.actions?.at(0);
+    if (firstAction?.type === "pause" && firstAction.duration != null) {
+      await delay(firstAction * 1e3);
+    } else {
+      await axios.request({
+        url: command.url,
+        method: command.method,
+        data: command.data
+      });
+    }
+  }
+}
+
+export { GptDriver as default };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "gpt-driver-node",
-  "version": "1.0.0-alpha.3",
+  "version": "1.0.0-alpha.4",
   "main": "./dist/index.cjs",
   "module": "./dist/index.mjs",
   "types": "./dist/index.d.cts",