gpt-driver-node 1.0.6 → 1.0.8

package/dist/index.cjs

@@ -954,6 +954,7 @@ class GptDriver {
   step_number = 1;
   organisationId;
   configFilePath;
+  additionalUserContext;
   // Session Execution Stats
   _stats_startTime = 0;
   _stats_executedSteps = 0;
@@ -989,6 +990,7 @@ class GptDriver {
     this.cachingMode = config.cachingMode ?? "NONE";
     this.organisationId = config.organisationId;
     this.configFilePath = config.configFilePath;
+    this.additionalUserContext = config.additionalUserContext;
     if (config.useGptDriverCloud) {
       if (config.serverConfig.device?.platform == null) {
         throw new Error("Platform is missing. Please specify the platform when using GPTDriver Cloud.");
@@ -1115,7 +1117,8 @@ class GptDriver {
         },
         use_internal_virtual_device: this.useGptDriverCloud,
         build_id: this.buildId,
-        caching_mode: this.cachingMode
+        caching_mode: this.cachingMode,
+        ...this.additionalUserContext && { additional_user_context: this.additionalUserContext }
       }
     );
     this.gptDriverSessionId = response.data.sessionId;
@@ -1484,31 +1487,40 @@ ${"=".repeat(50)}`);
     globalLogger.debug("Capturing screenshot...");
     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 serverHostname = new URL(appiumSessionConfig.serverUrl).hostname;
+      const isLocalhost = serverHostname === "localhost" || serverHostname === "127.0.0.1" || serverHostname === "::1";
+      let rawBuffer;
+      if (isLocalhost) {
+        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}` : "";
+          rawBuffer = execSync(`adb ${deviceArg} exec-out screencap -p`, {
+            encoding: "buffer",
+            maxBuffer: 50 * 1024 * 1024
+          });
+        } catch (e) {
+          globalLogger.warn("ADB screenshot failed, falling back to Appium screenshot");
+          const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/screenshot`);
+          const screenshotResponse = await axios.get(url);
+          rawBuffer = Buffer.from(screenshotResponse.data.value, "base64");
         }
-        const deviceArg = udid ? `-s ${udid}` : "";
-        const buffer = execSync(`adb ${deviceArg} exec-out screencap -p`, {
-          encoding: "buffer",
-          maxBuffer: 50 * 1024 * 1024
-        });
-        screenshot = buffer.toString("base64");
-      } catch (e) {
-        globalLogger.warn("ADB screenshot failed, falling back to Appium screenshot");
+      } else {
         const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/screenshot`);
         const screenshotResponse = await axios.get(url);
-        screenshot = screenshotResponse.data.value;
+        rawBuffer = Buffer.from(screenshotResponse.data.value, "base64");
       }
+      screenshot = (await sharp(rawBuffer).jpeg({ quality: 80 }).toBuffer()).toString("base64");
     } else {
       const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/screenshot`);
       const screenshotResponse = await axios.get(url);
@@ -1617,7 +1629,7 @@ ${"=".repeat(50)}`);
       const smartLoopEndTime = performance.now();
       globalLogger.debug(`[Performance] Smart loop completed in ${(smartLoopEndTime - smartLoopStartTime).toFixed(2)}ms`);
       if (!result.success) {
-        throw new Error(result.error || "Smart loop execution failed");
+        await this.setSessionFailed();
       }
       if (result.cacheHitCount) {
         this._stats_cacheHits += result.cacheHitCount;

package/dist/index.d.cts

@@ -22,6 +22,7 @@ interface GptDriverConfig {
     useGptDriverCloud?: boolean;
     buildId?: string;
     cachingMode?: CachingMode;
+    additionalUserContext?: string;
     organisationId?: string;
     configFilePath?: string;
 }
@@ -512,6 +513,7 @@ declare class GptDriver {
     private step_number;
     private organisationId?;
     private configFilePath?;
+    private additionalUserContext?;
     private _stats_startTime;
     private _stats_executedSteps;
     private _stats_cacheHits;

package/dist/index.mjs

@@ -952,6 +952,7 @@ class GptDriver {
   step_number = 1;
   organisationId;
   configFilePath;
+  additionalUserContext;
   // Session Execution Stats
   _stats_startTime = 0;
   _stats_executedSteps = 0;
@@ -987,6 +988,7 @@ class GptDriver {
     this.cachingMode = config.cachingMode ?? "NONE";
     this.organisationId = config.organisationId;
     this.configFilePath = config.configFilePath;
+    this.additionalUserContext = config.additionalUserContext;
     if (config.useGptDriverCloud) {
       if (config.serverConfig.device?.platform == null) {
         throw new Error("Platform is missing. Please specify the platform when using GPTDriver Cloud.");
@@ -1113,7 +1115,8 @@ class GptDriver {
         },
         use_internal_virtual_device: this.useGptDriverCloud,
         build_id: this.buildId,
-        caching_mode: this.cachingMode
+        caching_mode: this.cachingMode,
+        ...this.additionalUserContext && { additional_user_context: this.additionalUserContext }
       }
     );
     this.gptDriverSessionId = response.data.sessionId;
@@ -1482,31 +1485,40 @@ ${"=".repeat(50)}`);
     globalLogger.debug("Capturing screenshot...");
     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 serverHostname = new URL(appiumSessionConfig.serverUrl).hostname;
+      const isLocalhost = serverHostname === "localhost" || serverHostname === "127.0.0.1" || serverHostname === "::1";
+      let rawBuffer;
+      if (isLocalhost) {
+        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}` : "";
+          rawBuffer = execSync(`adb ${deviceArg} exec-out screencap -p`, {
+            encoding: "buffer",
+            maxBuffer: 50 * 1024 * 1024
+          });
+        } catch (e) {
+          globalLogger.warn("ADB screenshot failed, falling back to Appium screenshot");
+          const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/screenshot`);
+          const screenshotResponse = await axios.get(url);
+          rawBuffer = Buffer.from(screenshotResponse.data.value, "base64");
         }
-        const deviceArg = udid ? `-s ${udid}` : "";
-        const buffer = execSync(`adb ${deviceArg} exec-out screencap -p`, {
-          encoding: "buffer",
-          maxBuffer: 50 * 1024 * 1024
-        });
-        screenshot = buffer.toString("base64");
-      } catch (e) {
-        globalLogger.warn("ADB screenshot failed, falling back to Appium screenshot");
+      } else {
         const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/screenshot`);
         const screenshotResponse = await axios.get(url);
-        screenshot = screenshotResponse.data.value;
+        rawBuffer = Buffer.from(screenshotResponse.data.value, "base64");
       }
+      screenshot = (await sharp(rawBuffer).jpeg({ quality: 80 }).toBuffer()).toString("base64");
     } else {
       const url = buildUrl(this.appiumSessionConfig.serverUrl, `/session/${this.appiumSessionConfig.id}/screenshot`);
       const screenshotResponse = await axios.get(url);
@@ -1615,7 +1627,7 @@ ${"=".repeat(50)}`);
       const smartLoopEndTime = performance.now();
       globalLogger.debug(`[Performance] Smart loop completed in ${(smartLoopEndTime - smartLoopStartTime).toFixed(2)}ms`);
       if (!result.success) {
-        throw new Error(result.error || "Smart loop execution failed");
+        await this.setSessionFailed();
       }
       if (result.cacheHitCount) {
         this._stats_cacheHits += result.cacheHitCount;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "gpt-driver-node",
-  "version": "1.0.6",
+  "version": "1.0.8",
   "main": "./dist/index.cjs",
   "module": "./dist/index.mjs",
   "types": "./dist/index.d.cts",