@midscene/shared 0.11.2 → 0.11.4-beta-20250219074605.0

package/dist/browser/img.d.ts

@@ -67,11 +67,11 @@ declare function transformImgPathToBase64(inputPath: string): Promise<string>;
  * @returns A Promise that resolves to a base64-encoded string representing the resized image
  * @throws An error if the width or height cannot be determined from the metadata
  */
-declare function resizeImg(inputData: Buffer, newSize?: {
+declare function resizeImg(inputData: Buffer, newSize: {
     width: number;
     height: number;
 }): Promise<Buffer>;
-declare function resizeImgBase64(inputData: string, newSize?: {
+declare function resizeImgBase64(inputBase64: string, newSize: {
     width: number;
     height: number;
 }): Promise<string>;
@@ -87,7 +87,7 @@ declare function resizeImgBase64(inputData: string, newSize?: {
  * @returns {Object} An object containing the new width and height.
  * @throws {Error} Throws an error if the width or height is not a positive number.
  */
-declare function calculateNewDimensions(originalWidth: number, originalHeight: number): {
+declare function zoomForGPT4o(originalWidth: number, originalHeight: number): {
     width: number;
     height: number;
 };
@@ -148,4 +148,4 @@ declare function savePositionImg(options: {
     outputPath: string;
 }): Promise<void>;
 
-export { base64Encoded, base64ToPngFormat, bufferFromBase64, calculateNewDimensions, compositeElementInfoImg, drawBoxOnImage, imageInfo, imageInfoOfBase64, processImageElementInfo, resizeImg, resizeImgBase64, saveBase64Image, savePositionImg, transformImgPathToBase64, trimImage };
+export { base64Encoded, base64ToPngFormat, bufferFromBase64, compositeElementInfoImg, drawBoxOnImage, imageInfo, imageInfoOfBase64, processImageElementInfo, resizeImg, resizeImgBase64, saveBase64Image, savePositionImg, transformImgPathToBase64, trimImage, zoomForGPT4o };

package/dist/browser/img.js

@@ -54,7 +54,6 @@ __export(img_exports, {
   base64Encoded: () => base64Encoded,
   base64ToPngFormat: () => base64ToPngFormat,
   bufferFromBase64: () => bufferFromBase64,
-  calculateNewDimensions: () => calculateNewDimensions,
   compositeElementInfoImg: () => compositeElementInfoImg,
   drawBoxOnImage: () => drawBoxOnImage,
   imageInfo: () => imageInfo,
@@ -65,7 +64,8 @@ __export(img_exports, {
   saveBase64Image: () => saveBase64Image,
   savePositionImg: () => savePositionImg,
   transformImgPathToBase64: () => transformImgPathToBase64,
-  trimImage: () => trimImage
+  trimImage: () => trimImage,
+  zoomForGPT4o: () => zoomForGPT4o
 });
 module.exports = __toCommonJS(img_exports);
 
@@ -138,6 +138,7 @@ function base64ToPngFormat(base64) {
 }
 
 // src/img/transform.ts
+var import_node_assert2 = __toESM(require("assert"));
 var import_node_buffer2 = require("buffer");
 function saveBase64Image(options) {
   return __async(this, null, function* () {
@@ -161,27 +162,29 @@ function resizeImg(inputData, newSize) {
   return __async(this, null, function* () {
     if (typeof inputData === "string")
       throw Error("inputData is base64, use resizeImgBase64 instead");
+    (0, import_node_assert2.default)(
+      newSize && newSize.width > 0 && newSize.height > 0,
+      "newSize must be positive"
+    );
     const Jimp = yield getJimp();
     const image = yield Jimp.read(inputData);
     const { width, height } = image.bitmap;
     if (!width || !height) {
       throw Error("Undefined width or height from the input image.");
     }
-    const finalNewSize = newSize || calculateNewDimensions(width, height);
-    image.resize(
-      finalNewSize.width,
-      finalNewSize.height,
-      Jimp.RESIZE_NEAREST_NEIGHBOR
-    );
+    if (newSize.width === width && newSize.height === height) {
+      return inputData;
+    }
+    image.resize(newSize.width, newSize.height, Jimp.RESIZE_NEAREST_NEIGHBOR);
     image.quality(90);
     const resizedBuffer = yield image.getBufferAsync(Jimp.MIME_JPEG);
     return resizedBuffer;
   });
 }
-function resizeImgBase64(inputData, newSize) {
+function resizeImgBase64(inputBase64, newSize) {
   return __async(this, null, function* () {
     const splitFlag = ";base64,";
-    const dataSplitted = inputData.split(splitFlag);
+    const dataSplitted = inputBase64.split(splitFlag);
     if (dataSplitted.length !== 2) {
       throw Error("Invalid base64 data");
     }
@@ -191,7 +194,7 @@ function resizeImgBase64(inputData, newSize) {
     return `${dataSplitted[0]}${splitFlag}${content}`;
   });
 }
-function calculateNewDimensions(originalWidth, originalHeight) {
+function zoomForGPT4o(originalWidth, originalHeight) {
   const maxWidth = 2048;
   const maxHeight = 768;
   let newWidth = originalWidth;
@@ -237,7 +240,7 @@ function trimImage(image) {
 }
 
 // src/img/box-select.ts
-var import_node_assert2 = __toESM(require("assert"));
+var import_node_assert3 = __toESM(require("assert"));
 var cachedFont = null;
 var createSvgOverlay = (elements, imageWidth, imageHeight) => __async(void 0, null, function* () {
   const Jimp = yield getJimp();
@@ -343,7 +346,7 @@ var createSvgOverlay = (elements, imageWidth, imageHeight) => __async(void 0, nu
   return image;
 });
 var compositeElementInfoImg = (options) => __async(void 0, null, function* () {
-  (0, import_node_assert2.default)(options.inputImgBase64, "inputImgBase64 is required");
+  (0, import_node_assert3.default)(options.inputImgBase64, "inputImgBase64 is required");
   let width = 0;
   let height = 0;
   let jimpImage;
@@ -393,7 +396,7 @@ var compositeElementInfoImg = (options) => __async(void 0, null, function* () {
 });
 var processImageElementInfo = (options) => __async(void 0, null, function* () {
   const base64Image = options.inputImgBase64.split(";base64,").pop();
-  (0, import_node_assert2.default)(base64Image, "base64Image is undefined");
+  (0, import_node_assert3.default)(base64Image, "base64Image is undefined");
   const [
     compositeElementInfoImgBase64,
     compositeElementInfoImgWithoutTextBase64

package/dist/es/img.d.ts

@@ -67,11 +67,11 @@ declare function transformImgPathToBase64(inputPath: string): Promise<string>;
  * @returns A Promise that resolves to a base64-encoded string representing the resized image
  * @throws An error if the width or height cannot be determined from the metadata
  */
-declare function resizeImg(inputData: Buffer, newSize?: {
+declare function resizeImg(inputData: Buffer, newSize: {
     width: number;
     height: number;
 }): Promise<Buffer>;
-declare function resizeImgBase64(inputData: string, newSize?: {
+declare function resizeImgBase64(inputBase64: string, newSize: {
     width: number;
     height: number;
 }): Promise<string>;
@@ -87,7 +87,7 @@ declare function resizeImgBase64(inputData: string, newSize?: {
  * @returns {Object} An object containing the new width and height.
  * @throws {Error} Throws an error if the width or height is not a positive number.
  */
-declare function calculateNewDimensions(originalWidth: number, originalHeight: number): {
+declare function zoomForGPT4o(originalWidth: number, originalHeight: number): {
     width: number;
     height: number;
 };
@@ -148,4 +148,4 @@ declare function savePositionImg(options: {
     outputPath: string;
 }): Promise<void>;
 
-export { base64Encoded, base64ToPngFormat, bufferFromBase64, calculateNewDimensions, compositeElementInfoImg, drawBoxOnImage, imageInfo, imageInfoOfBase64, processImageElementInfo, resizeImg, resizeImgBase64, saveBase64Image, savePositionImg, transformImgPathToBase64, trimImage };
+export { base64Encoded, base64ToPngFormat, bufferFromBase64, compositeElementInfoImg, drawBoxOnImage, imageInfo, imageInfoOfBase64, processImageElementInfo, resizeImg, resizeImgBase64, saveBase64Image, savePositionImg, transformImgPathToBase64, trimImage, zoomForGPT4o };

package/dist/es/img.js

@@ -89,6 +89,7 @@ function base64ToPngFormat(base64) {
 }
 
 // src/img/transform.ts
+import assert2 from "assert";
 import { Buffer as Buffer3 } from "buffer";
 function saveBase64Image(options) {
   return __async(this, null, function* () {
@@ -112,27 +113,29 @@ function resizeImg(inputData, newSize) {
   return __async(this, null, function* () {
     if (typeof inputData === "string")
       throw Error("inputData is base64, use resizeImgBase64 instead");
+    assert2(
+      newSize && newSize.width > 0 && newSize.height > 0,
+      "newSize must be positive"
+    );
     const Jimp = yield getJimp();
     const image = yield Jimp.read(inputData);
     const { width, height } = image.bitmap;
     if (!width || !height) {
       throw Error("Undefined width or height from the input image.");
     }
-    const finalNewSize = newSize || calculateNewDimensions(width, height);
-    image.resize(
-      finalNewSize.width,
-      finalNewSize.height,
-      Jimp.RESIZE_NEAREST_NEIGHBOR
-    );
+    if (newSize.width === width && newSize.height === height) {
+      return inputData;
+    }
+    image.resize(newSize.width, newSize.height, Jimp.RESIZE_NEAREST_NEIGHBOR);
     image.quality(90);
     const resizedBuffer = yield image.getBufferAsync(Jimp.MIME_JPEG);
     return resizedBuffer;
   });
 }
-function resizeImgBase64(inputData, newSize) {
+function resizeImgBase64(inputBase64, newSize) {
   return __async(this, null, function* () {
     const splitFlag = ";base64,";
-    const dataSplitted = inputData.split(splitFlag);
+    const dataSplitted = inputBase64.split(splitFlag);
     if (dataSplitted.length !== 2) {
       throw Error("Invalid base64 data");
     }
@@ -142,7 +145,7 @@ function resizeImgBase64(inputData, newSize) {
     return `${dataSplitted[0]}${splitFlag}${content}`;
   });
 }
-function calculateNewDimensions(originalWidth, originalHeight) {
+function zoomForGPT4o(originalWidth, originalHeight) {
   const maxWidth = 2048;
   const maxHeight = 768;
   let newWidth = originalWidth;
@@ -188,7 +191,7 @@ function trimImage(image) {
 }
 
 // src/img/box-select.ts
-import assert2 from "assert";
+import assert3 from "assert";
 var cachedFont = null;
 var createSvgOverlay = (elements, imageWidth, imageHeight) => __async(void 0, null, function* () {
   const Jimp = yield getJimp();
@@ -294,7 +297,7 @@ var createSvgOverlay = (elements, imageWidth, imageHeight) => __async(void 0, nu
   return image;
 });
 var compositeElementInfoImg = (options) => __async(void 0, null, function* () {
-  assert2(options.inputImgBase64, "inputImgBase64 is required");
+  assert3(options.inputImgBase64, "inputImgBase64 is required");
   let width = 0;
   let height = 0;
   let jimpImage;
@@ -344,7 +347,7 @@ var compositeElementInfoImg = (options) => __async(void 0, null, function* () {
 });
 var processImageElementInfo = (options) => __async(void 0, null, function* () {
   const base64Image = options.inputImgBase64.split(";base64,").pop();
-  assert2(base64Image, "base64Image is undefined");
+  assert3(base64Image, "base64Image is undefined");
   const [
     compositeElementInfoImgBase64,
     compositeElementInfoImgWithoutTextBase64
@@ -409,7 +412,6 @@ export {
   base64Encoded,
   base64ToPngFormat,
   bufferFromBase64,
-  calculateNewDimensions,
   compositeElementInfoImg,
   drawBoxOnImage,
   imageInfo,
@@ -420,5 +422,6 @@ export {
   saveBase64Image,
   savePositionImg,
   transformImgPathToBase64,
-  trimImage
+  trimImage,
+  zoomForGPT4o
 };

package/dist/lib/img.d.ts

@@ -67,11 +67,11 @@ declare function transformImgPathToBase64(inputPath: string): Promise<string>;
  * @returns A Promise that resolves to a base64-encoded string representing the resized image
  * @throws An error if the width or height cannot be determined from the metadata
  */
-declare function resizeImg(inputData: Buffer, newSize?: {
+declare function resizeImg(inputData: Buffer, newSize: {
     width: number;
     height: number;
 }): Promise<Buffer>;
-declare function resizeImgBase64(inputData: string, newSize?: {
+declare function resizeImgBase64(inputBase64: string, newSize: {
     width: number;
     height: number;
 }): Promise<string>;
@@ -87,7 +87,7 @@ declare function resizeImgBase64(inputData: string, newSize?: {
  * @returns {Object} An object containing the new width and height.
  * @throws {Error} Throws an error if the width or height is not a positive number.
  */
-declare function calculateNewDimensions(originalWidth: number, originalHeight: number): {
+declare function zoomForGPT4o(originalWidth: number, originalHeight: number): {
     width: number;
     height: number;
 };
@@ -148,4 +148,4 @@ declare function savePositionImg(options: {
     outputPath: string;
 }): Promise<void>;
 
-export { base64Encoded, base64ToPngFormat, bufferFromBase64, calculateNewDimensions, compositeElementInfoImg, drawBoxOnImage, imageInfo, imageInfoOfBase64, processImageElementInfo, resizeImg, resizeImgBase64, saveBase64Image, savePositionImg, transformImgPathToBase64, trimImage };
+export { base64Encoded, base64ToPngFormat, bufferFromBase64, compositeElementInfoImg, drawBoxOnImage, imageInfo, imageInfoOfBase64, processImageElementInfo, resizeImg, resizeImgBase64, saveBase64Image, savePositionImg, transformImgPathToBase64, trimImage, zoomForGPT4o };

package/dist/lib/img.js

@@ -54,7 +54,6 @@ __export(img_exports, {
   base64Encoded: () => base64Encoded,
   base64ToPngFormat: () => base64ToPngFormat,
   bufferFromBase64: () => bufferFromBase64,
-  calculateNewDimensions: () => calculateNewDimensions,
   compositeElementInfoImg: () => compositeElementInfoImg,
   drawBoxOnImage: () => drawBoxOnImage,
   imageInfo: () => imageInfo,
@@ -65,7 +64,8 @@ __export(img_exports, {
   saveBase64Image: () => saveBase64Image,
   savePositionImg: () => savePositionImg,
   transformImgPathToBase64: () => transformImgPathToBase64,
-  trimImage: () => trimImage
+  trimImage: () => trimImage,
+  zoomForGPT4o: () => zoomForGPT4o
 });
 module.exports = __toCommonJS(img_exports);
 
@@ -138,6 +138,7 @@ function base64ToPngFormat(base64) {
 }
 
 // src/img/transform.ts
+var import_node_assert2 = __toESM(require("assert"));
 var import_node_buffer2 = require("buffer");
 function saveBase64Image(options) {
   return __async(this, null, function* () {
@@ -161,27 +162,29 @@ function resizeImg(inputData, newSize) {
   return __async(this, null, function* () {
     if (typeof inputData === "string")
       throw Error("inputData is base64, use resizeImgBase64 instead");
+    (0, import_node_assert2.default)(
+      newSize && newSize.width > 0 && newSize.height > 0,
+      "newSize must be positive"
+    );
     const Jimp = yield getJimp();
     const image = yield Jimp.read(inputData);
     const { width, height } = image.bitmap;
     if (!width || !height) {
       throw Error("Undefined width or height from the input image.");
     }
-    const finalNewSize = newSize || calculateNewDimensions(width, height);
-    image.resize(
-      finalNewSize.width,
-      finalNewSize.height,
-      Jimp.RESIZE_NEAREST_NEIGHBOR
-    );
+    if (newSize.width === width && newSize.height === height) {
+      return inputData;
+    }
+    image.resize(newSize.width, newSize.height, Jimp.RESIZE_NEAREST_NEIGHBOR);
     image.quality(90);
     const resizedBuffer = yield image.getBufferAsync(Jimp.MIME_JPEG);
     return resizedBuffer;
   });
 }
-function resizeImgBase64(inputData, newSize) {
+function resizeImgBase64(inputBase64, newSize) {
   return __async(this, null, function* () {
     const splitFlag = ";base64,";
-    const dataSplitted = inputData.split(splitFlag);
+    const dataSplitted = inputBase64.split(splitFlag);
     if (dataSplitted.length !== 2) {
       throw Error("Invalid base64 data");
     }
@@ -191,7 +194,7 @@ function resizeImgBase64(inputData, newSize) {
     return `${dataSplitted[0]}${splitFlag}${content}`;
   });
 }
-function calculateNewDimensions(originalWidth, originalHeight) {
+function zoomForGPT4o(originalWidth, originalHeight) {
   const maxWidth = 2048;
   const maxHeight = 768;
   let newWidth = originalWidth;
@@ -237,7 +240,7 @@ function trimImage(image) {
 }
 
 // src/img/box-select.ts
-var import_node_assert2 = __toESM(require("assert"));
+var import_node_assert3 = __toESM(require("assert"));
 var cachedFont = null;
 var createSvgOverlay = (elements, imageWidth, imageHeight) => __async(void 0, null, function* () {
   const Jimp = yield getJimp();
@@ -343,7 +346,7 @@ var createSvgOverlay = (elements, imageWidth, imageHeight) => __async(void 0, nu
   return image;
 });
 var compositeElementInfoImg = (options) => __async(void 0, null, function* () {
-  (0, import_node_assert2.default)(options.inputImgBase64, "inputImgBase64 is required");
+  (0, import_node_assert3.default)(options.inputImgBase64, "inputImgBase64 is required");
   let width = 0;
   let height = 0;
   let jimpImage;
@@ -393,7 +396,7 @@ var compositeElementInfoImg = (options) => __async(void 0, null, function* () {
 });
 var processImageElementInfo = (options) => __async(void 0, null, function* () {
   const base64Image = options.inputImgBase64.split(";base64,").pop();
-  (0, import_node_assert2.default)(base64Image, "base64Image is undefined");
+  (0, import_node_assert3.default)(base64Image, "base64Image is undefined");
   const [
     compositeElementInfoImgBase64,
     compositeElementInfoImgWithoutTextBase64
@@ -459,7 +462,6 @@ function savePositionImg(options) {
   base64Encoded,
   base64ToPngFormat,
   bufferFromBase64,
-  calculateNewDimensions,
   compositeElementInfoImg,
   drawBoxOnImage,
   imageInfo,
@@ -470,5 +472,6 @@ function savePositionImg(options) {
   saveBase64Image,
   savePositionImg,
   transformImgPathToBase64,
-  trimImage
+  trimImage,
+  zoomForGPT4o
 });

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@midscene/shared",
-  "version": "0.11.2",
+  "version": "0.11.4-beta-20250219074605.0",
   "repository": "https://github.com/web-infra-dev/midscene",
   "homepage": "https://midscenejs.com/",
   "types": "./src/index.ts",

package/src/img/index.ts

@@ -7,10 +7,10 @@ export {
 } from './info';
 export {
   trimImage,
-  calculateNewDimensions,
   resizeImg,
   resizeImgBase64,
   transformImgPathToBase64,
+  zoomForGPT4o,
   saveBase64Image,
 } from './transform';
 export { processImageElementInfo, compositeElementInfoImg } from './box-select';

package/src/img/transform.ts

@@ -1,3 +1,4 @@
+import assert from 'node:assert';
 import { Buffer } from 'node:buffer';
 import getJimp from './get-jimp';
 
@@ -49,7 +50,7 @@ export async function transformImgPathToBase64(inputPath: string) {
  */
 export async function resizeImg(
   inputData: Buffer,
-  newSize?: {
+  newSize: {
     width: number;
     height: number;
   },
@@ -57,6 +58,11 @@ export async function resizeImg(
   if (typeof inputData === 'string')
     throw Error('inputData is base64, use resizeImgBase64 instead');
 
+  assert(
+    newSize && newSize.width > 0 && newSize.height > 0,
+    'newSize must be positive',
+  );
+
   const Jimp = await getJimp();
   const image = await Jimp.read(inputData);
   const { width, height } = image.bitmap;
@@ -65,13 +71,11 @@ export async function resizeImg(
     throw Error('Undefined width or height from the input image.');
   }
 
-  const finalNewSize = newSize || calculateNewDimensions(width, height);
+  if (newSize.width === width && newSize.height === height) {
+    return inputData;
+  }
 
-  image.resize(
-    finalNewSize.width,
-    finalNewSize.height,
-    Jimp.RESIZE_NEAREST_NEIGHBOR,
-  );
+  image.resize(newSize.width, newSize.height, Jimp.RESIZE_NEAREST_NEIGHBOR);
   image.quality(90);
   const resizedBuffer = await image.getBufferAsync(Jimp.MIME_JPEG);
 
@@ -79,14 +83,14 @@ export async function resizeImg(
 }
 
 export async function resizeImgBase64(
-  inputData: string,
-  newSize?: {
+  inputBase64: string,
+  newSize: {
     width: number;
     height: number;
   },
 ): Promise<string> {
   const splitFlag = ';base64,';
-  const dataSplitted = inputData.split(splitFlag);
+  const dataSplitted = inputBase64.split(splitFlag);
   if (dataSplitted.length !== 2) {
     throw Error('Invalid base64 data');
   }
@@ -109,10 +113,7 @@ export async function resizeImgBase64(
  * @returns {Object} An object containing the new width and height.
  * @throws {Error} Throws an error if the width or height is not a positive number.
  */
-export function calculateNewDimensions(
-  originalWidth: number,
-  originalHeight: number,
-) {
+export function zoomForGPT4o(originalWidth: number, originalHeight: number) {
   // In low mode, the image is scaled to 512x512 pixels and 85 tokens are used to represent the image.
   // In high mode, the model looks at low-resolution images and then creates detailed crop images, using 170 tokens for each 512x512 pixel tile. In practical applications, it is recommended to control the image size within 2048x768 pixels
   const maxWidth = 2048; // Maximum width