feat(core): calculate dpr in agent

Author: yuyutaotao

packages/core/src/agent/agent.ts

@@ -48,7 +48,7 @@ import {
   globalConfigManager,
   globalModelConfigManager,
 } from '@midscene/shared/env';
-import { resizeImgBase64 } from '@midscene/shared/img';
+import { imageInfoOfBase64, resizeImgBase64 } from '@midscene/shared/img';
 import { getDebug } from '@midscene/shared/logger';
 import { assert } from '@midscene/shared/utils';
 // import type { AndroidDeviceInputOpt } from '../device';
@@ -125,10 +125,15 @@ export class Agent<
   private hasWarnedNonVLModel = false;
 
   /**
-   * Screenshot scale factor for AI model processing
+   * Screenshot scale factor derived from actual screenshot dimensions
    */
   private screenshotScale?: number;
 
+  /**
+   * Internal promise to deduplicate screenshot scale computation
+   */
+  private screenshotScalePromise?: Promise<number>;
+
   // @deprecated use .interface instead
   get page() {
     return this.interface;
@@ -150,6 +155,52 @@ export class Agent<
     }
   }
 
+  /**
+   * Lazily compute the ratio between the physical screenshot width and the logical page width
+   */
+  private async getScreenshotScale(context: UIContext): Promise<number> {
+    if (this.screenshotScale !== undefined) {
+      return this.screenshotScale;
+    }
+
+    if (!this.screenshotScalePromise) {
+      this.screenshotScalePromise = (async () => {
+        const pageWidth = context.size?.width;
+        assert(
+          pageWidth && pageWidth > 0,
+          `Invalid page width when computing screenshot scale: ${pageWidth}`,
+        );
+
+        const { width: screenshotWidth } = await imageInfoOfBase64(
+          context.screenshotBase64,
+        );
+
+        assert(
+          Number.isFinite(screenshotWidth) && screenshotWidth > 0,
+          `Invalid screenshot width when computing screenshot scale: ${screenshotWidth}`,
+        );
+
+        const computedScale = screenshotWidth / pageWidth;
+        assert(
+          Number.isFinite(computedScale) && computedScale > 0,
+          `Invalid computed screenshot scale: ${computedScale}`,
+        );
+
+        debug(
+          `Computed screenshot scale ${computedScale} from screenshot width ${screenshotWidth} and page width ${pageWidth}`,
+        );
+        return computedScale;
+      })();
+    }
+
+    try {
+      this.screenshotScale = await this.screenshotScalePromise;
+      return this.screenshotScale;
+    } finally {
+      this.screenshotScalePromise = undefined;
+    }
+  }
+
   constructor(interfaceInstance: InterfaceType, opts?: AgentOpt) {
     this.interface = interfaceInstance;
     this.opts = Object.assign(
@@ -171,7 +222,6 @@ export class Agent<
       ? new ModelConfigManager(opts.modelConfig)
       : globalModelConfigManager;
 
-    this.screenshotScale = opts?.screenshotScale;
     this.onTaskStartTip = this.opts.onTaskStartTip;
 
     this.insight = new Insight(async (action: InsightAction) => {
@@ -226,33 +276,22 @@ export class Agent<
       });
     }
 
-    // Unified screenshot scaling: prioritize screenshotScale, otherwise use DPR
-    let targetWidth = context.size.width;
-    let targetHeight = context.size.height;
-    let needResize = false;
-
-    if (this.screenshotScale && this.screenshotScale !== 1) {
-      // User-specified scaling ratio
-      debug(`Applying user screenshot scale: ${this.screenshotScale}`);
-      targetWidth = Math.round(context.size.width * this.screenshotScale);
-      targetHeight = Math.round(context.size.height * this.screenshotScale);
-      needResize = true;
-    } else if (context.size.dpr && context.size.dpr !== 1) {
-      // No user-specified scaling, use DPR scaling to logical size
+    const computedScreenshotScale = await this.getScreenshotScale(context);
+
+    if (computedScreenshotScale !== 1) {
+      const scaleForLog = Number.parseFloat(computedScreenshotScale.toFixed(4));
       debug(
-        `Applying DPR scaling: ${context.size.dpr} (resize to logical size)`,
+        `Applying computed screenshot scale: ${scaleForLog} (resize to logical size)`,
       );
-      // Target is logical size, no need to change targetWidth/targetHeight
-      needResize = true;
-    }
-
-    // Execute scaling
-    if (needResize) {
+      const targetWidth = Math.round(context.size.width);
+      const targetHeight = Math.round(context.size.height);
       debug(`Resizing screenshot to ${targetWidth}x${targetHeight}`);
       context.screenshotBase64 = await resizeImgBase64(
         context.screenshotBase64,
         { width: targetWidth, height: targetHeight },
       );
+    } else {
+      debug(`screenshot scale=${computedScreenshotScale}`);
     }
 
     return context;
@@ -856,12 +895,18 @@ export class Agent<
 
     const { element } = output;
 
+    const dprValue = await (this.interface.size() as any).dpr;
+    const dprEntry = dprValue
+      ? {
+          dpr: dprValue,
+        }
+      : {};
     return {
       rect: element?.rect,
       center: element?.center,
-      dpr: (await this.interface.size()).dpr,
+      ...dprEntry,
     } as Pick<LocateResultElement, 'rect' | 'center'> & {
-      dpr: number;
+      dpr?: number; // this field is deprecated
     };
   }
 

packages/shared/src/types/index.ts

@@ -7,9 +7,9 @@ export interface Point {
 }
 
 export interface Size {
-  width: number; // logical pixel size
-  height: number;
-  dpr?: number; // dpr is the ratio of the physical pixel to the logical pixel. For example, the dpr is 2 when the screenshotBase64 returned is 2000x1000 when the logical width and height are 1000x500 here. Overriding the dpr will affect how the screenshotBase64 is resized before being sent to the AI model.
+  width: number; // The image sent to AI model will be resized to this width. usually you should set it to the logical pixel size
+  height: number; // The image sent to AI model will be resized to this height. usually you should set it to the logical pixel size
+  dpr?: number; // this is deprecated, do NOT use it
 }
 
 export type Rect = Point & Size & { zoom?: number };