@vm0/cli 9.104.0 → 9.105.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vm0/cli",
-  "version": "9.104.0",
+  "version": "9.105.0",
   "description": "CLI application",
   "repository": {
     "type": "git",

package/zero.js

@@ -115,7 +115,7 @@ import {
   upsertZeroOrgModelProvider,
   withErrorHandler,
   zeroAgentCustomSkillNameSchema
-} from "./chunk-FKDGKPGQ.js";
+} from "./chunk-2SZAWYHO.js";
 
 // src/zero.ts
 import { Command as Command77 } from "commander";
@@ -4735,8 +4735,19 @@ async function captureScreenshot() {
   const tmpPath = join3(tmpdir(), `vm0-screenshot-${randomUUID()}.jpg`);
   try {
     await execFileAsync("screencapture", ["-x", "-t", "jpg", tmpPath]);
-    const buffer = await readFile(tmpPath);
     const info = await getScreenInfo();
+    if (info.scaleFactor > 1) {
+      await execFileAsync("sips", [
+        "-z",
+        String(info.height),
+        String(info.width),
+        "-s",
+        "formatOptions",
+        "80",
+        tmpPath
+      ]);
+    }
+    const buffer = await readFile(tmpPath);
     return {
       image: buffer.toString("base64"),
       width: info.width,
@@ -4761,12 +4772,24 @@ async function captureRegionScreenshot(region) {
       regionArg,
       tmpPath
     ]);
+    const info = await getScreenInfo();
+    if (info.scaleFactor > 1) {
+      await execFileAsync("sips", [
+        "-z",
+        String(region.height),
+        String(region.width),
+        "-s",
+        "formatOptions",
+        "80",
+        tmpPath
+      ]);
+    }
     const buffer = await readFile(tmpPath);
     return {
       image: buffer.toString("base64"),
       width: region.width,
       height: region.height,
-      scaleFactor: 1,
+      scaleFactor: info.scaleFactor,
       format: "jpg"
     };
   } finally {
@@ -4786,14 +4809,30 @@ async function getScreenInfo() {
     for (const screen of screens) {
       const pixelStr = screen._spdisplays_pixels;
       if (pixelStr) {
-        const match = pixelStr.match(/(\d+)\s*x\s*(\d+)/);
-        if (match?.[1] && match[2]) {
-          const width = parseInt(match[1], 10);
-          const height = parseInt(match[2], 10);
+        const pixelMatch = pixelStr.match(/(\d+)\s*x\s*(\d+)/);
+        if (pixelMatch?.[1] && pixelMatch[2]) {
+          const physicalWidth = parseInt(pixelMatch[1], 10);
+          const physicalHeight = parseInt(pixelMatch[2], 10);
           const resStr = screen._spdisplays_resolution ?? screen.spdisplays_resolution ?? "";
-          const isRetina = /retina/i.test(resStr);
-          const scaleFactor = isRetina ? 2 : 1;
-          return { width, height, scaleFactor };
+          const resMatch = resStr.match(/(\d+)\s*x\s*(\d+)/);
+          let scaleFactor;
+          let logicalWidth;
+          let logicalHeight;
+          if (resMatch?.[1] && resMatch[2]) {
+            logicalWidth = parseInt(resMatch[1], 10);
+            logicalHeight = parseInt(resMatch[2], 10);
+            scaleFactor = Math.round(physicalWidth / logicalWidth);
+          } else {
+            const isRetina = /retina/i.test(resStr);
+            scaleFactor = isRetina ? 2 : 1;
+            logicalWidth = Math.floor(physicalWidth / scaleFactor);
+            logicalHeight = Math.floor(physicalHeight / scaleFactor);
+          }
+          return {
+            width: logicalWidth,
+            height: logicalHeight,
+            scaleFactor
+          };
         }
       }
     }
@@ -4908,17 +4947,20 @@ function parseKeyCombo(keys) {
   }
   return { modifiers, mainKey };
 }
+function keyAction(key) {
+  return VALID_SPECIAL_KEYS.has(key) ? `kp:${key}` : `t:${key}`;
+}
 async function pressKey(keys) {
   const { modifiers, mainKey } = parseKeyCombo(keys);
   if (modifiers.length === 0) {
-    await execFileAsync2("cliclick", [`kp:${mainKey}`]);
+    await execFileAsync2("cliclick", [keyAction(mainKey)]);
     return;
   }
   const args = [];
   for (const mod of modifiers) {
     args.push(`kd:${mod}`);
   }
-  args.push(`kp:${mainKey}`);
+  args.push(keyAction(mainKey));
   for (let i = modifiers.length - 1; i >= 0; i--) {
     args.push(`ku:${modifiers[i]}`);
   }
@@ -5101,12 +5143,10 @@ async function handleMouseRequest(req, res) {
       return;
     }
     const info = await getScreenInfo();
-    const maxX = Math.floor(info.width / info.scaleFactor);
-    const maxY = Math.floor(info.height / info.scaleFactor);
-    if (x < 0 || x >= maxX || y < 0 || y >= maxY) {
+    if (x < 0 || x >= info.width || y < 0 || y >= info.height) {
       res.writeHead(400, { "Content-Type": "text/plain" });
       res.end(
-        `Coordinates out of bounds. Screen size: ${maxX}x${maxY} (points)`
+        `Coordinates out of bounds. Screen size: ${info.width}x${info.height} (points)`
       );
       return;
     }
@@ -5677,31 +5717,23 @@ var clientCursorPositionCommand = new Command75().name("cursor-position").descri
 // src/commands/zero/computer-use/index.ts
 var hostCommand = new Command76().name("host").description("Manage computer-use host daemon").addCommand(hostStartCommand).addCommand(hostStopCommand);
 var clientCommand = new Command76().name("client").description("Interact with remote computer-use host").addCommand(clientScreenshotCommand).addCommand(clientZoomCommand).addCommand(clientInfoCommand).addCommand(clientLeftClickCommand).addCommand(clientRightClickCommand).addCommand(clientMiddleClickCommand).addCommand(clientDoubleClickCommand).addCommand(clientTripleClickCommand).addCommand(clientLeftClickDragCommand).addCommand(clientLeftMouseDownCommand).addCommand(clientLeftMouseUpCommand).addCommand(clientScrollCommand).addCommand(clientReadClipboardCommand).addCommand(clientWriteClipboardCommand).addCommand(clientKeyCommand).addCommand(clientHoldKeyCommand).addCommand(clientTypeCommand).addCommand(clientOpenAppCommand).addCommand(clientMouseMoveCommand).addCommand(clientCursorPositionCommand);
-var zeroComputerUseCommand = new Command76().name("computer-use").description("Remote desktop control for cloud agents").addCommand(hostCommand).addCommand(clientCommand).addHelpText(
+clientCommand.addHelpText(
   "after",
   `
+Coordinate System:
+  All coordinates use macOS logical points, not physical pixels.
+  On Retina displays, logical size = physical size / scaleFactor.
+  Run "info" to check your screen's logical dimensions.
+
 Examples:
-  Start the host daemon (on macOS):  zero computer-use host start
-  Stop the host daemon:              zero computer-use host stop
-  Take a screenshot (from agent):    zero computer-use client screenshot
-  Zoom into a region (from agent):   zero computer-use client zoom --x 0 --y 0 --width 500 --height 500
-  Get screen info (from agent):      zero computer-use client info
-  Left click at (500, 300):          zero computer-use client left-click 500 300
-  Double click at (100, 200):        zero computer-use client double-click 100 200
-  Drag from A to B:                  zero computer-use client left-click-drag 100 100 500 500
-  Press mouse button:                zero computer-use client left-mouse-down 200 300
-  Release mouse button:              zero computer-use client left-mouse-up 500 500
-  Scroll down at position:           zero computer-use client scroll 500 300 down 5
-  Read clipboard text:               zero computer-use client read-clipboard
-  Write clipboard text:              zero computer-use client write-clipboard "hello"
-  Press key combo:                   zero computer-use client key "cmd+c"
-  Hold shift for 2 seconds:          zero computer-use client hold-key "shift" 2000
-  Type text:                         zero computer-use client type "Hello, world!"
-  Open an application:               zero computer-use client open-app Safari
-  Open by bundle ID:                 zero computer-use client open-app "com.apple.Safari"
-  Move mouse to (100, 200):          zero computer-use client mouse-move 100 200
-  Get cursor position:               zero computer-use client cursor-position`
+  zero computer-use client screenshot
+  zero computer-use client zoom --x 0 --y 0 --width 500 --height 500
+  zero computer-use client info
+  zero computer-use client left-click 500 300
+  zero computer-use client scroll 500 300 down 5
+  zero computer-use client key "cmd+c"`
 );
+var zeroComputerUseCommand = new Command76().name("computer-use").description("Remote desktop control for cloud agents").addCommand(hostCommand).addCommand(clientCommand);
 
 // src/zero.ts
 var COMMAND_CAPABILITY_MAP = {
@@ -5753,7 +5785,7 @@ function registerZeroCommands(prog, commands) {
 var program = new Command77();
 program.name("zero").description(
   "Zero CLI \u2014 interact with the zero platform from inside the sandbox"
-).version("9.104.0").addHelpText(
+).version("9.105.0").addHelpText(
   "after",
   `
 Examples: