visionclaw 0.1.187-beta.9 → 0.1.187-dev.refactor-computer-use-direct-coordinates.1

package/dist-agent/bundle.cjs

@@ -14607,63 +14607,6 @@ var init_types = __esm({
 });
 
 // dist/config/index.js
-var config_exports = {};
-__export(config_exports, {
-  addCommandIdToPendingUpgrade: () => addCommandIdToPendingUpgrade,
-  buildPricingUrl: () => buildPricingUrl,
-  configExists: () => configExists,
-  consumePendingRestart: () => consumePendingRestart,
-  consumePendingUpgrade: () => consumePendingUpgrade,
-  deleteOnboardingSession: () => deleteOnboardingSession,
-  ensureConfigDir: () => ensureConfigDir,
-  getBaseDir: () => getBaseDir,
-  getClaudeAutoMemoryDir: () => getClaudeAutoMemoryDir,
-  getClaudeLocalSettingsFile: () => getClaudeLocalSettingsFile,
-  getConfigDir: () => getConfigDir,
-  getConfigFile: () => getConfigFile,
-  getDynamicMcpServersFile: () => getDynamicMcpServersFile,
-  getFailedUploadsDir: () => getFailedUploadsDir,
-  getMemoriesDir: () => getMemoriesDir,
-  getOnboardingSessionFile: () => getOnboardingSessionFile,
-  getOwnerFile: () => getOwnerFile,
-  getPlaywrightProfileDir: () => getPlaywrightProfileDir,
-  getProfile: () => getProfile,
-  getProfilePhotosDir: () => getProfilePhotosDir,
-  getProfileTunnelConfigPath: () => getProfileTunnelConfigPath,
-  getProfileTunnelCredentialsPath: () => getProfileTunnelCredentialsPath,
-  getProfileTunnelDir: () => getProfileTunnelDir,
-  getProfilesDir: () => getProfilesDir,
-  getScreenshotsDir: () => getScreenshotsDir,
-  getSessionFile: () => getSessionFile,
-  getSkillsDir: () => getSkillsDir,
-  getTOTPSecretsFile: () => getTOTPSecretsFile,
-  getTelegramAccessFile: () => getTelegramAccessFile,
-  getTokensFile: () => getTokensFile,
-  getUserCatalogDir: () => getUserCatalogDir,
-  isOnboardingComplete: () => isOnboardingComplete,
-  loadConfig: () => loadConfig,
-  loadDynamicMcpServers: () => loadDynamicMcpServers,
-  loadGoogleTokens: () => loadGoogleTokens,
-  loadOnboardingSessionId: () => loadOnboardingSessionId,
-  loadOwnerConfig: () => loadOwnerConfig,
-  loadSessionId: () => loadSessionId,
-  loadTOTPSecrets: () => loadTOTPSecrets,
-  loadTelegramAccessConfig: () => loadTelegramAccessConfig,
-  loadUsageSnapshot: () => loadUsageSnapshot,
-  ownerConfigExists: () => ownerConfigExists,
-  saveConfig: () => saveConfig,
-  saveDynamicMcpServers: () => saveDynamicMcpServers,
-  saveGoogleTokens: () => saveGoogleTokens,
-  saveOnboardingSessionId: () => saveOnboardingSessionId,
-  saveOwnerConfig: () => saveOwnerConfig,
-  saveSessionId: () => saveSessionId,
-  saveTOTPSecrets: () => saveTOTPSecrets,
-  saveTelegramAccessConfig: () => saveTelegramAccessConfig,
-  saveUsageSnapshot: () => saveUsageSnapshot,
-  setProfile: () => setProfile,
-  writePendingRestart: () => writePendingRestart,
-  writePendingUpgrade: () => writePendingUpgrade
-});
 function setProfile(name) {
   if (!/^[a-z0-9][a-z0-9-]*$/.test(name)) {
     throw new Error(`Invalid profile name "${name}". Use lowercase letters, numbers, and hyphens.`);
@@ -14676,9 +14619,6 @@ function getProfile() {
 function getBaseDir() {
   return BASE_DIR;
 }
-function getProfilesDir() {
-  return PROFILES_DIR;
-}
 function getConfigDir() {
   return import_node_path.default.join(PROFILES_DIR, currentProfile);
 }
@@ -15007,22 +14947,6 @@ function saveSessionId(sessionId, mode) {
   existing.updatedAt = (/* @__PURE__ */ new Date()).toISOString();
   import_node_fs.default.writeFileSync(sessionFile, JSON.stringify(existing, null, 2), "utf-8");
 }
-function loadOnboardingSessionId() {
-  const file2 = getOnboardingSessionFile();
-  if (!import_node_fs.default.existsSync(file2))
-    return null;
-  try {
-    const raw = JSON.parse(import_node_fs.default.readFileSync(file2, "utf-8"));
-    const parsed = raw;
-    return parsed.sessionId ?? null;
-  } catch {
-    return null;
-  }
-}
-function saveOnboardingSessionId(sessionId) {
-  ensureConfigDir();
-  import_node_fs.default.writeFileSync(getOnboardingSessionFile(), JSON.stringify({ sessionId, updatedAt: (/* @__PURE__ */ new Date()).toISOString() }, null, 2), "utf-8");
-}
 function deleteOnboardingSession() {
   const file2 = getOnboardingSessionFile();
   if (import_node_fs.default.existsSync(file2)) {
@@ -20445,7 +20369,7 @@ var require_prompts3 = __commonJS({
 
 // dist/utils/version-check.js
 function isBundled() {
-  const v16 = "0.1.187-beta.9";
+  const v16 = "0.1.187-dev.refactor-computer-use-direct-coordinates.1";
   return typeof v16 === "string" && v16 !== "undefined";
 }
 function getPackageRoot() {
@@ -20462,7 +20386,7 @@ function getInstallationInfo() {
   };
 }
 function getCurrentVersion() {
-  const bundledVersion = "0.1.187-beta.9";
+  const bundledVersion = "0.1.187-dev.refactor-computer-use-direct-coordinates.1";
   if (bundledVersion && bundledVersion !== "undefined") {
     return bundledVersion;
   }
@@ -24995,7 +24919,7 @@ var require_retry = __commonJS({
     Object.defineProperty(exports2, "__esModule", { value: true });
     exports2.getRetryConfig = getRetryConfig;
     async function getRetryConfig(err7) {
-      let config2 = getConfig4(err7);
+      let config2 = getConfig3(err7);
       if (!err7 || !err7.config || !config2 && !err7.config.retry) {
         return { shouldRetry: false };
       }
@@ -25047,7 +24971,7 @@ var require_retry = __commonJS({
     }
     function shouldRetryRequest(err7) {
       var _a7;
-      const config2 = getConfig4(err7);
+      const config2 = getConfig3(err7);
       if (err7.name === "AbortError" || ((_a7 = err7.error) === null || _a7 === void 0 ? void 0 : _a7.name) === "AbortError") {
         return false;
       }
@@ -25079,7 +25003,7 @@ var require_retry = __commonJS({
       }
       return true;
     }
-    function getConfig4(err7) {
+    function getConfig3(err7) {
       if (err7 && err7.config && err7.config.retryConfig) {
         return err7.config.retryConfig;
       }
@@ -686554,108 +686478,14 @@ function generateAndSetWallpaper(agentName) {
   cleanupOldWallpapers(dest);
   return dest;
 }
-function wallpaperHelperSwiftSource() {
-  return `
-import AppKit
-
-let args = CommandLine.arguments
-guard args.count >= 2 else {
-  fputs("Usage: wallpaper-helper set <path> | get\\n", stderr)
-  exit(1)
-}
-
-let command = args[1]
-
-switch command {
-case "set":
-  guard args.count >= 3 else {
-    fputs("Usage: wallpaper-helper set <imagePath>\\n", stderr)
-    exit(1)
-  }
-  let imagePath = args[2]
-  let url = URL(fileURLWithPath: imagePath)
-  do {
-    for screen in NSScreen.screens {
-      // Preserve the screen's existing display options (scaling, fill colour,
-      // clipping). Apple's docs note these come from System Settings; if we
-      // pass [:] we silently reset them to defaults, which can change the
-      // visual layout of the wallpaper on some setups.
-      let existingOptions = NSWorkspace.shared.desktopImageOptions(for: screen) ?? [:]
-      try NSWorkspace.shared.setDesktopImageURL(url, for: screen, options: existingOptions)
-    }
-  } catch {
-    fputs("Failed to set wallpaper: \\(error)\\n", stderr)
-    exit(1)
-  }
-
-case "get":
-  if let screen = NSScreen.main,
-     let url = NSWorkspace.shared.desktopImageURL(for: screen) {
-    print(url.path)
-  }
-
-default:
-  fputs("Unknown command: \\(command). Use 'set' or 'get'.\\n", stderr)
-  exit(1)
-}
-`;
-}
-function getHelperHash() {
-  _helperHash ??= (0, import_node_crypto3.createHash)("sha256").update(wallpaperHelperSwiftSource()).digest("hex").slice(0, 12);
-  return _helperHash;
-}
-function getWallpaperHelperBin() {
-  const hash3 = getHelperHash();
-  const binFile = import_node_path11.default.join(import_node_os7.default.tmpdir(), `visionclaw-wallpaper-helper-${hash3}`);
-  if (!import_node_fs14.default.existsSync(binFile)) {
-    const srcFile = import_node_path11.default.join(import_node_os7.default.tmpdir(), `visionclaw-wallpaper-helper-${hash3}.swift`);
-    import_node_fs14.default.writeFileSync(srcFile, wallpaperHelperSwiftSource(), "utf-8");
-    const compile = (0, import_node_child_process9.spawnSync)("swiftc", ["-O", "-o", binFile, srcFile], {
-      encoding: "utf-8",
-      timeout: 6e4,
-      stdio: ["pipe", "pipe", "pipe"]
-    });
-    try {
-      import_node_fs14.default.unlinkSync(srcFile);
-    } catch {
-    }
-    if (compile.status !== 0) {
-      throw new Error(`Wallpaper helper compilation failed: ${compile.stderr}`);
-    }
-  }
-  return binFile;
-}
 function setDesktopWallpaper(imagePath) {
   if (process.platform !== "darwin")
     return;
-  try {
-    const bin = getWallpaperHelperBin();
-    const result = (0, import_node_child_process9.spawnSync)(bin, ["set", imagePath], {
-      encoding: "utf-8",
-      timeout: 15e3,
-      stdio: ["pipe", "pipe", "pipe"]
-    });
-    if (result.status === 0)
-      return;
-  } catch {
-  }
   (0, import_node_child_process9.execSync)(`osascript -e 'tell application "Finder" to set desktop picture to POSIX file "${imagePath}"'`, { stdio: "ignore", timeout: 15e3 });
 }
 function getCurrentDesktopWallpaper() {
   if (process.platform !== "darwin")
     return "";
-  try {
-    const bin = getWallpaperHelperBin();
-    const result = (0, import_node_child_process9.spawnSync)(bin, ["get"], {
-      encoding: "utf-8",
-      timeout: 15e3,
-      stdio: ["pipe", "pipe", "pipe"]
-    });
-    if (result.status === 0 && result.stdout.trim()) {
-      return result.stdout.trim();
-    }
-  } catch {
-  }
   try {
     return (0, import_node_child_process9.execSync)(`osascript -e 'tell application "Finder" to get POSIX path of (desktop picture as alias)'`, { encoding: "utf-8", timeout: 15e3, stdio: ["pipe", "pipe", "pipe"] }).trim();
   } catch {
@@ -686679,7 +686509,7 @@ function ensureWallpaper(agentName) {
   }
   return false;
 }
-var import_node_child_process9, import_node_crypto3, import_node_fs14, import_node_os7, import_node_path11, WALLPAPER_META_FILENAME, BG_SETUP, BG_RUNNING, _rendererHash, MACOS_CODENAMES, _helperHash;
+var import_node_child_process9, import_node_crypto3, import_node_fs14, import_node_os7, import_node_path11, WALLPAPER_META_FILENAME, BG_SETUP, BG_RUNNING, _rendererHash, MACOS_CODENAMES;
 var init_generate_wallpaper = __esm({
   "dist/onboarding/generate-wallpaper.js"() {
     "use strict";
@@ -810608,7 +810438,7 @@ var require_retry4 = __commonJS({
     Object.defineProperty(exports2, "__esModule", { value: true });
     exports2.getRetryConfig = getRetryConfig;
     async function getRetryConfig(err7) {
-      let config2 = getConfig4(err7);
+      let config2 = getConfig3(err7);
       if (!err7 || !err7.config || !config2 && !err7.config.retry) {
         return { shouldRetry: false };
       }
@@ -810659,7 +810489,7 @@ var require_retry4 = __commonJS({
       return { shouldRetry: true, config: err7.config };
     }
     function shouldRetryRequest(err7) {
-      const config2 = getConfig4(err7);
+      const config2 = getConfig3(err7);
       if (err7.config.signal?.aborted && err7.code !== "TimeoutError" || err7.code === "AbortError") {
         return false;
       }
@@ -810691,7 +810521,7 @@ var require_retry4 = __commonJS({
       }
       return true;
     }
-    function getConfig4(err7) {
+    function getConfig3(err7) {
       if (err7 && err7.config && err7.config.retryConfig) {
         return err7.config.retryConfig;
       }
@@ -974873,7 +974703,7 @@ async function sendMessage(params) {
     label: "sendMessage"
   });
 }
-async function getConfig3(params) {
+async function getConfig2(params) {
   const rawText = await apiPostFetch({
     baseUrl: params.baseUrl,
     endpoint: "ilink/bot/getconfig",
@@ -974942,7 +974772,7 @@ var init_config_cache = __esm({
         if (shouldFetch) {
           let fetchOk = false;
           try {
-            const resp = await getConfig3({
+            const resp = await getConfig2({
               baseUrl: this.apiOpts.baseUrl,
               token: this.apiOpts.token,
               ilinkUserId: userId,
@@ -976538,9 +976368,9 @@ var init_weixin = __esm({
           setContextToken(deps.accountId, fromUserId, full.context_token);
           this.contextTokens.set(fromUserId, full.context_token);
         }
-        const cachedConfig2 = await deps.configManager.getForUser(fromUserId, full.context_token);
-        if (cachedConfig2.typingTicket) {
-          this.typingTickets.set(fromUserId, cachedConfig2.typingTicket);
+        const cachedConfig = await deps.configManager.getForUser(fromUserId, full.context_token);
+        if (cachedConfig.typingTicket) {
+          this.typingTickets.set(fromUserId, cachedConfig.typingTicket);
         }
         const textBody = extractTextBody(full.item_list);
         const attachments = [];
@@ -991680,9 +991510,7 @@ Even if you are using Playwright to operate the browser, you should bring the br
 
 ### Desktop Interactions (Computer Use)
 
-For desktop applications and UI outside the browser, use the available desktop/computer-use tools. Some runtimes expose a native computer tool, while others expose the \`computer_use_*\` tool family with AI-powered coordinate resolution.
-
-If you describe something but the computer use tool can not resolve the coordinates, try describe the target in more details. Also, the screenshot will always show the current cursor location.
+For desktop applications and UI outside the browser, use the available desktop/computer-use tools. The \`computer_use_*\` tools accept pixel coordinates from screenshots \u2014 always take a screenshot first with \`computer_use_screenshot\` to see the screen and determine coordinates, then use those coordinates for click/move/scroll/drag actions.
 
 **Prefer ${config2.browserBackend === "playwriter" ? "Playwriter" : "Playwright"} browser tools** for all in-browser work. Use desktop/computer-use tools only for desktop apps, situations outside the browser, or as a fallback when the browser tools can't do the job.
 
@@ -996065,84 +995893,6 @@ init_transcribe_audio();
 init_zod();
 init_sdk2();
 init_desktop_executor_factory();
-
-// dist/tools/coordinate-resolver.js
-init_logger();
-init_screenshot();
-init_client_factory();
-function clamp(n6, min, max) {
-  return Math.max(min, Math.min(max, n6));
-}
-async function resolveCoordinates(screenshotBase64, instruction, displayWidth, displayHeight, config2, screenshotFilePath) {
-  const dims = screenshotFilePath ? await getImageDimensionsFromFile(screenshotFilePath) : await getImageDimensions(Buffer.from(screenshotBase64, "base64"));
-  if (!dims) {
-    logger.warn("Could not read screenshot dimensions, aborting coordinate resolution");
-    return null;
-  }
-  const imgW = dims.width;
-  const imgH = dims.height;
-  const scaleX = displayWidth / imgW;
-  const scaleY = displayHeight / imgH;
-  logger.info(`Resolving coordinates: "${instruction}" (display ${displayWidth}x${displayHeight}, image ${imgW}x${imgH}, scaleX=${scaleX.toFixed(3)}, scaleY=${scaleY.toFixed(3)})`, { instruction, displayWidth, displayHeight, imgW, imgH, scaleX, scaleY });
-  const client = createClient(config2);
-  const response = await client.beta.messages.create({
-    model: getModelId(config2),
-    max_tokens: 1024,
-    betas: ["computer-use-2025-11-24"],
-    tools: [
-      {
-        type: "computer_20251124",
-        name: "computer",
-        display_width_px: imgW,
-        display_height_px: imgH,
-        display_number: 1
-      }
-    ],
-    system: "You are a helpful assistant that can resolve natural language targets to pixel coordinates using the computer tool. Call the tool directly, no questions or explanations.",
-    tool_choice: { type: "tool", name: "computer" },
-    messages: [
-      {
-        role: "user",
-        content: [
-          {
-            type: "image",
-            source: {
-              type: "base64",
-              media_type: "image/png",
-              data: screenshotBase64
-            }
-          },
-          {
-            type: "text",
-            text: instruction
-          }
-        ]
-      }
-    ]
-  });
-  for (const block of response.content) {
-    if (block.type === "tool_use" && block.name === "computer") {
-      const input = block.input;
-      const coord = input.coordinate;
-      const action = input.action ?? "left_click";
-      logger.debug(`Coordinate resolver raw response: action=${action} coordinate=${JSON.stringify(coord)}`, { instruction, action, coordinate: coord, fullInput: input });
-      if (Array.isArray(coord) && coord.length >= 2) {
-        const rawX = coord[0];
-        const rawY = coord[1];
-        if (typeof rawX === "number" && typeof rawY === "number" && Number.isFinite(rawX) && Number.isFinite(rawY)) {
-          const x14 = clamp(Math.round(rawX * scaleX), 0, displayWidth);
-          const y12 = clamp(Math.round(rawY * scaleY), 0, displayHeight);
-          logger.info(`Resolved coordinates: action=${action} raw=(${rawX},${rawY}) unscaled=(${x14},${y12}) scaleX=${scaleX.toFixed(3)} scaleY=${scaleY.toFixed(3)}`, { instruction, action, rawX, rawY, x: x14, y: y12, scaleX, scaleY });
-          return { action, x: x14, y: y12 };
-        }
-      }
-    }
-  }
-  logger.info(`Could not resolve coordinates: "${instruction}"`, { instruction });
-  return null;
-}
-
-// dist/tools/computer-use.js
 init_screenshot();
 init_logger();
 function requireDesktop() {
@@ -996165,48 +995915,32 @@ function requireDesktop() {
   }
   return null;
 }
-var cachedConfig = null;
-async function getConfig() {
-  if (cachedConfig)
-    return cachedConfig;
-  const { loadConfig: loadConfig2 } = await Promise.resolve().then(() => (init_config(), config_exports));
-  try {
-    cachedConfig = loadConfig2();
-    return cachedConfig;
-  } catch {
-    if (process.env.ANTHROPIC_API_KEY) {
-      cachedConfig = {
-        model: "claude-sonnet-4-6",
-        provider: "anthropic",
-        anthropicApiKey: process.env.ANTHROPIC_API_KEY
-      };
-      return cachedConfig;
-    }
-    return null;
-  }
+function getScreenshotSize(displaySize) {
+  const scale = getVisionScaleFactor(displaySize.width, displaySize.height);
+  if (scale >= 1)
+    return { ...displaySize };
+  return {
+    width: Math.max(1, Math.round(displaySize.width * scale)),
+    height: Math.max(1, Math.round(displaySize.height * scale))
+  };
 }
-async function resolveAndExecute(instruction, config2, fn2) {
-  const { base64: screenshot, filePath: screenshotPath } = await captureScreen3();
-  const size = getDisplaySize3();
-  const resolved = await resolveCoordinates(screenshot, instruction, size.width, size.height, config2, screenshotPath);
-  if (!resolved) {
-    return {
-      content: [
-        {
-          type: "text",
-          text: `Could not resolve coordinates for: ${instruction}. Please try again with a clearer instruction.`
-        }
-      ],
-      isError: true
-    };
-  }
-  await fn2(resolved.x, resolved.y);
-  await new Promise((resolve) => setTimeout(resolve, 1e3));
+function screenshotToDisplay(point, screenshotSize, displaySize) {
+  return {
+    x: Math.round(point.x / screenshotSize.width * displaySize.width),
+    y: Math.round(point.y / screenshotSize.height * displaySize.height)
+  };
+}
+var pointSchema = external_exports.object({
+  x: external_exports.number().min(0).describe("X pixel coordinate in the screenshot image (0 = left edge)"),
+  y: external_exports.number().min(0).describe("Y pixel coordinate in the screenshot image (0 = top edge)")
+});
+async function actionResult(text, delayMs = 1e3) {
+  await new Promise((resolve) => setTimeout(resolve, delayMs));
   const { base64: resultScreenshot } = await captureScreen3();
   logger.screenshot(resultScreenshot);
   return {
     content: [
-      { type: "text", text: `Executed at (${resolved.x}, ${resolved.y})` },
+      { type: "text", text },
       {
         type: "image",
         data: resultScreenshot,
@@ -996215,184 +995949,88 @@ async function resolveAndExecute(instruction, config2, fn2) {
     ]
   };
 }
-var computerUseClickTool = gl("computer_use_click", "Click on a UI element described in natural language (e.g. 'the Chrome icon', 'the submit button'). Resolves target via AI, then performs left/right/double-click.", {
-  target: external_exports.string().describe("Natural language description of the element to click"),
+function withDesktopCheck(toolName, fn2) {
+  return (async () => {
+    const err7 = requireDesktop();
+    if (err7)
+      return err7;
+    try {
+      return await fn2();
+    } catch (e11) {
+      logger.err(`${toolName} failed`, { error: String(e11) });
+      return {
+        content: [
+          {
+            type: "text",
+            text: `${toolName} failed: ${e11 instanceof Error ? e11.message : String(e11)}`
+          }
+        ],
+        isError: true
+      };
+    }
+  })();
+}
+var computerUseClickTool = gl("computer_use_click", "Click at a point on the desktop screen. Coordinates are pixel positions in the screenshot image returned by computer_use_screenshot. Supports left-click, right-click, and double-click.", {
+  point: pointSchema.describe("Pixel coordinates in the screenshot image to click"),
   button: external_exports.enum(["left", "right", "double"]).optional().default("left")
 }, async (args) => {
-  const err7 = requireDesktop();
-  if (err7)
-    return err7;
-  const cfg = await getConfig();
-  if (!cfg) {
-    return {
-      content: [{ type: "text", text: "Claude API credentials not configured" }],
-      isError: true
-    };
-  }
-  const instruction = args.button === "right" ? `Right-click on: ${args.target}` : args.button === "double" ? `Double-click on: ${args.target}` : `Click on: ${args.target}`;
-  const fn2 = args.button === "right" ? (x14, y12) => rightClick3(x14, y12) : args.button === "double" ? (x14, y12) => doubleClick3(x14, y12) : (x14, y12) => click3(x14, y12);
-  try {
-    return await resolveAndExecute(instruction, cfg, fn2);
-  } catch (e11) {
-    logger.err("computer_use_click failed", { error: String(e11) });
-    return {
-      content: [
-        {
-          type: "text",
-          text: `computer_use_click failed: ${e11 instanceof Error ? e11.message : String(e11)}`
-        }
-      ],
-      isError: true
-    };
-  }
+  return withDesktopCheck("computer_use_click", async () => {
+    const displaySize = getDisplaySize3();
+    const ssSize = getScreenshotSize(displaySize);
+    const abs = screenshotToDisplay(args.point, ssSize, displaySize);
+    if (args.button === "right") {
+      await rightClick3(abs.x, abs.y);
+    } else if (args.button === "double") {
+      await doubleClick3(abs.x, abs.y);
+    } else {
+      await click3(abs.x, abs.y);
+    }
+    return actionResult(`${args.button === "double" ? "Double-clicked" : args.button === "right" ? "Right-clicked" : "Clicked"} at screenshot (${Math.round(args.point.x)}, ${Math.round(args.point.y)}) \u2192 display (${abs.x}, ${abs.y})`);
+  });
 });
-var computerUseMoveTool = gl("computer_use_move", "Move the mouse cursor to a UI element described in natural language (e.g. 'the search bar', 'the file menu'). Resolves target via AI, then moves cursor without clicking.", {
-  target: external_exports.string().describe("Natural language description of the element to move the cursor to")
+var computerUseMoveTool = gl("computer_use_move", "Move the mouse cursor to a point on the desktop screen. Coordinates are pixel positions in the screenshot image.", {
+  point: pointSchema.describe("Pixel coordinates in the screenshot image to move the cursor to")
 }, async (args) => {
-  const err7 = requireDesktop();
-  if (err7)
-    return err7;
-  const cfg = await getConfig();
-  if (!cfg) {
-    return {
-      content: [{ type: "text", text: "Claude API credentials not configured" }],
-      isError: true
-    };
-  }
-  const instruction = `Move cursor to the center of: ${args.target}`;
-  try {
-    return await resolveAndExecute(instruction, cfg, async (x14, y12) => {
-      await moveTo3(x14, y12);
-    });
-  } catch (e11) {
-    logger.err("computer_use_move failed", { error: String(e11) });
-    return {
-      content: [
-        {
-          type: "text",
-          text: `computer_use_move failed: ${e11 instanceof Error ? e11.message : String(e11)}`
-        }
-      ],
-      isError: true
-    };
-  }
+  return withDesktopCheck("computer_use_move", async () => {
+    const displaySize = getDisplaySize3();
+    const ssSize = getScreenshotSize(displaySize);
+    const abs = screenshotToDisplay(args.point, ssSize, displaySize);
+    await moveTo3(abs.x, abs.y);
+    return actionResult(`Moved cursor to screenshot (${Math.round(args.point.x)}, ${Math.round(args.point.y)}) \u2192 display (${abs.x}, ${abs.y})`);
+  });
 });
-var computerUseScrollTool = gl("computer_use_scroll", "Scroll at a target location in a direction. Resolves the scroll target via AI, then scrolls.", {
-  target: external_exports.string().describe("Natural language description of where to scroll"),
+var computerUseScrollTool = gl("computer_use_scroll", "Scroll at a point on the desktop screen in a direction. Coordinates are pixel positions in the screenshot image.", {
+  point: pointSchema.describe("Pixel coordinates in the screenshot image for scroll origin"),
   direction: external_exports.enum(["up", "down", "left", "right"]),
   amount: external_exports.number().optional().default(3).describe("Scroll amount (default 3)")
 }, async (args) => {
-  const err7 = requireDesktop();
-  if (err7)
-    return err7;
-  const cfg = await getConfig();
-  if (!cfg) {
-    return {
-      content: [{ type: "text", text: "Claude API credentials not configured" }],
-      isError: true
-    };
-  }
-  const instruction = `Move cursor to the center of: ${args.target} (for scrolling)`;
-  const amount = args.amount;
-  try {
-    return await resolveAndExecute(instruction, cfg, async (x14, y12) => {
-      await scroll3(x14, y12, args.direction, amount);
-    });
-  } catch (e11) {
-    logger.err("computer_use_scroll failed", { error: String(e11) });
-    return {
-      content: [
-        {
-          type: "text",
-          text: `computer_use_scroll failed: ${e11 instanceof Error ? e11.message : String(e11)}`
-        }
-      ],
-      isError: true
-    };
-  }
+  return withDesktopCheck("computer_use_scroll", async () => {
+    const displaySize = getDisplaySize3();
+    const ssSize = getScreenshotSize(displaySize);
+    const abs = screenshotToDisplay(args.point, ssSize, displaySize);
+    await scroll3(abs.x, abs.y, args.direction, args.amount);
+    return actionResult(`Scrolled ${args.direction} at screenshot (${Math.round(args.point.x)}, ${Math.round(args.point.y)}) \u2192 display (${abs.x}, ${abs.y})`);
+  });
 });
-var computerUseDragTool = gl("computer_use_drag", "Drag from one element to another. Resolves both 'from' and 'to' via AI, then performs drag.", {
-  from: external_exports.string().describe("Natural language description of drag start"),
-  to: external_exports.string().describe("Natural language description of drag end")
+var computerUseDragTool = gl("computer_use_drag", "Drag from one point to another on the desktop screen. Coordinates are pixel positions in the screenshot image.", {
+  startPoint: pointSchema.describe("Pixel coordinates in the screenshot image (drag source)"),
+  endPoint: pointSchema.describe("Pixel coordinates in the screenshot image (drop target)")
 }, async (args) => {
-  const err7 = requireDesktop();
-  if (err7)
-    return err7;
-  const cfg = await getConfig();
-  if (!cfg) {
-    return {
-      content: [{ type: "text", text: "Claude API credentials not configured" }],
-      isError: true
-    };
-  }
-  const { base64: screenshot, filePath: screenshotPath } = await captureScreen3();
-  const size = getDisplaySize3();
-  const [fromResolved, toResolved] = await Promise.all([
-    resolveCoordinates(screenshot, `Move cursor to the center of: ${args.from} (for dragging)`, size.width, size.height, cfg, screenshotPath),
-    resolveCoordinates(screenshot, `Move cursor to the center of: ${args.to} (for dragging)`, size.width, size.height, cfg, screenshotPath)
-  ]);
-  if (!fromResolved) {
-    return {
-      content: [
-        {
-          type: "text",
-          text: `Could not resolve 'from' coordinates: ${args.from}. Please try again with a clearer instruction.`
-        }
-      ],
-      isError: true
-    };
-  }
-  if (!toResolved) {
-    return {
-      content: [
-        {
-          type: "text",
-          text: `Could not resolve 'to' coordinates: ${args.to}. Please try again with a clearer instruction.`
-        }
-      ],
-      isError: true
-    };
-  }
-  try {
-    await drag3(fromResolved.x, fromResolved.y, toResolved.x, toResolved.y);
-    await new Promise((resolve) => setTimeout(resolve, 1e3));
-    const { base64: resultScreenshot } = await captureScreen3();
-    logger.screenshot(resultScreenshot);
-    return {
-      content: [
-        {
-          type: "text",
-          text: `Dragged from (${fromResolved.x}, ${fromResolved.y}) to (${toResolved.x}, ${toResolved.y})`
-        },
-        {
-          type: "image",
-          data: resultScreenshot,
-          mimeType: "image/png"
-        }
-      ]
-    };
-  } catch (e11) {
-    logger.err("computer_use_drag failed", { error: String(e11) });
-    return {
-      content: [
-        {
-          type: "text",
-          text: `computer_use_drag failed: ${e11 instanceof Error ? e11.message : String(e11)}`
-        }
-      ],
-      isError: true
-    };
-  }
+  return withDesktopCheck("computer_use_drag", async () => {
+    const displaySize = getDisplaySize3();
+    const ssSize = getScreenshotSize(displaySize);
+    const start = screenshotToDisplay(args.startPoint, ssSize, displaySize);
+    const end = screenshotToDisplay(args.endPoint, ssSize, displaySize);
+    await drag3(start.x, start.y, end.x, end.y);
+    return actionResult(`Dragged from screenshot (${Math.round(args.startPoint.x)}, ${Math.round(args.startPoint.y)}) to (${Math.round(args.endPoint.x)}, ${Math.round(args.endPoint.y)}) \u2192 display (${start.x}, ${start.y}) to (${end.x}, ${end.y})`);
+  });
 });
 var computerUseTypeTool = gl("computer_use_type", "Type text into the focused field. Optionally clear the field first (replace) and/or press Enter after.", {
   text: external_exports.string().describe("Text to type"),
   replace: external_exports.boolean().optional().default(false).describe("If true, select-all and replace existing text first"),
   press_enter: external_exports.boolean().optional().default(false).describe("If true, press Enter after typing")
 }, async (args) => {
-  const err7 = requireDesktop();
-  if (err7)
-    return err7;
-  try {
+  return withDesktopCheck("computer_use_type", async () => {
     if (args.replace) {
       await pressKey3(process.platform === "win32" ? "ctrl+a" : "cmd+a");
     }
@@ -996400,78 +996038,33 @@ var computerUseTypeTool = gl("computer_use_type", "Type text into the focused fi
     if (args.press_enter) {
       await pressKey3("enter");
     }
-    await new Promise((resolve) => setTimeout(resolve, 1e3));
-    const { base64: resultScreenshot } = await captureScreen3();
-    logger.screenshot(resultScreenshot);
-    return {
-      content: [
-        { type: "text", text: `Typed text${args.press_enter ? " and pressed Enter" : ""}` },
-        {
-          type: "image",
-          data: resultScreenshot,
-          mimeType: "image/png"
-        }
-      ]
-    };
-  } catch (e11) {
-    logger.err("computer_use_type failed", { error: String(e11) });
-    return {
-      content: [
-        {
-          type: "text",
-          text: `computer_use_type failed: ${e11 instanceof Error ? e11.message : String(e11)}`
-        }
-      ],
-      isError: true
-    };
-  }
+    return actionResult(`Typed text${args.press_enter ? " and pressed Enter" : ""}`);
+  });
 });
 var computerUseKeyTool = gl("computer_use_key", "Press a key or key combination (e.g. 'enter', 'escape', 'cmd+s', 'ctrl+shift+t').", {
   key: external_exports.string().describe("Key or combo: 'enter', 'escape', 'cmd+s', 'ctrl+shift+t', etc.")
 }, async (args) => {
-  const err7 = requireDesktop();
-  if (err7)
-    return err7;
-  try {
+  return withDesktopCheck("computer_use_key", async () => {
     await pressKey3(args.key);
-    await new Promise((resolve) => setTimeout(resolve, 1e3));
-    const { base64: resultScreenshot } = await captureScreen3();
-    logger.screenshot(resultScreenshot);
-    return {
-      content: [
-        { type: "text", text: `Pressed: ${args.key}` },
-        {
-          type: "image",
-          data: resultScreenshot,
-          mimeType: "image/png"
-        }
-      ]
-    };
-  } catch (e11) {
-    logger.err("computer_use_key failed", { error: String(e11) });
-    return {
-      content: [
-        {
-          type: "text",
-          text: `computer_use_key failed: ${e11 instanceof Error ? e11.message : String(e11)}`
-        }
-      ],
-      isError: true
-    };
-  }
+    return actionResult(`Pressed: ${args.key}`);
+  });
 });
-var computerUseScreenshotTool = gl("computer_use_screenshot", "Capture a screenshot of the current desktop screen. Returns the image as base64 PNG and the file path where it is saved.", {}, async (_args) => {
+var computerUseScreenshotTool = gl("computer_use_screenshot", "Capture a screenshot of the current desktop screen. Returns the image as base64 PNG. Use the pixel coordinates from this image when calling click, move, scroll, or drag tools.", {}, async (_args) => {
   const err7 = requireDesktop();
   if (err7)
     return err7;
   try {
+    const displaySize = getDisplaySize3();
+    const ssSize = getScreenshotSize(displaySize);
     const { base64: base643, filePath } = await takeScreenshot();
     logger.screenshot(base643);
     return {
       content: [
         {
           type: "text",
-          text: `Screenshot saved to ${filePath}`
+          text: `Screenshot saved to ${filePath}
+Screenshot image size: ${ssSize.width}x${ssSize.height}px | Display resolution: ${displaySize.width}x${displaySize.height}px
+Use pixel coordinates from this image when calling click/move/scroll/drag tools.`
         },
         {
           type: "image",
@@ -996910,7 +996503,7 @@ async function ensureDisplaySize() {
     await initDisplaySize4();
   }
 }
-function getScreenshotSize(deviceSize) {
+function getScreenshotSize2(deviceSize) {
   const scale = getVisionScaleFactor(deviceSize.width, deviceSize.height);
   if (scale >= 1)
     return { ...deviceSize };
@@ -996925,11 +996518,11 @@ function screenshotToDevice(point, screenshotSize, deviceSize) {
     y: Math.round(point.y / screenshotSize.height * deviceSize.height)
   };
 }
-var pointSchema = external_exports.object({
+var pointSchema2 = external_exports.object({
   x: external_exports.number().min(0).describe("X pixel coordinate in the screenshot image (0 = left edge)"),
   y: external_exports.number().min(0).describe("Y pixel coordinate in the screenshot image (0 = top edge)")
 });
-async function actionResult(text, delayMs = 1e3) {
+async function actionResult2(text, delayMs = 1e3) {
   await new Promise((resolve) => setTimeout(resolve, delayMs));
   const { base64: resultScreenshot } = await captureScreen4();
   logger.screenshot(resultScreenshot);
@@ -996991,34 +996584,34 @@ function normalizeKey(raw) {
   return KEY_ALIASES[normalized] ?? normalized;
 }
 var androidUseClickTool = gl("android_use_click", "Click (tap) at a point on the Android device screen. Coordinates are pixel positions in the screenshot image returned by android_use_screenshot. Supports single tap, double tap, and button variants.", {
-  point: pointSchema.describe("Pixel coordinates in the screenshot image to click"),
+  point: pointSchema2.describe("Pixel coordinates in the screenshot image to click"),
   button: external_exports.enum(["left", "right", "middle"]).optional().default("left").describe("Click button (default: left). On Android, all buttons map to tap."),
   double: external_exports.boolean().optional().default(false).describe("If true, perform a double-tap instead of a single tap")
 }, async (args) => {
   return withDeviceCheck("android_use_click", async () => {
     await ensureDisplaySize();
     const deviceSize = getDisplaySize4();
-    const ssSize = getScreenshotSize(deviceSize);
+    const ssSize = getScreenshotSize2(deviceSize);
     const abs = screenshotToDevice(args.point, ssSize, deviceSize);
     if (args.double) {
       await doubleTap(abs.x, abs.y);
     } else {
       await tap(abs.x, abs.y);
     }
-    return actionResult(`${args.double ? "Double-clicked" : "Clicked"} at screenshot (${Math.round(args.point.x)}, ${Math.round(args.point.y)}) \u2192 device (${abs.x}, ${abs.y})`);
+    return actionResult2(`${args.double ? "Double-clicked" : "Clicked"} at screenshot (${Math.round(args.point.x)}, ${Math.round(args.point.y)}) \u2192 device (${abs.x}, ${abs.y})`);
   });
 });
 var androidUseLongPressTool = gl("android_use_long_press", "Long press at a point on the Android device screen. Coordinates are pixel positions in the screenshot image.", {
-  point: pointSchema.describe("Pixel coordinates in the screenshot image to long-press"),
+  point: pointSchema2.describe("Pixel coordinates in the screenshot image to long-press"),
   durationMs: external_exports.number().int().positive().optional().default(1e3).describe("Long press duration in milliseconds (default: 1000)")
 }, async (args) => {
   return withDeviceCheck("android_use_long_press", async () => {
     await ensureDisplaySize();
     const deviceSize = getDisplaySize4();
-    const ssSize = getScreenshotSize(deviceSize);
+    const ssSize = getScreenshotSize2(deviceSize);
     const abs = screenshotToDevice(args.point, ssSize, deviceSize);
     await longPress(abs.x, abs.y, args.durationMs);
-    return actionResult(`Long-pressed at screenshot (${Math.round(args.point.x)}, ${Math.round(args.point.y)}) \u2192 device (${abs.x}, ${abs.y}) for ${args.durationMs}ms`);
+    return actionResult2(`Long-pressed at screenshot (${Math.round(args.point.x)}, ${Math.round(args.point.y)}) \u2192 device (${abs.x}, ${abs.y}) for ${args.durationMs}ms`);
   });
 });
 var androidUseTypeTool = gl("android_use_type", "Type text into the currently focused input field on the Android device. Optionally replace existing text first (select all + delete before typing).", {
@@ -997027,18 +996620,18 @@ var androidUseTypeTool = gl("android_use_type", "Type text into the currently fo
 }, async (args) => {
   return withDeviceCheck("android_use_type", async () => {
     await typeText4(args.content, { replace: args.replace });
-    return actionResult(`Typed "${args.content.length > 50 ? args.content.slice(0, 50) + "..." : args.content}"${args.replace ? " (replaced existing text)" : ""}`);
+    return actionResult2(`Typed "${args.content.length > 50 ? args.content.slice(0, 50) + "..." : args.content}"${args.replace ? " (replaced existing text)" : ""}`);
   });
 });
 var androidUseScrollTool = gl("android_use_scroll", "Scroll the Android device screen in a direction from a point. Coordinates are pixel positions in the screenshot image. Amount is a float in [0, 1] representing scroll distance relative to screen size.", {
   direction: external_exports.enum(["up", "down", "left", "right"]).describe("Scroll direction"),
   amount: external_exports.number().min(0).max(1).describe("Scroll distance as fraction of screen size (0.0 to 1.0, e.g. 0.25 = quarter screen)"),
-  point: pointSchema.optional().describe("Pixel coordinates in the screenshot image for scroll origin (default: center of screen)")
+  point: pointSchema2.optional().describe("Pixel coordinates in the screenshot image for scroll origin (default: center of screen)")
 }, async (args) => {
   return withDeviceCheck("android_use_scroll", async () => {
     await ensureDisplaySize();
     const deviceSize = getDisplaySize4();
-    const ssSize = getScreenshotSize(deviceSize);
+    const ssSize = getScreenshotSize2(deviceSize);
     const ssPoint = args.point ?? {
       x: ssSize.width / 2,
       y: ssSize.height / 2
@@ -997065,22 +996658,22 @@ var androidUseScrollTool = gl("android_use_scroll", "Scroll the Android device s
     endX = Math.max(0, Math.min(deviceSize.width, endX));
     endY = Math.max(0, Math.min(deviceSize.height, endY));
     await swipe(origin.x, origin.y, endX, endY, 300);
-    return actionResult(`Scrolled ${args.direction} by ${(args.amount * 100).toFixed(0)}% from screenshot (${Math.round(ssPoint.x)}, ${Math.round(ssPoint.y)})`);
+    return actionResult2(`Scrolled ${args.direction} by ${(args.amount * 100).toFixed(0)}% from screenshot (${Math.round(ssPoint.x)}, ${Math.round(ssPoint.y)})`);
   });
 });
 var androidUseDragTool = gl("android_use_drag", "Drag from one point to another on the Android device. Coordinates are pixel positions in the screenshot image. Suitable for drag-and-drop operations (moving icons, reordering list items).", {
-  startPoint: pointSchema.describe("Pixel coordinates in the screenshot image (drag source)"),
-  endPoint: pointSchema.describe("Pixel coordinates in the screenshot image (drop target)"),
+  startPoint: pointSchema2.describe("Pixel coordinates in the screenshot image (drag source)"),
+  endPoint: pointSchema2.describe("Pixel coordinates in the screenshot image (drop target)"),
   durationMs: external_exports.number().int().positive().optional().default(800).describe("Drag duration in milliseconds (default: 800, longer than swipe for drag-and-drop)")
 }, async (args) => {
   return withDeviceCheck("android_use_drag", async () => {
     await ensureDisplaySize();
     const deviceSize = getDisplaySize4();
-    const ssSize = getScreenshotSize(deviceSize);
+    const ssSize = getScreenshotSize2(deviceSize);
     const start = screenshotToDevice(args.startPoint, ssSize, deviceSize);
     const end = screenshotToDevice(args.endPoint, ssSize, deviceSize);
     await swipe(start.x, start.y, end.x, end.y, args.durationMs);
-    return actionResult(`Dragged from screenshot (${Math.round(args.startPoint.x)}, ${Math.round(args.startPoint.y)}) to (${Math.round(args.endPoint.x)}, ${Math.round(args.endPoint.y)}) over ${args.durationMs}ms`);
+    return actionResult2(`Dragged from screenshot (${Math.round(args.startPoint.x)}, ${Math.round(args.startPoint.y)}) to (${Math.round(args.endPoint.x)}, ${Math.round(args.endPoint.y)}) over ${args.durationMs}ms`);
   });
 });
 var androidUsePressTool = gl("android_use_press", "Press a key or button on the Android device. Supports standard keys: enter, tab, escape, backspace, delete, space, arrow keys (up/down/left/right), mobile keys (home, back, menu, app_switch, volume_up, volume_down, power), and function keys (f1-f12).", {
@@ -997089,7 +996682,7 @@ var androidUsePressTool = gl("android_use_press", "Press a key or button on the
   return withDeviceCheck("android_use_press", async () => {
     const normalized = normalizeKey(args.key);
     await pressKey4(normalized);
-    return actionResult(`Pressed: ${normalized}${normalized !== args.key ? ` (normalized from "${args.key}")` : ""}`);
+    return actionResult2(`Pressed: ${normalized}${normalized !== args.key ? ` (normalized from "${args.key}")` : ""}`);
   });
 });
 var androidUseScreenshotTool = gl("android_use_screenshot", "Capture a screenshot of the connected Android device screen. Returns the image as base64 PNG. The image is resized for vision model processing. Use the pixel coordinates from this image when calling click, long_press, scroll, or drag tools.", {}, async (_args) => {
@@ -997099,7 +996692,7 @@ var androidUseScreenshotTool = gl("android_use_screenshot", "Capture a screensho
   try {
     await ensureDisplaySize();
     const deviceSize = getDisplaySize4();
-    const ssSize = getScreenshotSize(deviceSize);
+    const ssSize = getScreenshotSize2(deviceSize);
     const { base64: base643, filePath } = await captureScreen4();
     logger.screenshot(base643);
     return {
@@ -997137,7 +996730,7 @@ var androidUseOpenTool = gl("android_use_open", "Open (launch) an app on the And
     await new Promise((resolve) => setTimeout(resolve, 2e3));
     await ensureDisplaySize();
     const deviceSize = getDisplaySize4();
-    const ssSize = getScreenshotSize(deviceSize);
+    const ssSize = getScreenshotSize2(deviceSize);
     const { base64: resultScreenshot } = await captureScreen4();
     logger.screenshot(resultScreenshot);
     let activity = "";
@@ -1029367,7 +1028960,7 @@ var _config = void 0;
 function resetArchiveConfigCache() {
   _config = void 0;
 }
-function getConfig2() {
+function getConfig() {
   if (_config === void 0) {
     try {
       _config = buildArchiveConfig(loadConfig());
@@ -1029378,7 +1028971,7 @@ function getConfig2() {
   return _config;
 }
 function isArchiveEnabled() {
-  return getConfig2() !== null;
+  return getConfig() !== null;
 }
 var TEXT_CHUNK_KEYWORD = "vc-archive-id";
 function pngCrc32(data2) {
@@ -1029526,7 +1029119,7 @@ async function drainFailedUploads(config2) {
   }
 }
 function archiveImage(base64Data, mediaType, sessionId) {
-  const config2 = getConfig2();
+  const config2 = getConfig();
   if (!config2)
     return null;
   const now = /* @__PURE__ */ new Date();