assistme 0.9.0 → 0.9.1

package/dist/{chunk-E27P57PM.js → chunk-O5K33FBW.js}

@@ -318,6 +318,7 @@ var CDP_SETTLE_MS = 300;
 var PAGE_TRANSITION_MS = 500;
 var MAX_WAIT_MS = 5e3;
 var SCREENSHOT_QUALITY = 80;
+var SCREENSHOT_MAX_DIM_PX = 1568;
 function delay(ms) {
   return new Promise((r) => {
     setTimeout(r, ms);
@@ -809,6 +810,56 @@ async function tryDismissOverlay(conn) {
   }
 }
 
+// src/browser/screenshot.ts
+async function getViewportMetrics(conn) {
+  try {
+    const result = await conn.send("Runtime.evaluate", {
+      expression: `JSON.stringify({ width: window.innerWidth, height: window.innerHeight, dpr: window.devicePixelRatio || 1 })`,
+      returnByValue: true
+    });
+    const value = result.result?.value;
+    if (value) {
+      const parsed = JSON.parse(value);
+      const width = Number(parsed.width) || 0;
+      const height = Number(parsed.height) || 0;
+      const dpr = Number(parsed.dpr) || 1;
+      if (width > 0 && height > 0) {
+        return { width, height, dpr };
+      }
+    }
+  } catch {
+  }
+  return { width: 0, height: 0, dpr: 1 };
+}
+async function captureBoundedScreenshot(conn, maxDim = SCREENSHOT_MAX_DIM_PX) {
+  const { width, height, dpr } = await getViewportMetrics(conn);
+  if (width === 0 || height === 0) {
+    const result2 = await conn.send("Page.captureScreenshot", {
+      format: "png",
+      quality: SCREENSHOT_QUALITY,
+      captureBeyondViewport: false
+    });
+    return result2.data || "";
+  }
+  const naturalMax = Math.max(width, height) * dpr;
+  if (naturalMax <= maxDim) {
+    const result2 = await conn.send("Page.captureScreenshot", {
+      format: "png",
+      quality: SCREENSHOT_QUALITY,
+      captureBeyondViewport: false
+    });
+    return result2.data || "";
+  }
+  const scale = Math.min(dpr, maxDim / Math.max(width, height));
+  const result = await conn.send("Page.captureScreenshot", {
+    format: "png",
+    quality: SCREENSHOT_QUALITY,
+    captureBeyondViewport: false,
+    clip: { x: 0, y: 0, width, height, scale }
+  });
+  return result.data || "";
+}
+
 // src/browser/actions.ts
 var Actions = class {
   constructor(conn, snapshotEngine) {
@@ -1004,12 +1055,7 @@ var Actions = class {
     let screenshot;
     if (takeScreenshot) {
       await delay(CDP_SETTLE_MS);
-      const screenshotResult = await this.conn.send("Page.captureScreenshot", {
-        format: "png",
-        quality: SCREENSHOT_QUALITY,
-        captureBeyondViewport: false
-      });
-      screenshot = screenshotResult.data || "";
+      screenshot = await captureBoundedScreenshot(this.conn);
     }
     return { results, screenshot };
   }
@@ -1378,12 +1424,7 @@ URL: ${info.url}`;
   }
   async screenshot() {
     this.conn.ensureConnected();
-    const result = await this.conn.send("Page.captureScreenshot", {
-      format: "png",
-      quality: SCREENSHOT_QUALITY,
-      captureBeyondViewport: false
-    });
-    return result.data || "";
+    return captureBoundedScreenshot(this.conn);
   }
   async detectLoginPage() {
     try {
@@ -1833,12 +1874,7 @@ var SnapshotEngine = class {
         expression: buildAnnotationOverlayJS(refsJson)
       });
     }
-    const screenshotResult = await this.conn.send("Page.captureScreenshot", {
-      format: "png",
-      quality: SCREENSHOT_QUALITY,
-      captureBeyondViewport: false
-    });
-    const image = screenshotResult.data || "";
+    const image = await captureBoundedScreenshot(this.conn);
     if (annotate) {
       await this.conn.send("Runtime.evaluate", {
         expression: `(function() { var el = document.getElementById('__assistme_refs__'); if (el) el.remove(); })()`

package/dist/index.js

@@ -34,7 +34,7 @@ import {
   setSessionBusy,
   toggleScheduledTask,
   updateHeartbeat
-} from "./chunk-E27P57PM.js";
+} from "./chunk-O5K33FBW.js";
 import {
   JobRunner,
   callMcpHandler

package/dist/workers/entry.js

@@ -2,7 +2,7 @@
 import {
   TaskProcessor,
   getBrowser
-} from "../chunk-E27P57PM.js";
+} from "../chunk-O5K33FBW.js";
 import "../chunk-RRMI6RDG.js";
 import {
   setLogTransport

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "assistme",
-  "version": "0.9.0",
+  "version": "0.9.1",
   "description": "AssistMe CLI Agent - AI-powered agentic assistant for code, browser, and automation",
   "type": "module",
   "main": "dist/index.js",

package/src/browser/actions.ts

@@ -25,9 +25,9 @@ import {
   KEY_PRESS_DELAY_MS,
   MAX_WAIT_MS,
   PAGE_TRANSITION_MS,
-  SCREENSHOT_QUALITY,
   SCROLL_SETTLE_MS,
 } from "./delays.js";
+import { captureBoundedScreenshot } from "./screenshot.js";
 import { clearFieldScript } from "./scripts.js";
 import type { SnapshotEngine } from "./snapshot.js";
 import type { ActionResult, ActionSpec, CDPEvalResult, RefActionResult } from "./types.js";
@@ -268,12 +268,7 @@ export class Actions {
     let screenshot: string | undefined;
     if (takeScreenshot) {
       await delay(CDP_SETTLE_MS);
-      const screenshotResult = await this.conn.send("Page.captureScreenshot", {
-        format: "png",
-        quality: SCREENSHOT_QUALITY,
-        captureBeyondViewport: false,
-      });
-      screenshot = (screenshotResult as { data?: string }).data || "";
+      screenshot = await captureBoundedScreenshot(this.conn);
     }
 
     return { results, screenshot };

package/src/browser/delays.ts

@@ -29,6 +29,16 @@ export const MAX_WAIT_MS = 5000;
 /** JPEG quality for CDP screenshots (0-100). */
 export const SCREENSHOT_QUALITY = 80;
 
+/**
+ * Maximum pixels on either side of a captured screenshot.
+ *
+ * Anthropic's Messages API rejects requests where any image exceeds 2000px
+ * per side once a conversation contains many images. Browser-agent sessions
+ * accumulate screenshots quickly, so we cap at a safe value below that limit.
+ * 1568 is also Anthropic's recommended maximum for best model performance.
+ */
+export const SCREENSHOT_MAX_DIM_PX = 1568;
+
 /** Promise-based delay helper. */
 export function delay(ms: number): Promise<void> {
   return new Promise((r) => { setTimeout(r, ms); });

package/src/browser/navigation.ts

@@ -3,7 +3,7 @@
  */
 
 import type { CDPConnection } from "./connection.js";
-import { SCREENSHOT_QUALITY } from "./delays.js";
+import { captureBoundedScreenshot } from "./screenshot.js";
 import type { CDPEvalResult } from "./types.js";
 
 export class Navigation {
@@ -100,12 +100,7 @@ export class Navigation {
 
   async screenshot(): Promise<string> {
     this.conn.ensureConnected();
-    const result = await this.conn.send("Page.captureScreenshot", {
-      format: "png",
-      quality: SCREENSHOT_QUALITY,
-      captureBeyondViewport: false,
-    });
-    return (result as { data?: string }).data || "";
+    return captureBoundedScreenshot(this.conn);
   }
 
   async detectLoginPage(): Promise<{ isLoginPage: boolean; reason: string }> {

package/src/browser/screenshot.ts

@@ -0,0 +1,94 @@
+/**
+ * Screenshot helper — captures a page screenshot via CDP while ensuring
+ * output dimensions stay within Anthropic's per-image size limit.
+ *
+ * Without this cap, high-DPI displays (Retina, 2× / 3×) produce PNGs that
+ * exceed 2000px per side. Once a conversation accumulates many images the
+ * API rejects the request with:
+ *   "An image in the conversation exceeds the dimension limit for
+ *    many-image requests (2000px)."
+ *
+ * Implementation uses CDP's native `clip.scale` so no image library is
+ * needed. We read the viewport size and devicePixelRatio, then downscale
+ * only when the natural capture would exceed the cap.
+ */
+
+import type { CDPConnection } from "./connection.js";
+import { SCREENSHOT_MAX_DIM_PX, SCREENSHOT_QUALITY } from "./delays.js";
+import type { CDPEvalResult } from "./types.js";
+
+interface ViewportMetrics {
+  width: number;
+  height: number;
+  dpr: number;
+}
+
+async function getViewportMetrics(conn: CDPConnection): Promise<ViewportMetrics> {
+  try {
+    const result = await conn.send("Runtime.evaluate", {
+      expression: `JSON.stringify({ width: window.innerWidth, height: window.innerHeight, dpr: window.devicePixelRatio || 1 })`,
+      returnByValue: true,
+    });
+    const value = (result as CDPEvalResult).result?.value as string | undefined;
+    if (value) {
+      const parsed = JSON.parse(value) as Partial<ViewportMetrics>;
+      const width = Number(parsed.width) || 0;
+      const height = Number(parsed.height) || 0;
+      const dpr = Number(parsed.dpr) || 1;
+      if (width > 0 && height > 0) {
+        return { width, height, dpr };
+      }
+    }
+  } catch {
+    // Fall through to defaults
+  }
+  // Safe defaults if viewport query fails — no clip, CDP uses natural capture
+  return { width: 0, height: 0, dpr: 1 };
+}
+
+/**
+ * Capture a PNG screenshot of the current page, downsampled so the longer
+ * side does not exceed {@link SCREENSHOT_MAX_DIM_PX}. Returns base64 data.
+ */
+export async function captureBoundedScreenshot(
+  conn: CDPConnection,
+  maxDim: number = SCREENSHOT_MAX_DIM_PX
+): Promise<string> {
+  const { width, height, dpr } = await getViewportMetrics(conn);
+
+  // If we couldn't determine metrics, fall back to unclipped capture.
+  if (width === 0 || height === 0) {
+    const result = await conn.send("Page.captureScreenshot", {
+      format: "png",
+      quality: SCREENSHOT_QUALITY,
+      captureBeyondViewport: false,
+    });
+    return (result as { data?: string }).data || "";
+  }
+
+  // Natural output dimensions (what we'd get without clip): viewport CSS px × DPR.
+  const naturalMax = Math.max(width, height) * dpr;
+
+  // If the natural capture already fits under the cap, no scaling needed.
+  if (naturalMax <= maxDim) {
+    const result = await conn.send("Page.captureScreenshot", {
+      format: "png",
+      quality: SCREENSHOT_QUALITY,
+      captureBeyondViewport: false,
+    });
+    return (result as { data?: string }).data || "";
+  }
+
+  // CDP clip.scale applies to CSS pixels: output = cssPx × scale.
+  // Pick the largest scale that keeps max(output) ≤ maxDim, capped at dpr
+  // so we never upscale beyond the native render.
+  const scale = Math.min(dpr, maxDim / Math.max(width, height));
+
+  const result = await conn.send("Page.captureScreenshot", {
+    format: "png",
+    quality: SCREENSHOT_QUALITY,
+    captureBeyondViewport: false,
+    clip: { x: 0, y: 0, width, height, scale },
+  });
+  return (result as { data?: string }).data || "";
+}

package/src/browser/snapshot.ts

@@ -6,7 +6,7 @@ import { SNAPSHOT_LOAD_TIMEOUT_MS } from "../utils/constants.js";
 import { safeJsonParse } from "../utils/errors.js";
 import { log } from "../utils/logger.js";
 import type { CDPConnection } from "./connection.js";
-import { SCREENSHOT_QUALITY } from "./delays.js";
+import { captureBoundedScreenshot } from "./screenshot.js";
 import {
   buildAnnotationOverlayJS,
   buildFrameDiscoverJS,
@@ -83,13 +83,8 @@ export class SnapshotEngine {
       });
     }
 
-    // 3. Take screenshot
-    const screenshotResult = await this.conn.send("Page.captureScreenshot", {
-      format: "png",
-      quality: SCREENSHOT_QUALITY,
-      captureBeyondViewport: false,
-    });
-    const image = (screenshotResult as { data?: string }).data || "";
+    // 3. Take screenshot (bounded to stay within Anthropic's many-image dim limit)
+    const image = await captureBoundedScreenshot(this.conn);
 
     // 4. Remove overlay if injected
     if (annotate) {

package/tests/browser/snapshot.test.ts

@@ -12,6 +12,7 @@ vi.mock("../../src/utils/logger.js", () => ({
 
 vi.mock("../../src/browser/delays.js", () => ({
   SCREENSHOT_QUALITY: 80,
+  SCREENSHOT_MAX_DIM_PX: 1568,
 }));
 
 // Create mock connection