eve 0.65.0 → 0.66.0

package/dist/src/extensions/code/extension/lib/computer-use-driver-source.d.ts

@@ -0,0 +1,14 @@
+export declare const SERVER_MJS_SOURCE = "import { setTimeout as delay } from \"node:timers/promises\";\nimport { execFile, execFileSync } from \"node:child_process\";\nimport { randomUUID } from \"node:crypto\";\nimport { chmod, rename, rm } from \"node:fs/promises\";\nimport { promisify } from \"node:util\";\nimport http from \"node:http\";\nimport { env } from \"node:process\";\nimport { dirname, resolve } from \"node:path\";\n\nimport { CuaDriver } from \"@trycua/cua-driver\";\n\nimport { pointerMotionPlan } from \"./pointer-motion.mjs\";\nimport { selectGroundingElements } from \"./state.mjs\";\nimport { naturalTypingPlan } from \"./typing.mjs\";\n\nconst socketPath = env.COMPUTER_USE_SOCKET_PATH;\nconst workspaceRoot = env.COMPUTER_USE_ROOT;\nconst maxRequestBytes = 32 * 1024;\nconst maxRequestMs = 120_000;\nconst maxRecordingMs = 300_000;\nconst nativeCallTimeoutMs = 10_000;\n// Only the queue owner sets this signal; recording finalization deliberately bypasses it.\nlet requestSignal;\nconst desktopMargin = 24;\nconst minimumWindow = { height: 240, width: 480 };\nconst layoutTolerance = { height: 8, width: 8, x: 4, y: 4 };\nconst maxClipboardBytes = 10_000;\nconst driver = await CuaDriver.create({ claudeCodeCompatibility: false });\nconst execFilePromise = promisify(execFile);\nfunction execFileAsync(file, args, options) {\n  return execFilePromise(file, args, {\n    timeout: nativeCallTimeoutMs,\n    ...options,\n    signal: requestSignal,\n  });\n}\n\nlet activeRecording = null;\nlet queue = Promise.resolve();\nlet shuttingDown = false;\n\nawait rm(socketPath, { force: true });\n\nconst server = http.createServer((request, response) => {\n  if (shuttingDown) return response.destroy();\n  if (request.method !== \"POST\" || request.url !== \"/call\") {\n    sendJson(response, 404, { error: \"not found\" });\n    return;\n  }\n  const controller = new AbortController();\n  const deadline = Number(request.headers[\"x-computer-use-deadline\"]);\n  const timeoutMs = Number.isFinite(deadline)\n    ? Math.max(0, Math.min(maxRequestMs, deadline - Date.now()))\n    : maxRequestMs;\n  const cancel = () =>\n    controller.abort(new Error(\"computer-use request cancelled or deadline exceeded\"));\n  const timer = setTimeout(cancel, timeoutMs);\n  response.once(\"close\", cancel);\n  controller.signal.addEventListener(\"abort\", () => response.destroy(), { once: true });\n  if (timeoutMs === 0) cancel();\n  // Read immediately: a body may finish or disconnect while another call owns the queue.\n  const body = readBody(request);\n  void body.catch(() => undefined);\n  const task = queue.then(async () => {\n    controller.signal.throwIfAborted();\n    if (shuttingDown) return;\n    requestSignal = controller.signal;\n    try {\n      await handleRequest(await body, response);\n    } finally {\n      requestSignal = undefined;\n    }\n  });\n  queue = task\n    .catch(() => response.destroy())\n    .finally(() => {\n      clearTimeout(timer);\n      response.removeListener(\"close\", cancel);\n    });\n});\n\nserver.listen(socketPath, () => {\n  void chmod(socketPath, 0o600);\n});\n\nfor (const signal of [\"SIGINT\", \"SIGTERM\"]) {\n  process.once(signal, () => {\n    void shutdown();\n  });\n}\n\nasync function shutdown() {\n  if (shuttingDown) return;\n  shuttingDown = true;\n  server.close();\n  server.closeAllConnections();\n  const timer = setTimeout(() => process.exit(1), nativeCallTimeoutMs * 2);\n  let exitCode = 0;\n  try {\n    await queue;\n    if (activeRecording !== null) await stopRecording();\n  } catch {\n    exitCode = 1;\n  } finally {\n    try {\n      await driver.shutdown();\n    } catch {\n      exitCode = 1;\n    } finally {\n      clearTimeout(timer);\n      process.exit(exitCode);\n    }\n  }\n}\n\nasync function handleRequest(body, response) {\n  try {\n    const input = JSON.parse(body);\n    if (input.action === \"health\") {\n      sendJson(response, 200, { backend: \"cua-driver\", version: \"0.12.5\" });\n      return;\n    }\n\n    validateRequest(input);\n    const actionResult = await performAction(input.action);\n    const state = await captureState(input.screenshotPath);\n    sendJson(response, 200, {\n      action: actionResult,\n      backend: \"cua-driver\",\n      screenshot: input.screenshotPath,\n      state,\n      version: \"0.12.5\",\n    });\n  } catch (error) {\n    if (requestSignal?.aborted && activeRecording !== null) {\n      await stopRecording().catch(() => undefined);\n    }\n    sendJson(response, 500, {\n      error: error instanceof Error ? error.message : String(error),\n    });\n  }\n}\n\nasync function performAction(action) {\n  requestSignal?.throwIfAborted();\n  validateActionCoordinates(action);\n  if (action.action === \"sequence\") {\n    const results = [];\n    try {\n      for (const nested of action.actions) {\n        const startedAt = performance.now();\n        const result = await performAction(nested);\n        results.push({\n          action: nested.action,\n          durationMs: Math.round(performance.now() - startedAt),\n          result,\n        });\n      }\n      return results;\n    } catch (error) {\n      // Never strand the durable driver in an active-recording state when a\n      // rehearsed step fails. Preserve the original action error.\n      if (activeRecording !== null) await stopRecording().catch(() => undefined);\n      throw error;\n    }\n  }\n  switch (action.action) {\n    case \"screenshot\":\n      return null;\n    case \"click\":\n      await moveCursor(action);\n      return click(action, 1);\n    case \"double_click\":\n      await moveCursor(action);\n      return click(action, 2);\n    case \"triple_click\":\n      await moveCursor(action);\n      return click(action, 3);\n    case \"drag\":\n      return dragCursor(action);\n    case \"keypress\":\n      return action.keys.length === 1\n        ? call(\"press_key\", {\n            scope: \"desktop\",\n            key: normalizeKey(action.keys[0]),\n          })\n        : call(\"hotkey\", {\n            scope: \"desktop\",\n            keys: action.keys.map(normalizeKey),\n          });\n    case \"move\":\n      return moveCursor(action);\n    case \"scroll\": {\n      await call(\"move_cursor\", {\n        scope: \"desktop\",\n        x: action.x,\n        y: action.y,\n      });\n      const results = [];\n      if (action.scrollY !== 0) {\n        results.push(await call(\"scroll\", scrollInput(action.scrollY, \"up\", \"down\")));\n      }\n      if (action.scrollX !== 0) {\n        results.push(await call(\"scroll\", scrollInput(action.scrollX, \"left\", \"right\")));\n      }\n      return results;\n    }\n    case \"type\":\n      return typeText(action);\n    case \"clipboard_read\":\n      return readClipboard(action.selection);\n    case \"clipboard_write\":\n      return writeClipboard(action.selection, action.text);\n    case \"wait\":\n      await sleep(action.durationMs ?? 2_000);\n      return null;\n    case \"arrange_windows\":\n      return arrangeWindows(action);\n    case \"record_start\":\n      return startRecording(action.path);\n    case \"record_stop\":\n      return stopRecording();\n    default:\n      throw new Error(`unsupported computer-use action: ${String(action.action)}`);\n  }\n}\n\nasync function captureState(screenshotPath) {\n  await call(\"get_desktop_state\", { screenshot_out_file: screenshotPath });\n  const pointer = structured(await call(\"get_cursor_position\", {}));\n  const screen = x11ScreenSize();\n  const windows = structured(await call(\"list_windows\", { on_screen_only: true })).windows;\n  const contentWindows = Array.isArray(windows) ? windows.filter(isContentWindow) : [];\n  const groundingWindow = contentWindows.toSorted((left, right) => right.z_index - left.z_index)[0];\n\n  let elements = [];\n  if (groundingWindow !== undefined) {\n    const windowState = structured(\n      await call(\"get_window_state\", {\n        pid: groundingWindow.pid,\n        window_id: groundingWindow.window_id,\n        include_screenshot: false,\n        max_depth: 20,\n        max_elements: 250,\n      }),\n    );\n    if (Array.isArray(windowState.elements)) {\n      elements = selectGroundingElements(windowState.elements, {\n        width: Number(screen.width) || 1920,\n        height: Number(screen.height) || 1080,\n      });\n    }\n  }\n\n  return formatState(groundingWindow, pointer, elements);\n}\n\nasync function startRecording(path) {\n  if (activeRecording !== null) {\n    throw new Error(\"a recording is already active\");\n  }\n\n  const outputDirectory = `${workspaceRoot}/recording-${randomUUID()}`;\n  const result = await call(\n    \"start_recording\",\n    {\n      output_dir: outputDirectory,\n      record_video: true,\n    },\n    false,\n  );\n  activeRecording = {\n    outputDirectory,\n    outputPath: path,\n    stopping: null,\n    timer: setTimeout(() => {\n      void stopRecording().catch(() => {\n        void shutdown();\n      });\n    }, maxRecordingMs),\n  };\n  return result;\n}\n\nasync function stopRecording() {\n  if (activeRecording === null) {\n    throw new Error(\"no recording is active\");\n  }\n\n  const recording = activeRecording;\n  if (recording.stopping !== null) return recording.stopping;\n  clearTimeout(recording.timer);\n  recording.stopping = finalizeRecording(recording);\n  return recording.stopping;\n}\n\nasync function finalizeRecording(recording) {\n  try {\n    const result = await call(\"stop_recording\", {}, false);\n    const output = structured(result);\n    const videoPath =\n      typeof output.last_video_path === \"string\"\n        ? output.last_video_path\n        : recording.outputDirectory + \"/recording.mp4\";\n    await rename(videoPath, recording.outputPath);\n    await rm(recording.outputDirectory, { force: true, recursive: true });\n    return result;\n  } catch (error) {\n    void shutdown();\n    throw error;\n  } finally {\n    activeRecording = null;\n  }\n}\n\nasync function call(name, input, cancellable = true) {\n  const signal = cancellable ? requestSignal : undefined;\n  signal?.throwIfAborted();\n  let timer;\n  const result = await Promise.race([\n    driver.callTool(name, JSON.stringify(input)),\n    new Promise((_, reject) => {\n      timer = setTimeout(() => {\n        // Native calls have no abort API. Do not reuse a driver with uncertain in-flight work.\n        void shutdown();\n        reject(new Error(\"computer-use native call timed out; restart with startComputerUse\"));\n      }, nativeCallTimeoutMs);\n    }),\n  ]).finally(() => clearTimeout(timer));\n  signal?.throwIfAborted();\n  if (result.isError) {\n    throw new Error(\n      `Cua ${name} failed${result.errorCode ? ` (${result.errorCode})` : \"\"}: ${result.text}`,\n    );\n  }\n  return {\n    degraded: result.degraded,\n    structuredJson: result.structuredJson,\n    text: result.text,\n    verified: result.verified,\n  };\n}\n\nfunction structured(result) {\n  if (result.structuredJson === undefined) return {};\n  return JSON.parse(result.structuredJson);\n}\n\nfunction pointerClickInput(x, y, button, count) {\n  return {\n    scope: \"desktop\",\n    x,\n    y,\n    button: button === \"wheel\" ? \"middle\" : button,\n    count,\n  };\n}\n\nasync function moveCursor(action) {\n  if (action.movementStyle === undefined) {\n    return call(\"move_cursor\", { scope: \"desktop\", x: action.x, y: action.y });\n  }\n  const current = structured(await call(\"get_cursor_position\", {}));\n  const start = { x: Number(current.x), y: Number(current.y) };\n  if (!Number.isFinite(start.x) || !Number.isFinite(start.y)) {\n    return call(\"move_cursor\", { scope: \"desktop\", x: action.x, y: action.y });\n  }\n  const options = { style: action.movementStyle };\n  if (action.durationMs !== undefined) options.durationMs = action.durationMs;\n  return followPointerPlan(pointerMotionPlan(start, { x: action.x, y: action.y }, options), \"x11\");\n}\n\nasync function dragCursor(action) {\n  const start = action.path[0];\n  const end = action.path.at(-1);\n  if (action.movementStyle === undefined) {\n    return call(\"drag\", {\n      scope: \"desktop\",\n      from_x: start.x,\n      from_y: start.y,\n      to_x: end.x,\n      to_y: end.y,\n      steps: Math.min(200, Math.max(20, action.path.length * 10)),\n    });\n  }\n  const style = action.movementStyle;\n  const current = structured(await call(\"get_cursor_position\", {}));\n  const currentPoint = { x: Number(current.x), y: Number(current.y) };\n  const approach =\n    Number.isFinite(currentPoint.x) && Number.isFinite(currentPoint.y)\n      ? pointerPlanArgs(pointerMotionPlan(currentPoint, start, { style: \"natural\" }))\n      : [\"mousemove\", String(start.x), String(start.y)];\n  const options = { style };\n  if (action.durationMs !== undefined) options.durationMs = action.durationMs;\n  const drag = pointerPlanArgs(pointerMotionPlan(start, end, options));\n  return xdotool([\n    ...approach,\n    \"sleep\",\n    \"0.025\",\n    \"mousedown\",\n    \"1\",\n    \"sleep\",\n    \"0.030\",\n    ...drag,\n    \"sleep\",\n    \"0.030\",\n    \"mouseup\",\n    \"1\",\n  ]);\n}\n\nasync function followPointerPlan(plan, backend = \"cua\") {\n  if (backend === \"x11\") return xdotool(pointerPlanArgs(plan));\n  const results = [];\n  for (let phaseIndex = 0; phaseIndex < plan.phases.length; phaseIndex += 1) {\n    for (const point of plan.phases[phaseIndex]) {\n      results.push(await call(\"move_cursor\", { scope: \"desktop\", x: point.x, y: point.y }));\n      if (point.delayMs > 0) await sleep(point.delayMs);\n    }\n    if (phaseIndex < plan.phases.length - 1 && plan.settleMs > 0) await sleep(plan.settleMs);\n  }\n  if (plan.phases.length === 1 && plan.settleMs > 0) await sleep(plan.settleMs);\n  return results;\n}\n\nfunction pointerPlanArgs(plan) {\n  const args = [];\n  for (let phaseIndex = 0; phaseIndex < plan.phases.length; phaseIndex += 1) {\n    for (const point of plan.phases[phaseIndex]) {\n      args.push(\"mousemove\", String(point.x), String(point.y));\n      if (point.delayMs > 0) args.push(\"sleep\", (point.delayMs / 1000).toFixed(3));\n    }\n    if (phaseIndex < plan.phases.length - 1 && plan.settleMs > 0) {\n      args.push(\"sleep\", (plan.settleMs / 1000).toFixed(3));\n    }\n  }\n  if (plan.phases.length === 1 && plan.settleMs > 0) {\n    args.push(\"sleep\", (plan.settleMs / 1000).toFixed(3));\n  }\n  return args;\n}\n\nasync function xdotool(args) {\n  const plannedSleepMs = args.reduce(\n    (total, value, index) => (args[index - 1] === \"sleep\" ? total + Number(value) * 1000 : total),\n    0,\n  );\n  const timeout = Math.max(2_000, Math.ceil(plannedSleepMs + 1_500));\n  const result = await execFileAsync(\"xdotool\", args, { env: displayEnv(), timeout });\n  return {\n    plannedSleepMs: Math.round(plannedSleepMs),\n    stderr: result.stderr,\n    stdout: result.stdout,\n  };\n}\n\nfunction typingPlan(action) {\n  if (action.typingStyle === \"natural\") return naturalTypingPlan(action.text);\n  return [...new Intl.Segmenter().segment(action.text)].map(({ segment }) => ({\n    grapheme: segment,\n    delayMs: action.typingDelayMs ?? 0,\n  }));\n}\n\nfunction validateTypingBudget(actions) {\n  let characters = 0;\n  let calls = 0;\n  let delayMs = 0;\n  for (const action of actions) {\n    if (action.action !== \"type\") continue;\n    if (typeof action.text !== \"string\" || action.text.length > 10_000) {\n      throw new Error(\"typing text must be at most 10000 characters\");\n    }\n    characters += action.text.length;\n    if (action.typingStyle === undefined && !action.typingDelayMs) continue;\n    const plan = typingPlan(action);\n    calls += plan.length;\n    delayMs += plan.reduce((total, step) => total + step.delayMs, 0);\n  }\n  if (characters > 10_000 || calls > 1_000 || delayMs > 30_000) {\n    throw new Error(\n      \"typing budget exceeded: at most 10000 characters, 1000 paced graphemes and 30000ms delay per request; use unpaced typing for longer text\",\n    );\n  }\n}\n\nasync function typeText(action) {\n  if (action.typingStyle === undefined && !action.typingDelayMs) {\n    return call(\"type_text\", { scope: \"desktop\", text: action.text });\n  }\n  let result;\n  let graphemesTyped = 0;\n  for (const { grapheme, delayMs } of typingPlan(action)) {\n    result = await call(\"type_text\", { scope: \"desktop\", text: grapheme });\n    graphemesTyped += 1;\n    if (delayMs > 0) await sleep(delayMs);\n  }\n  return { graphemesTyped, result };\n}\n\nasync function click(action, count) {\n  if (action.button !== \"back\" && action.button !== \"forward\") {\n    return call(\"click\", pointerClickInput(action.x, action.y, action.button, count));\n  }\n\n  await call(\"move_cursor\", {\n    scope: \"desktop\",\n    x: action.x,\n    y: action.y,\n  });\n  const results = [];\n  for (let index = 0; index < count; index += 1) {\n    results.push(\n      await call(\"hotkey\", {\n        scope: \"desktop\",\n        keys: [\"alt\", action.button === \"back\" ? \"left\" : \"right\"],\n      }),\n    );\n  }\n  return results;\n}\n\nasync function readClipboard(selection) {\n  try {\n    const result = await execFileAsync(\"xclip\", [\"-selection\", xclipSelection(selection), \"-out\"], {\n      env: displayEnv(),\n      maxBuffer: maxClipboardBytes + 1,\n    });\n    if (Buffer.byteLength(result.stdout) > maxClipboardBytes) {\n      throw new Error(`clipboard content exceeds ${maxClipboardBytes} bytes`);\n    }\n    return { selection, text: result.stdout };\n  } catch (error) {\n    if (error instanceof Error && error.message.includes(\"clipboard content exceeds\")) throw error;\n    return { selection, text: \"\" };\n  }\n}\n\nasync function writeClipboard(selection, text) {\n  if (Buffer.byteLength(text) > maxClipboardBytes) {\n    throw new Error(`clipboard content exceeds ${maxClipboardBytes} bytes`);\n  }\n  await new Promise((resolve, reject) => {\n    const child = execFile(\"xclip\", [\"-selection\", xclipSelection(selection), \"-in\"], {\n      env: displayEnv(),\n      timeout: nativeCallTimeoutMs,\n      signal: requestSignal,\n    });\n    child.once(\"error\", reject);\n    child.once(\"exit\", (code) =>\n      code === 0 ? resolve() : reject(new Error(`xclip exited ${code}`)),\n    );\n    child.stdin.end(text);\n  });\n  return { bytes: Buffer.byteLength(text), selection };\n}\n\nfunction xclipSelection(selection) {\n  return selection === \"primary\" ? \"primary\" : \"clipboard\";\n}\n\nasync function arrangeWindows(action) {\n  const scene = validateWindowScene(action.windows);\n  const deadline = Date.now() + action.waitForMs;\n  let resolved = null;\n  while (resolved === null) {\n    const windows = structured(await call(\"list_windows\", { on_screen_only: true })).windows;\n    resolved = resolveSceneWindows(scene, Array.isArray(windows) ? windows : []);\n    if (resolved !== null) break;\n    if (Date.now() >= deadline) {\n      throw new Error(\n        `arrange_windows timed out waiting for: ${missingSceneSelectors(scene, Array.isArray(windows) ? windows : []).join(\", \")}`,\n      );\n    }\n    await sleep(100);\n  }\n\n  // Resolve the complete scene before mutating any window. A missing selector\n  // therefore leaves the existing desktop untouched.\n  for (const item of resolved) {\n    const windowId = String(item.window.window_id);\n    await execFileAsync(\"xdotool\", [\"windowactivate\", \"--sync\", windowId], { env: displayEnv() });\n    await execFileAsync(\n      \"xdotool\",\n      [\"windowmove\", windowId, String(item.bounds.x), String(item.bounds.y)],\n      {\n        env: displayEnv(),\n      },\n    );\n    await execFileAsync(\n      \"xdotool\",\n      [\"windowsize\", windowId, String(item.bounds.width), String(item.bounds.height)],\n      { env: displayEnv() },\n    );\n  }\n\n  let verified = null;\n  let windows = [];\n  for (let attempt = 0; attempt < 10; attempt += 1) {\n    const response = structured(await call(\"list_windows\", { on_screen_only: true })).windows;\n    windows = Array.isArray(response) ? response : [];\n    verified = resolveSceneWindows(scene, windows);\n    if (\n      verified !== null &&\n      verified.every(({ bounds, window }) => approximatelyEqual(bounds, window))\n    ) {\n      return verified.map(layoutResult);\n    }\n    await sleep(100);\n  }\n\n  const screenshotPath = `${workspaceRoot}/layout-failure-${Date.now()}.png`;\n  await call(\"get_desktop_state\", { screenshot_out_file: screenshotPath });\n  const details =\n    verified === null\n      ? { stage: \"selector_resolution\", missing: missingSceneSelectors(scene, windows) }\n      : {\n          stage: \"verification\",\n          tolerance: layoutTolerance,\n          windows: verified.map(layoutResult),\n        };\n  throw new Error(\n    `arrange_windows failed: ${JSON.stringify({ ...details, screenshot: screenshotPath })}`,\n  );\n}\n\nfunction validateWindowScene(windows) {\n  if (!Array.isArray(windows) || windows.length < 1 || windows.length > 4) {\n    throw new Error(\"arrange_windows requires 1 to 4 windows\");\n  }\n  const scene = windows.map((item) => ({ ...item, bounds: regionBounds(item.region) }));\n  const selectors = scene.map((item) => selectorLabel(item.selector));\n  if (new Set(selectors).size !== selectors.length) {\n    throw new Error(\"arrange_windows selectors must be unique\");\n  }\n  for (let left = 0; left < scene.length; left += 1) {\n    for (let right = left + 1; right < scene.length; right += 1) {\n      if (overlaps(scene[left].bounds, scene[right].bounds)) {\n        throw new Error(\"arrange_windows regions must not overlap\");\n      }\n    }\n  }\n  return scene;\n}\n\nfunction desktopBounds() {\n  const screen = x11ScreenSize();\n  return {\n    x: desktopMargin,\n    y: desktopMargin,\n    width: screen.width - desktopMargin * 2,\n    height: screen.height - desktopMargin * 2,\n  };\n}\n\nfunction x11ScreenSize() {\n  // The X server is the source of truth after set_desktop_size. Avoid CUA's\n  // session-oriented get_screen_size contract, which differs from pointer\n  // actions and does not accept scope.\n  const output = execFileSync(\"xdpyinfo\", [\"-display\", \":99\"], {\n    encoding: \"utf8\",\n    timeout: 2_000,\n  });\n  const match = /dimensions:\\s+(\\d+)x(\\d+) pixels/u.exec(output);\n  if (match === null) throw new Error(\"could not determine desktop dimensions\");\n  return { width: Number(match[1]), height: Number(match[2]) };\n}\n\nfunction regionBounds(region) {\n  if (\n    typeof region !== \"object\" ||\n    region === null ||\n    ![region.x, region.y, region.width, region.height].every(Number.isFinite) ||\n    region.x < 0 ||\n    region.y < 0 ||\n    region.width <= 0 ||\n    region.height <= 0 ||\n    region.x + region.width > 1 ||\n    region.y + region.height > 1\n  ) {\n    throw new Error(\"arrange_windows regions must fit within the desktop\");\n  }\n  const desktop = desktopBounds();\n  const left = Math.round(desktop.x + desktop.width * region.x);\n  const top = Math.round(desktop.y + desktop.height * region.y);\n  const right = Math.round(desktop.x + desktop.width * (region.x + region.width));\n  const bottom = Math.round(desktop.y + desktop.height * (region.y + region.height));\n  const bounds = { x: left, y: top, width: right - left, height: bottom - top };\n  if (bounds.width < minimumWindow.width || bounds.height < minimumWindow.height) {\n    throw new Error(\n      `arrange_windows regions must be at least ${minimumWindow.width}x${minimumWindow.height}`,\n    );\n  }\n  return bounds;\n}\n\nfunction resolveSceneWindows(scene, windows) {\n  const resolved = scene.map((item) => {\n    const window = windows.find((candidate) => matchesSelector(candidate, item.selector));\n    return window === undefined ? null : { bounds: item.bounds, window };\n  });\n  return resolved.includes(null) ? null : resolved;\n}\n\nfunction missingSceneSelectors(scene, windows) {\n  return scene\n    .filter((item) => !windows.some((window) => matchesSelector(window, item.selector)))\n    .map((item) => selectorLabel(item.selector));\n}\n\nfunction matchesSelector(window, selector) {\n  const identity = `${window.app_name ?? \"\"}\\n${window.title ?? \"\"}`.toLowerCase();\n  if (selector.app === \"firefox\") return identity.includes(\"firefox\");\n  const title = selector.instance === \"primary\" ? \"terminal 1\" : \"terminal 2\";\n  return identity.includes(title);\n}\n\nfunction selectorLabel(selector) {\n  return selector.app === \"firefox\" ? \"firefox\" : `xterm:${selector.instance}`;\n}\n\nfunction overlaps(left, right) {\n  return (\n    left.x < right.x + right.width &&\n    left.x + left.width > right.x &&\n    left.y < right.y + right.height &&\n    left.y + left.height > right.y\n  );\n}\n\nfunction layoutResult({ bounds, window }) {\n  const observed = {\n    x: Number(window.x),\n    y: Number(window.y),\n    width: Number(window.width),\n    height: Number(window.height),\n  };\n  return {\n    selector: `${window.app_name ?? \"\"}:${window.title ?? \"\"}`,\n    requested: bounds,\n    observed,\n    delta: {\n      x: observed.x - bounds.x,\n      y: observed.y - bounds.y,\n      width: observed.width - bounds.width,\n      height: observed.height - bounds.height,\n    },\n    tolerance: layoutTolerance,\n    title: window.title,\n    window_id: window.window_id,\n  };\n}\n\nfunction displayEnv() {\n  return { ...env, DISPLAY: \":99\" };\n}\n\nfunction approximatelyEqual(bounds, window) {\n  return (\n    Math.abs(bounds.x - window.x) <= layoutTolerance.x &&\n    Math.abs(bounds.y - window.y) <= layoutTolerance.y &&\n    Math.abs(bounds.width - window.width) <= layoutTolerance.width &&\n    Math.abs(bounds.height - window.height) <= layoutTolerance.height\n  );\n}\n\nfunction sleep(milliseconds) {\n  return delay(milliseconds, undefined, { signal: requestSignal });\n}\n\nfunction scrollInput(delta, negativeDirection, positiveDirection) {\n  return {\n    scope: \"desktop\",\n    direction: delta < 0 ? negativeDirection : positiveDirection,\n    amount: Math.min(50, Math.max(1, Math.round(Math.abs(delta) / 100))),\n    by: \"line\",\n  };\n}\n\nfunction normalizeKey(key) {\n  const aliases = {\n    ARROWDOWN: \"down\",\n    ARROWLEFT: \"left\",\n    ARROWRIGHT: \"right\",\n    ARROWUP: \"up\",\n    CMD: \"meta\",\n    COMMAND: \"meta\",\n    CONTROL: \"ctrl\",\n    DEL: \"delete\",\n    ENTER: \"enter\",\n    ESC: \"escape\",\n    OPTION: \"alt\",\n    RETURN: \"enter\",\n  };\n  return aliases[key] ?? key.toLowerCase();\n}\n\nfunction isContentWindow(window) {\n  const identity = `${window.app_name ?? \"\"}\\n${window.title ?? \"\"}`.toLowerCase();\n  return (\n    window.width >= 200 &&\n    window.height >= 100 &&\n    !identity.includes(\"xfce4-panel\") &&\n    !identity.includes(\"xfdesktop\") &&\n    window.title !== \"Desktop\"\n  );\n}\n\nfunction formatState(groundingWindow, pointer, elements) {\n  const lines = [\n    groundingWindow === undefined\n      ? \"grounding_window\\t(none)\"\n      : `grounding_window\\t${groundingWindow.title} pid=${groundingWindow.pid} window_id=${groundingWindow.window_id}`,\n    `pointer\\tx=${pointer.x ?? \"unknown\"} y=${pointer.y ?? \"unknown\"}`,\n    \"accessibility_elements\",\n  ];\n\n  for (const element of elements) {\n    const frame =\n      element.frame === undefined\n        ? \"\"\n        : ` x=${element.frame.x} y=${element.frame.y} width=${element.frame.w} height=${element.frame.h}`;\n    lines.push(\n      `[${element.element_index}] role=${element.role}${frame} label=${JSON.stringify(element.label ?? \"\")}`,\n    );\n  }\n  return lines.join(\"\\n\");\n}\n\nfunction validateActionCoordinates(action) {\n  const screen = x11ScreenSize();\n  const points = [];\n  if ([\"click\", \"double_click\", \"triple_click\", \"move\", \"scroll\"].includes(action.action)) {\n    points.push({ x: action.x, y: action.y });\n  }\n  if (action.action === \"drag\") points.push(...action.path);\n  for (const point of points) {\n    if (\n      !Number.isFinite(point.x) ||\n      !Number.isFinite(point.y) ||\n      point.x < 0 ||\n      point.y < 0 ||\n      point.x >= screen.width ||\n      point.y >= screen.height\n    ) {\n      throw new Error(\n        `computer-use coordinate is outside the ${screen.width}x${screen.height} desktop`,\n      );\n    }\n  }\n}\n\nfunction validateRequest(input) {\n  if (typeof input !== \"object\" || input === null) {\n    throw new Error(\"request must be an object\");\n  }\n  if (!isArtifactPath(input.screenshotPath, \".png\")) {\n    throw new Error(\"screenshot path must be a PNG under the computer-use workspace\");\n  }\n  if (typeof input.action !== \"object\" || input.action === null) {\n    throw new Error(\"request action must be an object\");\n  }\n  const actions = new Set([\n    \"arrange_windows\",\n    \"click\",\n    \"clipboard_read\",\n    \"clipboard_write\",\n    \"double_click\",\n    \"drag\",\n    \"keypress\",\n    \"move\",\n    \"record_start\",\n    \"record_stop\",\n    \"screenshot\",\n    \"scroll\",\n    \"sequence\",\n    \"triple_click\",\n    \"type\",\n    \"wait\",\n  ]);\n  if (!actions.has(input.action.action)) {\n    throw new Error(`unsupported computer-use action: ${String(input.action.action)}`);\n  }\n  if (input.action.action === \"sequence\") {\n    if (\n      !Array.isArray(input.action.actions) ||\n      input.action.actions.length < 1 ||\n      input.action.actions.length > 30\n    ) {\n      throw new Error(\"sequence must contain 1 to 30 actions\");\n    }\n    for (const action of input.action.actions) validateNestedSequenceAction(action);\n  }\n  if (input.action.action === \"arrange_windows\") validateArrangeWindowsAction(input.action);\n  if (input.action.action === \"record_start\" && !isArtifactPath(input.action.path, \".mp4\")) {\n    throw new Error(\"recording path must be an MP4 under the computer-use workspace\");\n  }\n  if (\n    [\"click\", \"double_click\", \"triple_click\", \"drag\", \"move\"].includes(input.action.action) &&\n    \"durationMs\" in input.action &&\n    (!Number.isInteger(input.action.durationMs) ||\n      input.action.durationMs < 0 ||\n      input.action.durationMs > 2000)\n  ) {\n    throw new Error(\"pointer durationMs must be an integer from 0 to 2000\");\n  }\n  validateMovementStyle(input.action, \"\");\n  if (\n    input.action.action === \"wait\" &&\n    (!Number.isInteger(input.action.durationMs) ||\n      input.action.durationMs < 100 ||\n      input.action.durationMs > 10_000)\n  ) {\n    throw new Error(\"wait durationMs must be an integer from 100 to 10000\");\n  }\n  validateClipboardOptions(input.action);\n  validateTypingOptions(input.action, \"\");\n  validateTypingBudget(input.action.action === \"sequence\" ? input.action.actions : [input.action]);\n}\n\nfunction validateNestedSequenceAction(action) {\n  if (typeof action !== \"object\" || action === null)\n    throw new Error(\"sequence action must be an object\");\n  const allowed = new Set([\n    \"arrange_windows\",\n    \"click\",\n    \"double_click\",\n    \"drag\",\n    \"keypress\",\n    \"move\",\n    \"record_start\",\n    \"record_stop\",\n    \"triple_click\",\n    \"type\",\n    \"wait\",\n  ]);\n  if (!allowed.has(action.action))\n    throw new Error(`unsupported sequence action: ${String(action.action)}`);\n  if (action.action === \"arrange_windows\") validateArrangeWindowsAction(action);\n  if (action.action === \"record_start\" && !isArtifactPath(action.path, \".mp4\")) {\n    throw new Error(\"sequence recording path must be an MP4 under the computer-use workspace\");\n  }\n  if (\n    [\"click\", \"double_click\", \"triple_click\", \"drag\", \"move\"].includes(action.action) &&\n    \"durationMs\" in action &&\n    (!Number.isInteger(action.durationMs) || action.durationMs < 0 || action.durationMs > 2000)\n  ) {\n    throw new Error(\"sequence pointer durationMs must be an integer from 0 to 2000\");\n  }\n  validateMovementStyle(action, \"sequence \");\n  if (\n    action.action === \"wait\" &&\n    (!Number.isInteger(action.durationMs) || action.durationMs < 100 || action.durationMs > 10_000)\n  ) {\n    throw new Error(\"sequence wait durationMs must be an integer from 100 to 10000\");\n  }\n  if (action.action === \"type\") validateTypingOptions(action, \"sequence \");\n}\n\nfunction validateMovementStyle(action, prefix) {\n  if (\n    \"movementStyle\" in action &&\n    action.movementStyle !== \"natural\" &&\n    action.movementStyle !== \"precision\"\n  ) {\n    throw new Error(`${prefix}movementStyle must be natural or precision`);\n  }\n  if (\n    action.action === \"drag\" &&\n    \"movementStyle\" in action &&\n    action.movementStyle !== \"precision\"\n  ) {\n    throw new Error(`${prefix}drag movementStyle must be precision`);\n  }\n}\n\nfunction validateClipboardOptions(action) {\n  if (\n    [\"clipboard_read\", \"clipboard_write\"].includes(action.action) &&\n    action.selection !== \"clipboard\" &&\n    action.selection !== \"primary\"\n  ) {\n    throw new Error(\"clipboard selection must be clipboard or primary\");\n  }\n  if (\n    action.action === \"clipboard_write\" &&\n    (typeof action.text !== \"string\" || Buffer.byteLength(action.text) > maxClipboardBytes)\n  ) {\n    throw new Error(`clipboard text must be at most ${maxClipboardBytes} bytes`);\n  }\n}\n\nfunction validateTypingOptions(action, prefix) {\n  if (\n    \"typingDelayMs\" in action &&\n    (!Number.isInteger(action.typingDelayMs) ||\n      action.typingDelayMs < 0 ||\n      action.typingDelayMs > 200)\n  ) {\n    throw new Error(`${prefix}typingDelayMs must be an integer from 0 to 200`);\n  }\n  if (\"typingStyle\" in action && action.typingStyle !== \"natural\") {\n    throw new Error(`${prefix}typingStyle must be natural`);\n  }\n  if (\"typingStyle\" in action && \"typingDelayMs\" in action) {\n    throw new Error(`${prefix}type action cannot set both typingStyle and typingDelayMs`);\n  }\n}\n\nfunction validateArrangeWindowsAction(action) {\n  if (!Number.isInteger(action.waitForMs) || action.waitForMs < 0 || action.waitForMs > 10_000) {\n    throw new Error(\"arrange_windows waitForMs must be an integer from 0 to 10000\");\n  }\n  validateWindowScene(action.windows);\n}\n\nfunction isArtifactPath(value, extension) {\n  if (typeof value !== \"string\" || !value.endsWith(extension)) return false;\n  const path = resolve(value);\n  return dirname(path) === workspaceRoot;\n}\n\nasync function readBody(request) {\n  let body = \"\";\n  for await (const chunk of request) {\n    body += chunk;\n    if (Buffer.byteLength(body) > maxRequestBytes) {\n      throw new Error(\"request exceeds 32 KiB\");\n    }\n  }\n  return body;\n}\n\nfunction sendJson(response, statusCode, body) {\n  if (response.destroyed) return;\n  const json = JSON.stringify(body);\n  response.writeHead(statusCode, {\n    \"content-length\": Buffer.byteLength(json),\n    \"content-type\": \"application/json\",\n  });\n  response.end(json);\n}\n";
+export declare const CLIENT_MJS_SOURCE = "import http from \"node:http\";\nimport { env } from \"node:process\";\n\nconst socketPath = env.COMPUTER_USE_SOCKET_PATH;\nconst body = env.COMPUTER_USE_REQUEST;\n\nif (body === undefined) {\n  throw new Error(\"COMPUTER_USE_REQUEST is required\");\n}\n\nconst deadline = Date.now() + 120_000;\nconst response = await new Promise((resolve, reject) => {\n  const request = http.request(\n    {\n      socketPath,\n      method: \"POST\",\n      path: \"/call\",\n      headers: {\n        \"content-length\": Buffer.byteLength(body),\n        \"content-type\": \"application/json\",\n        \"x-computer-use-deadline\": String(deadline),\n      },\n    },\n    (incoming) => {\n      incoming.on(\"error\", reject);\n      incoming.setEncoding(\"utf8\");\n      let responseBody = \"\";\n      incoming.on(\"data\", (chunk) => {\n        responseBody += chunk;\n      });\n      incoming.on(\"end\", () => {\n        resolve({ body: responseBody, statusCode: incoming.statusCode ?? 500 });\n      });\n    },\n  );\n  const timer = setTimeout(\n    () => request.destroy(new Error(\"computer-use request deadline exceeded\")),\n    Math.max(0, deadline - Date.now()),\n  );\n  request.once(\"close\", () => clearTimeout(timer));\n  request.on(\"error\", reject);\n  request.end(body);\n});\n\nif (response.statusCode !== 200) {\n  process.stderr.write(response.body);\n  process.exitCode = 1;\n} else {\n  process.stdout.write(response.body);\n}\n";
+export declare const STATE_MJS_SOURCE = "const MAX_GROUNDING_ELEMENTS = 80;\n\n/**\n * @typedef {{\n *   element_index?: number;\n *   frame?: { h?: number; w?: number; x?: number; y?: number };\n *   label?: unknown;\n *   [key: string]: unknown;\n * }} GroundingElement\n */\n\n/**\n * Keep the model-facing state aligned with the visible desktop screenshot.\n *\n * @param {unknown[]} elements\n * @returns {GroundingElement[]}\n */\nexport function selectGroundingElements(elements, screen = { width: 1920, height: 1080 }) {\n  return elements\n    .filter((element) => isVisibleLabeledElement(element, screen))\n    .slice(0, MAX_GROUNDING_ELEMENTS);\n}\n\n/**\n * @param {unknown} value\n * @returns {value is GroundingElement}\n */\nfunction isVisibleLabeledElement(value, screen) {\n  if (typeof value !== \"object\" || value === null) return false;\n  const element = /** @type {GroundingElement} */ (value);\n  const frame = element.frame;\n  if (typeof element.label !== \"string\" || element.label.trim().length === 0) return false;\n  if (typeof frame !== \"object\" || frame === null) return false;\n  if (\n    !Number.isFinite(frame.x) ||\n    !Number.isFinite(frame.y) ||\n    !Number.isFinite(frame.w) ||\n    !Number.isFinite(frame.h)\n  ) {\n    return false;\n  }\n\n  const x = /** @type {number} */ (frame.x);\n  const y = /** @type {number} */ (frame.y);\n  const width = /** @type {number} */ (frame.w);\n  const height = /** @type {number} */ (frame.h);\n  return (\n    width > 0 &&\n    height > 0 &&\n    x < screen.width &&\n    y < screen.height &&\n    x + width > 0 &&\n    y + height > 0\n  );\n}\n";
+export declare const TYPING_MJS_SOURCE = "const DEMO_TYPING_SPEED = 1.2;\nconst MIN_DELAY_MS = 35;\nconst MAX_DELAY_MS = 220;\n\nexport function naturalTypingPlan(text) {\n  const graphemes = [...new Intl.Segmenter().segment(text)].map(({ segment }) => segment);\n  let state = hashText(text) || 1;\n  return graphemes.map((grapheme, index) => {\n    state = nextState(state);\n    return {\n      grapheme,\n      delayMs: index === graphemes.length - 1 ? 0 : naturalDelay(grapheme, state),\n    };\n  });\n}\n\nfunction naturalDelay(grapheme, state) {\n  const jitter = state / 0x1_0000_0000;\n  const range = delayRange(grapheme);\n  return Math.round((range.min + (range.max - range.min) * jitter) / DEMO_TYPING_SPEED);\n}\n\nfunction delayRange(grapheme) {\n  if (/\\s/u.test(grapheme)) return { min: 75, max: 125 };\n  if (/[.!?]/u.test(grapheme)) return { min: 145, max: MAX_DELAY_MS };\n  if (/[/|;&]/u.test(grapheme)) return { min: 105, max: 175 };\n  if (/[-_=,:]/u.test(grapheme)) return { min: 70, max: 125 };\n  return { min: MIN_DELAY_MS, max: 85 };\n}\n\nfunction hashText(text) {\n  let hash = 2166136261;\n  for (const character of text) {\n    hash ^= character.codePointAt(0);\n    hash = Math.imul(hash, 16777619);\n  }\n  return hash >>> 0;\n}\n\nfunction nextState(state) {\n  let value = state;\n  value ^= value << 13;\n  value ^= value >>> 17;\n  value ^= value << 5;\n  return value >>> 0;\n}\n";
+export declare const POINTER_MOTION_MJS_SOURCE = "const FRAME_MS = 16;\n\nexport function pointerMotionPlan(start, end, options = {}) {\n  const style = options.style ?? \"natural\";\n  const distance = Math.hypot(end.x - start.x, end.y - start.y);\n  const durationMs = options.durationMs ?? naturalDuration(distance, style);\n  if (distance < 2 || durationMs === 0) {\n    return { phases: [[{ ...end, delayMs: 0 }]], settleMs: 0 };\n  }\n\n  if (style === \"precision\" && distance >= 120) {\n    const ratio = 0.92 + deterministicUnit(start, end, 7) * 0.04;\n    const undershoot = {\n      x: Math.round(start.x + (end.x - start.x) * ratio),\n      y: Math.round(start.y + (end.y - start.y) * ratio),\n    };\n    // Cover most of the distance quickly, then spend a disproportionate\n    // share of the duration continuously correcting the final 4\u20138%.\n    const firstDuration = Math.round(durationMs * 0.62);\n    return {\n      phases: [\n        bezierPhase(start, undershoot, firstDuration, \"natural\", 11),\n        bezierPhase(undershoot, end, durationMs - firstDuration, \"precision\", 23),\n      ],\n      settleMs: 0,\n    };\n  }\n\n  return {\n    phases: [bezierPhase(start, end, durationMs, style, 17)],\n    settleMs: style === \"natural\" ? 15 + Math.round(deterministicUnit(start, end, 31) * 15) : 0,\n  };\n}\n\nfunction bezierPhase(start, end, durationMs, style, salt) {\n  const distance = Math.hypot(end.x - start.x, end.y - start.y);\n  const steps = Math.max(2, Math.min(60, Math.round(durationMs / FRAME_MS)));\n  const perpendicular = unitPerpendicular(start, end);\n  const bendLimit = style === \"precision\" ? 8 : Math.min(70, distance * 0.09);\n  const bend = (deterministicUnit(start, end, salt) * 2 - 1) * bendLimit;\n  const c1 = {\n    x: start.x + (end.x - start.x) * 0.28 + perpendicular.x * bend,\n    y: start.y + (end.y - start.y) * 0.28 + perpendicular.y * bend,\n  };\n  const c2 = {\n    x: start.x + (end.x - start.x) * 0.76 + perpendicular.x * bend * 0.65,\n    y: start.y + (end.y - start.y) * 0.76 + perpendicular.y * bend * 0.65,\n  };\n  const delayMs = durationMs / steps;\n  return Array.from({ length: steps }, (_, index) => {\n    // This asymmetric timing curve accelerates quickly and eases into the\n    // target without the perfectly mirrored feel of smoothstep.\n    const progress = (index + 1) / steps;\n    const t = 1 - (1 - progress) ** 2.35;\n    const point = cubicBezier(start, c1, c2, end, t);\n    return {\n      x: index === steps - 1 ? end.x : Math.round(point.x),\n      y: index === steps - 1 ? end.y : Math.round(point.y),\n      delayMs: index === steps - 1 ? 0 : delayMs,\n    };\n  });\n}\n\nfunction cubicBezier(start, c1, c2, end, t) {\n  const inverse = 1 - t;\n  return {\n    x:\n      inverse ** 3 * start.x +\n      3 * inverse ** 2 * t * c1.x +\n      3 * inverse * t ** 2 * c2.x +\n      t ** 3 * end.x,\n    y:\n      inverse ** 3 * start.y +\n      3 * inverse ** 2 * t * c1.y +\n      3 * inverse * t ** 2 * c2.y +\n      t ** 3 * end.y,\n  };\n}\n\nfunction naturalDuration(distance, style) {\n  const base = style === \"precision\" ? 150 : 110;\n  const scale = style === \"precision\" ? 0.35 : 0.28;\n  const minimum = style === \"precision\" ? 180 : 140;\n  const maximum = style === \"precision\" ? 650 : 550;\n  return Math.round(Math.min(maximum, Math.max(minimum, base + distance * scale)));\n}\n\nfunction unitPerpendicular(start, end) {\n  const dx = end.x - start.x;\n  const dy = end.y - start.y;\n  const length = Math.hypot(dx, dy) || 1;\n  return { x: -dy / length, y: dx / length };\n}\n\nfunction deterministicUnit(start, end, salt) {\n  let value =\n    (Math.imul(Math.round(start.x), 73_856_093) ^\n      Math.imul(Math.round(start.y), 19_349_663) ^\n      Math.imul(Math.round(end.x), 83_492_791) ^\n      Math.imul(Math.round(end.y), 2_654_435_761) ^\n      salt) >>>\n    0;\n  value ^= value << 13;\n  value ^= value >>> 17;\n  value ^= value << 5;\n  return (value >>> 0) / 0x1_0000_0000;\n}\n";
+export declare const PACKAGE_JSON_SOURCE = "{\n  \"name\": \"computer-use-driver\",\n  \"private\": true,\n  \"type\": \"module\",\n  \"dependencies\": {\n    \"@trycua/cua-driver\": \"0.12.5\"\n  }\n}\n";
+export declare const COMPUTER_USE_DRIVER_SOURCES: {
+    readonly "server.mjs": "import { setTimeout as delay } from \"node:timers/promises\";\nimport { execFile, execFileSync } from \"node:child_process\";\nimport { randomUUID } from \"node:crypto\";\nimport { chmod, rename, rm } from \"node:fs/promises\";\nimport { promisify } from \"node:util\";\nimport http from \"node:http\";\nimport { env } from \"node:process\";\nimport { dirname, resolve } from \"node:path\";\n\nimport { CuaDriver } from \"@trycua/cua-driver\";\n\nimport { pointerMotionPlan } from \"./pointer-motion.mjs\";\nimport { selectGroundingElements } from \"./state.mjs\";\nimport { naturalTypingPlan } from \"./typing.mjs\";\n\nconst socketPath = env.COMPUTER_USE_SOCKET_PATH;\nconst workspaceRoot = env.COMPUTER_USE_ROOT;\nconst maxRequestBytes = 32 * 1024;\nconst maxRequestMs = 120_000;\nconst maxRecordingMs = 300_000;\nconst nativeCallTimeoutMs = 10_000;\n// Only the queue owner sets this signal; recording finalization deliberately bypasses it.\nlet requestSignal;\nconst desktopMargin = 24;\nconst minimumWindow = { height: 240, width: 480 };\nconst layoutTolerance = { height: 8, width: 8, x: 4, y: 4 };\nconst maxClipboardBytes = 10_000;\nconst driver = await CuaDriver.create({ claudeCodeCompatibility: false });\nconst execFilePromise = promisify(execFile);\nfunction execFileAsync(file, args, options) {\n  return execFilePromise(file, args, {\n    timeout: nativeCallTimeoutMs,\n    ...options,\n    signal: requestSignal,\n  });\n}\n\nlet activeRecording = null;\nlet queue = Promise.resolve();\nlet shuttingDown = false;\n\nawait rm(socketPath, { force: true });\n\nconst server = http.createServer((request, response) => {\n  if (shuttingDown) return response.destroy();\n  if (request.method !== \"POST\" || request.url !== \"/call\") {\n    sendJson(response, 404, { error: \"not found\" });\n    return;\n  }\n  const controller = new AbortController();\n  const deadline = Number(request.headers[\"x-computer-use-deadline\"]);\n  const timeoutMs = Number.isFinite(deadline)\n    ? Math.max(0, Math.min(maxRequestMs, deadline - Date.now()))\n    : maxRequestMs;\n  const cancel = () =>\n    controller.abort(new Error(\"computer-use request cancelled or deadline exceeded\"));\n  const timer = setTimeout(cancel, timeoutMs);\n  response.once(\"close\", cancel);\n  controller.signal.addEventListener(\"abort\", () => response.destroy(), { once: true });\n  if (timeoutMs === 0) cancel();\n  // Read immediately: a body may finish or disconnect while another call owns the queue.\n  const body = readBody(request);\n  void body.catch(() => undefined);\n  const task = queue.then(async () => {\n    controller.signal.throwIfAborted();\n    if (shuttingDown) return;\n    requestSignal = controller.signal;\n    try {\n      await handleRequest(await body, response);\n    } finally {\n      requestSignal = undefined;\n    }\n  });\n  queue = task\n    .catch(() => response.destroy())\n    .finally(() => {\n      clearTimeout(timer);\n      response.removeListener(\"close\", cancel);\n    });\n});\n\nserver.listen(socketPath, () => {\n  void chmod(socketPath, 0o600);\n});\n\nfor (const signal of [\"SIGINT\", \"SIGTERM\"]) {\n  process.once(signal, () => {\n    void shutdown();\n  });\n}\n\nasync function shutdown() {\n  if (shuttingDown) return;\n  shuttingDown = true;\n  server.close();\n  server.closeAllConnections();\n  const timer = setTimeout(() => process.exit(1), nativeCallTimeoutMs * 2);\n  let exitCode = 0;\n  try {\n    await queue;\n    if (activeRecording !== null) await stopRecording();\n  } catch {\n    exitCode = 1;\n  } finally {\n    try {\n      await driver.shutdown();\n    } catch {\n      exitCode = 1;\n    } finally {\n      clearTimeout(timer);\n      process.exit(exitCode);\n    }\n  }\n}\n\nasync function handleRequest(body, response) {\n  try {\n    const input = JSON.parse(body);\n    if (input.action === \"health\") {\n      sendJson(response, 200, { backend: \"cua-driver\", version: \"0.12.5\" });\n      return;\n    }\n\n    validateRequest(input);\n    const actionResult = await performAction(input.action);\n    const state = await captureState(input.screenshotPath);\n    sendJson(response, 200, {\n      action: actionResult,\n      backend: \"cua-driver\",\n      screenshot: input.screenshotPath,\n      state,\n      version: \"0.12.5\",\n    });\n  } catch (error) {\n    if (requestSignal?.aborted && activeRecording !== null) {\n      await stopRecording().catch(() => undefined);\n    }\n    sendJson(response, 500, {\n      error: error instanceof Error ? error.message : String(error),\n    });\n  }\n}\n\nasync function performAction(action) {\n  requestSignal?.throwIfAborted();\n  validateActionCoordinates(action);\n  if (action.action === \"sequence\") {\n    const results = [];\n    try {\n      for (const nested of action.actions) {\n        const startedAt = performance.now();\n        const result = await performAction(nested);\n        results.push({\n          action: nested.action,\n          durationMs: Math.round(performance.now() - startedAt),\n          result,\n        });\n      }\n      return results;\n    } catch (error) {\n      // Never strand the durable driver in an active-recording state when a\n      // rehearsed step fails. Preserve the original action error.\n      if (activeRecording !== null) await stopRecording().catch(() => undefined);\n      throw error;\n    }\n  }\n  switch (action.action) {\n    case \"screenshot\":\n      return null;\n    case \"click\":\n      await moveCursor(action);\n      return click(action, 1);\n    case \"double_click\":\n      await moveCursor(action);\n      return click(action, 2);\n    case \"triple_click\":\n      await moveCursor(action);\n      return click(action, 3);\n    case \"drag\":\n      return dragCursor(action);\n    case \"keypress\":\n      return action.keys.length === 1\n        ? call(\"press_key\", {\n            scope: \"desktop\",\n            key: normalizeKey(action.keys[0]),\n          })\n        : call(\"hotkey\", {\n            scope: \"desktop\",\n            keys: action.keys.map(normalizeKey),\n          });\n    case \"move\":\n      return moveCursor(action);\n    case \"scroll\": {\n      await call(\"move_cursor\", {\n        scope: \"desktop\",\n        x: action.x,\n        y: action.y,\n      });\n      const results = [];\n      if (action.scrollY !== 0) {\n        results.push(await call(\"scroll\", scrollInput(action.scrollY, \"up\", \"down\")));\n      }\n      if (action.scrollX !== 0) {\n        results.push(await call(\"scroll\", scrollInput(action.scrollX, \"left\", \"right\")));\n      }\n      return results;\n    }\n    case \"type\":\n      return typeText(action);\n    case \"clipboard_read\":\n      return readClipboard(action.selection);\n    case \"clipboard_write\":\n      return writeClipboard(action.selection, action.text);\n    case \"wait\":\n      await sleep(action.durationMs ?? 2_000);\n      return null;\n    case \"arrange_windows\":\n      return arrangeWindows(action);\n    case \"record_start\":\n      return startRecording(action.path);\n    case \"record_stop\":\n      return stopRecording();\n    default:\n      throw new Error(`unsupported computer-use action: ${String(action.action)}`);\n  }\n}\n\nasync function captureState(screenshotPath) {\n  await call(\"get_desktop_state\", { screenshot_out_file: screenshotPath });\n  const pointer = structured(await call(\"get_cursor_position\", {}));\n  const screen = x11ScreenSize();\n  const windows = structured(await call(\"list_windows\", { on_screen_only: true })).windows;\n  const contentWindows = Array.isArray(windows) ? windows.filter(isContentWindow) : [];\n  const groundingWindow = contentWindows.toSorted((left, right) => right.z_index - left.z_index)[0];\n\n  let elements = [];\n  if (groundingWindow !== undefined) {\n    const windowState = structured(\n      await call(\"get_window_state\", {\n        pid: groundingWindow.pid,\n        window_id: groundingWindow.window_id,\n        include_screenshot: false,\n        max_depth: 20,\n        max_elements: 250,\n      }),\n    );\n    if (Array.isArray(windowState.elements)) {\n      elements = selectGroundingElements(windowState.elements, {\n        width: Number(screen.width) || 1920,\n        height: Number(screen.height) || 1080,\n      });\n    }\n  }\n\n  return formatState(groundingWindow, pointer, elements);\n}\n\nasync function startRecording(path) {\n  if (activeRecording !== null) {\n    throw new Error(\"a recording is already active\");\n  }\n\n  const outputDirectory = `${workspaceRoot}/recording-${randomUUID()}`;\n  const result = await call(\n    \"start_recording\",\n    {\n      output_dir: outputDirectory,\n      record_video: true,\n    },\n    false,\n  );\n  activeRecording = {\n    outputDirectory,\n    outputPath: path,\n    stopping: null,\n    timer: setTimeout(() => {\n      void stopRecording().catch(() => {\n        void shutdown();\n      });\n    }, maxRecordingMs),\n  };\n  return result;\n}\n\nasync function stopRecording() {\n  if (activeRecording === null) {\n    throw new Error(\"no recording is active\");\n  }\n\n  const recording = activeRecording;\n  if (recording.stopping !== null) return recording.stopping;\n  clearTimeout(recording.timer);\n  recording.stopping = finalizeRecording(recording);\n  return recording.stopping;\n}\n\nasync function finalizeRecording(recording) {\n  try {\n    const result = await call(\"stop_recording\", {}, false);\n    const output = structured(result);\n    const videoPath =\n      typeof output.last_video_path === \"string\"\n        ? output.last_video_path\n        : recording.outputDirectory + \"/recording.mp4\";\n    await rename(videoPath, recording.outputPath);\n    await rm(recording.outputDirectory, { force: true, recursive: true });\n    return result;\n  } catch (error) {\n    void shutdown();\n    throw error;\n  } finally {\n    activeRecording = null;\n  }\n}\n\nasync function call(name, input, cancellable = true) {\n  const signal = cancellable ? requestSignal : undefined;\n  signal?.throwIfAborted();\n  let timer;\n  const result = await Promise.race([\n    driver.callTool(name, JSON.stringify(input)),\n    new Promise((_, reject) => {\n      timer = setTimeout(() => {\n        // Native calls have no abort API. Do not reuse a driver with uncertain in-flight work.\n        void shutdown();\n        reject(new Error(\"computer-use native call timed out; restart with startComputerUse\"));\n      }, nativeCallTimeoutMs);\n    }),\n  ]).finally(() => clearTimeout(timer));\n  signal?.throwIfAborted();\n  if (result.isError) {\n    throw new Error(\n      `Cua ${name} failed${result.errorCode ? ` (${result.errorCode})` : \"\"}: ${result.text}`,\n    );\n  }\n  return {\n    degraded: result.degraded,\n    structuredJson: result.structuredJson,\n    text: result.text,\n    verified: result.verified,\n  };\n}\n\nfunction structured(result) {\n  if (result.structuredJson === undefined) return {};\n  return JSON.parse(result.structuredJson);\n}\n\nfunction pointerClickInput(x, y, button, count) {\n  return {\n    scope: \"desktop\",\n    x,\n    y,\n    button: button === \"wheel\" ? \"middle\" : button,\n    count,\n  };\n}\n\nasync function moveCursor(action) {\n  if (action.movementStyle === undefined) {\n    return call(\"move_cursor\", { scope: \"desktop\", x: action.x, y: action.y });\n  }\n  const current = structured(await call(\"get_cursor_position\", {}));\n  const start = { x: Number(current.x), y: Number(current.y) };\n  if (!Number.isFinite(start.x) || !Number.isFinite(start.y)) {\n    return call(\"move_cursor\", { scope: \"desktop\", x: action.x, y: action.y });\n  }\n  const options = { style: action.movementStyle };\n  if (action.durationMs !== undefined) options.durationMs = action.durationMs;\n  return followPointerPlan(pointerMotionPlan(start, { x: action.x, y: action.y }, options), \"x11\");\n}\n\nasync function dragCursor(action) {\n  const start = action.path[0];\n  const end = action.path.at(-1);\n  if (action.movementStyle === undefined) {\n    return call(\"drag\", {\n      scope: \"desktop\",\n      from_x: start.x,\n      from_y: start.y,\n      to_x: end.x,\n      to_y: end.y,\n      steps: Math.min(200, Math.max(20, action.path.length * 10)),\n    });\n  }\n  const style = action.movementStyle;\n  const current = structured(await call(\"get_cursor_position\", {}));\n  const currentPoint = { x: Number(current.x), y: Number(current.y) };\n  const approach =\n    Number.isFinite(currentPoint.x) && Number.isFinite(currentPoint.y)\n      ? pointerPlanArgs(pointerMotionPlan(currentPoint, start, { style: \"natural\" }))\n      : [\"mousemove\", String(start.x), String(start.y)];\n  const options = { style };\n  if (action.durationMs !== undefined) options.durationMs = action.durationMs;\n  const drag = pointerPlanArgs(pointerMotionPlan(start, end, options));\n  return xdotool([\n    ...approach,\n    \"sleep\",\n    \"0.025\",\n    \"mousedown\",\n    \"1\",\n    \"sleep\",\n    \"0.030\",\n    ...drag,\n    \"sleep\",\n    \"0.030\",\n    \"mouseup\",\n    \"1\",\n  ]);\n}\n\nasync function followPointerPlan(plan, backend = \"cua\") {\n  if (backend === \"x11\") return xdotool(pointerPlanArgs(plan));\n  const results = [];\n  for (let phaseIndex = 0; phaseIndex < plan.phases.length; phaseIndex += 1) {\n    for (const point of plan.phases[phaseIndex]) {\n      results.push(await call(\"move_cursor\", { scope: \"desktop\", x: point.x, y: point.y }));\n      if (point.delayMs > 0) await sleep(point.delayMs);\n    }\n    if (phaseIndex < plan.phases.length - 1 && plan.settleMs > 0) await sleep(plan.settleMs);\n  }\n  if (plan.phases.length === 1 && plan.settleMs > 0) await sleep(plan.settleMs);\n  return results;\n}\n\nfunction pointerPlanArgs(plan) {\n  const args = [];\n  for (let phaseIndex = 0; phaseIndex < plan.phases.length; phaseIndex += 1) {\n    for (const point of plan.phases[phaseIndex]) {\n      args.push(\"mousemove\", String(point.x), String(point.y));\n      if (point.delayMs > 0) args.push(\"sleep\", (point.delayMs / 1000).toFixed(3));\n    }\n    if (phaseIndex < plan.phases.length - 1 && plan.settleMs > 0) {\n      args.push(\"sleep\", (plan.settleMs / 1000).toFixed(3));\n    }\n  }\n  if (plan.phases.length === 1 && plan.settleMs > 0) {\n    args.push(\"sleep\", (plan.settleMs / 1000).toFixed(3));\n  }\n  return args;\n}\n\nasync function xdotool(args) {\n  const plannedSleepMs = args.reduce(\n    (total, value, index) => (args[index - 1] === \"sleep\" ? total + Number(value) * 1000 : total),\n    0,\n  );\n  const timeout = Math.max(2_000, Math.ceil(plannedSleepMs + 1_500));\n  const result = await execFileAsync(\"xdotool\", args, { env: displayEnv(), timeout });\n  return {\n    plannedSleepMs: Math.round(plannedSleepMs),\n    stderr: result.stderr,\n    stdout: result.stdout,\n  };\n}\n\nfunction typingPlan(action) {\n  if (action.typingStyle === \"natural\") return naturalTypingPlan(action.text);\n  return [...new Intl.Segmenter().segment(action.text)].map(({ segment }) => ({\n    grapheme: segment,\n    delayMs: action.typingDelayMs ?? 0,\n  }));\n}\n\nfunction validateTypingBudget(actions) {\n  let characters = 0;\n  let calls = 0;\n  let delayMs = 0;\n  for (const action of actions) {\n    if (action.action !== \"type\") continue;\n    if (typeof action.text !== \"string\" || action.text.length > 10_000) {\n      throw new Error(\"typing text must be at most 10000 characters\");\n    }\n    characters += action.text.length;\n    if (action.typingStyle === undefined && !action.typingDelayMs) continue;\n    const plan = typingPlan(action);\n    calls += plan.length;\n    delayMs += plan.reduce((total, step) => total + step.delayMs, 0);\n  }\n  if (characters > 10_000 || calls > 1_000 || delayMs > 30_000) {\n    throw new Error(\n      \"typing budget exceeded: at most 10000 characters, 1000 paced graphemes and 30000ms delay per request; use unpaced typing for longer text\",\n    );\n  }\n}\n\nasync function typeText(action) {\n  if (action.typingStyle === undefined && !action.typingDelayMs) {\n    return call(\"type_text\", { scope: \"desktop\", text: action.text });\n  }\n  let result;\n  let graphemesTyped = 0;\n  for (const { grapheme, delayMs } of typingPlan(action)) {\n    result = await call(\"type_text\", { scope: \"desktop\", text: grapheme });\n    graphemesTyped += 1;\n    if (delayMs > 0) await sleep(delayMs);\n  }\n  return { graphemesTyped, result };\n}\n\nasync function click(action, count) {\n  if (action.button !== \"back\" && action.button !== \"forward\") {\n    return call(\"click\", pointerClickInput(action.x, action.y, action.button, count));\n  }\n\n  await call(\"move_cursor\", {\n    scope: \"desktop\",\n    x: action.x,\n    y: action.y,\n  });\n  const results = [];\n  for (let index = 0; index < count; index += 1) {\n    results.push(\n      await call(\"hotkey\", {\n        scope: \"desktop\",\n        keys: [\"alt\", action.button === \"back\" ? \"left\" : \"right\"],\n      }),\n    );\n  }\n  return results;\n}\n\nasync function readClipboard(selection) {\n  try {\n    const result = await execFileAsync(\"xclip\", [\"-selection\", xclipSelection(selection), \"-out\"], {\n      env: displayEnv(),\n      maxBuffer: maxClipboardBytes + 1,\n    });\n    if (Buffer.byteLength(result.stdout) > maxClipboardBytes) {\n      throw new Error(`clipboard content exceeds ${maxClipboardBytes} bytes`);\n    }\n    return { selection, text: result.stdout };\n  } catch (error) {\n    if (error instanceof Error && error.message.includes(\"clipboard content exceeds\")) throw error;\n    return { selection, text: \"\" };\n  }\n}\n\nasync function writeClipboard(selection, text) {\n  if (Buffer.byteLength(text) > maxClipboardBytes) {\n    throw new Error(`clipboard content exceeds ${maxClipboardBytes} bytes`);\n  }\n  await new Promise((resolve, reject) => {\n    const child = execFile(\"xclip\", [\"-selection\", xclipSelection(selection), \"-in\"], {\n      env: displayEnv(),\n      timeout: nativeCallTimeoutMs,\n      signal: requestSignal,\n    });\n    child.once(\"error\", reject);\n    child.once(\"exit\", (code) =>\n      code === 0 ? resolve() : reject(new Error(`xclip exited ${code}`)),\n    );\n    child.stdin.end(text);\n  });\n  return { bytes: Buffer.byteLength(text), selection };\n}\n\nfunction xclipSelection(selection) {\n  return selection === \"primary\" ? \"primary\" : \"clipboard\";\n}\n\nasync function arrangeWindows(action) {\n  const scene = validateWindowScene(action.windows);\n  const deadline = Date.now() + action.waitForMs;\n  let resolved = null;\n  while (resolved === null) {\n    const windows = structured(await call(\"list_windows\", { on_screen_only: true })).windows;\n    resolved = resolveSceneWindows(scene, Array.isArray(windows) ? windows : []);\n    if (resolved !== null) break;\n    if (Date.now() >= deadline) {\n      throw new Error(\n        `arrange_windows timed out waiting for: ${missingSceneSelectors(scene, Array.isArray(windows) ? windows : []).join(\", \")}`,\n      );\n    }\n    await sleep(100);\n  }\n\n  // Resolve the complete scene before mutating any window. A missing selector\n  // therefore leaves the existing desktop untouched.\n  for (const item of resolved) {\n    const windowId = String(item.window.window_id);\n    await execFileAsync(\"xdotool\", [\"windowactivate\", \"--sync\", windowId], { env: displayEnv() });\n    await execFileAsync(\n      \"xdotool\",\n      [\"windowmove\", windowId, String(item.bounds.x), String(item.bounds.y)],\n      {\n        env: displayEnv(),\n      },\n    );\n    await execFileAsync(\n      \"xdotool\",\n      [\"windowsize\", windowId, String(item.bounds.width), String(item.bounds.height)],\n      { env: displayEnv() },\n    );\n  }\n\n  let verified = null;\n  let windows = [];\n  for (let attempt = 0; attempt < 10; attempt += 1) {\n    const response = structured(await call(\"list_windows\", { on_screen_only: true })).windows;\n    windows = Array.isArray(response) ? response : [];\n    verified = resolveSceneWindows(scene, windows);\n    if (\n      verified !== null &&\n      verified.every(({ bounds, window }) => approximatelyEqual(bounds, window))\n    ) {\n      return verified.map(layoutResult);\n    }\n    await sleep(100);\n  }\n\n  const screenshotPath = `${workspaceRoot}/layout-failure-${Date.now()}.png`;\n  await call(\"get_desktop_state\", { screenshot_out_file: screenshotPath });\n  const details =\n    verified === null\n      ? { stage: \"selector_resolution\", missing: missingSceneSelectors(scene, windows) }\n      : {\n          stage: \"verification\",\n          tolerance: layoutTolerance,\n          windows: verified.map(layoutResult),\n        };\n  throw new Error(\n    `arrange_windows failed: ${JSON.stringify({ ...details, screenshot: screenshotPath })}`,\n  );\n}\n\nfunction validateWindowScene(windows) {\n  if (!Array.isArray(windows) || windows.length < 1 || windows.length > 4) {\n    throw new Error(\"arrange_windows requires 1 to 4 windows\");\n  }\n  const scene = windows.map((item) => ({ ...item, bounds: regionBounds(item.region) }));\n  const selectors = scene.map((item) => selectorLabel(item.selector));\n  if (new Set(selectors).size !== selectors.length) {\n    throw new Error(\"arrange_windows selectors must be unique\");\n  }\n  for (let left = 0; left < scene.length; left += 1) {\n    for (let right = left + 1; right < scene.length; right += 1) {\n      if (overlaps(scene[left].bounds, scene[right].bounds)) {\n        throw new Error(\"arrange_windows regions must not overlap\");\n      }\n    }\n  }\n  return scene;\n}\n\nfunction desktopBounds() {\n  const screen = x11ScreenSize();\n  return {\n    x: desktopMargin,\n    y: desktopMargin,\n    width: screen.width - desktopMargin * 2,\n    height: screen.height - desktopMargin * 2,\n  };\n}\n\nfunction x11ScreenSize() {\n  // The X server is the source of truth after set_desktop_size. Avoid CUA's\n  // session-oriented get_screen_size contract, which differs from pointer\n  // actions and does not accept scope.\n  const output = execFileSync(\"xdpyinfo\", [\"-display\", \":99\"], {\n    encoding: \"utf8\",\n    timeout: 2_000,\n  });\n  const match = /dimensions:\\s+(\\d+)x(\\d+) pixels/u.exec(output);\n  if (match === null) throw new Error(\"could not determine desktop dimensions\");\n  return { width: Number(match[1]), height: Number(match[2]) };\n}\n\nfunction regionBounds(region) {\n  if (\n    typeof region !== \"object\" ||\n    region === null ||\n    ![region.x, region.y, region.width, region.height].every(Number.isFinite) ||\n    region.x < 0 ||\n    region.y < 0 ||\n    region.width <= 0 ||\n    region.height <= 0 ||\n    region.x + region.width > 1 ||\n    region.y + region.height > 1\n  ) {\n    throw new Error(\"arrange_windows regions must fit within the desktop\");\n  }\n  const desktop = desktopBounds();\n  const left = Math.round(desktop.x + desktop.width * region.x);\n  const top = Math.round(desktop.y + desktop.height * region.y);\n  const right = Math.round(desktop.x + desktop.width * (region.x + region.width));\n  const bottom = Math.round(desktop.y + desktop.height * (region.y + region.height));\n  const bounds = { x: left, y: top, width: right - left, height: bottom - top };\n  if (bounds.width < minimumWindow.width || bounds.height < minimumWindow.height) {\n    throw new Error(\n      `arrange_windows regions must be at least ${minimumWindow.width}x${minimumWindow.height}`,\n    );\n  }\n  return bounds;\n}\n\nfunction resolveSceneWindows(scene, windows) {\n  const resolved = scene.map((item) => {\n    const window = windows.find((candidate) => matchesSelector(candidate, item.selector));\n    return window === undefined ? null : { bounds: item.bounds, window };\n  });\n  return resolved.includes(null) ? null : resolved;\n}\n\nfunction missingSceneSelectors(scene, windows) {\n  return scene\n    .filter((item) => !windows.some((window) => matchesSelector(window, item.selector)))\n    .map((item) => selectorLabel(item.selector));\n}\n\nfunction matchesSelector(window, selector) {\n  const identity = `${window.app_name ?? \"\"}\\n${window.title ?? \"\"}`.toLowerCase();\n  if (selector.app === \"firefox\") return identity.includes(\"firefox\");\n  const title = selector.instance === \"primary\" ? \"terminal 1\" : \"terminal 2\";\n  return identity.includes(title);\n}\n\nfunction selectorLabel(selector) {\n  return selector.app === \"firefox\" ? \"firefox\" : `xterm:${selector.instance}`;\n}\n\nfunction overlaps(left, right) {\n  return (\n    left.x < right.x + right.width &&\n    left.x + left.width > right.x &&\n    left.y < right.y + right.height &&\n    left.y + left.height > right.y\n  );\n}\n\nfunction layoutResult({ bounds, window }) {\n  const observed = {\n    x: Number(window.x),\n    y: Number(window.y),\n    width: Number(window.width),\n    height: Number(window.height),\n  };\n  return {\n    selector: `${window.app_name ?? \"\"}:${window.title ?? \"\"}`,\n    requested: bounds,\n    observed,\n    delta: {\n      x: observed.x - bounds.x,\n      y: observed.y - bounds.y,\n      width: observed.width - bounds.width,\n      height: observed.height - bounds.height,\n    },\n    tolerance: layoutTolerance,\n    title: window.title,\n    window_id: window.window_id,\n  };\n}\n\nfunction displayEnv() {\n  return { ...env, DISPLAY: \":99\" };\n}\n\nfunction approximatelyEqual(bounds, window) {\n  return (\n    Math.abs(bounds.x - window.x) <= layoutTolerance.x &&\n    Math.abs(bounds.y - window.y) <= layoutTolerance.y &&\n    Math.abs(bounds.width - window.width) <= layoutTolerance.width &&\n    Math.abs(bounds.height - window.height) <= layoutTolerance.height\n  );\n}\n\nfunction sleep(milliseconds) {\n  return delay(milliseconds, undefined, { signal: requestSignal });\n}\n\nfunction scrollInput(delta, negativeDirection, positiveDirection) {\n  return {\n    scope: \"desktop\",\n    direction: delta < 0 ? negativeDirection : positiveDirection,\n    amount: Math.min(50, Math.max(1, Math.round(Math.abs(delta) / 100))),\n    by: \"line\",\n  };\n}\n\nfunction normalizeKey(key) {\n  const aliases = {\n    ARROWDOWN: \"down\",\n    ARROWLEFT: \"left\",\n    ARROWRIGHT: \"right\",\n    ARROWUP: \"up\",\n    CMD: \"meta\",\n    COMMAND: \"meta\",\n    CONTROL: \"ctrl\",\n    DEL: \"delete\",\n    ENTER: \"enter\",\n    ESC: \"escape\",\n    OPTION: \"alt\",\n    RETURN: \"enter\",\n  };\n  return aliases[key] ?? key.toLowerCase();\n}\n\nfunction isContentWindow(window) {\n  const identity = `${window.app_name ?? \"\"}\\n${window.title ?? \"\"}`.toLowerCase();\n  return (\n    window.width >= 200 &&\n    window.height >= 100 &&\n    !identity.includes(\"xfce4-panel\") &&\n    !identity.includes(\"xfdesktop\") &&\n    window.title !== \"Desktop\"\n  );\n}\n\nfunction formatState(groundingWindow, pointer, elements) {\n  const lines = [\n    groundingWindow === undefined\n      ? \"grounding_window\\t(none)\"\n      : `grounding_window\\t${groundingWindow.title} pid=${groundingWindow.pid} window_id=${groundingWindow.window_id}`,\n    `pointer\\tx=${pointer.x ?? \"unknown\"} y=${pointer.y ?? \"unknown\"}`,\n    \"accessibility_elements\",\n  ];\n\n  for (const element of elements) {\n    const frame =\n      element.frame === undefined\n        ? \"\"\n        : ` x=${element.frame.x} y=${element.frame.y} width=${element.frame.w} height=${element.frame.h}`;\n    lines.push(\n      `[${element.element_index}] role=${element.role}${frame} label=${JSON.stringify(element.label ?? \"\")}`,\n    );\n  }\n  return lines.join(\"\\n\");\n}\n\nfunction validateActionCoordinates(action) {\n  const screen = x11ScreenSize();\n  const points = [];\n  if ([\"click\", \"double_click\", \"triple_click\", \"move\", \"scroll\"].includes(action.action)) {\n    points.push({ x: action.x, y: action.y });\n  }\n  if (action.action === \"drag\") points.push(...action.path);\n  for (const point of points) {\n    if (\n      !Number.isFinite(point.x) ||\n      !Number.isFinite(point.y) ||\n      point.x < 0 ||\n      point.y < 0 ||\n      point.x >= screen.width ||\n      point.y >= screen.height\n    ) {\n      throw new Error(\n        `computer-use coordinate is outside the ${screen.width}x${screen.height} desktop`,\n      );\n    }\n  }\n}\n\nfunction validateRequest(input) {\n  if (typeof input !== \"object\" || input === null) {\n    throw new Error(\"request must be an object\");\n  }\n  if (!isArtifactPath(input.screenshotPath, \".png\")) {\n    throw new Error(\"screenshot path must be a PNG under the computer-use workspace\");\n  }\n  if (typeof input.action !== \"object\" || input.action === null) {\n    throw new Error(\"request action must be an object\");\n  }\n  const actions = new Set([\n    \"arrange_windows\",\n    \"click\",\n    \"clipboard_read\",\n    \"clipboard_write\",\n    \"double_click\",\n    \"drag\",\n    \"keypress\",\n    \"move\",\n    \"record_start\",\n    \"record_stop\",\n    \"screenshot\",\n    \"scroll\",\n    \"sequence\",\n    \"triple_click\",\n    \"type\",\n    \"wait\",\n  ]);\n  if (!actions.has(input.action.action)) {\n    throw new Error(`unsupported computer-use action: ${String(input.action.action)}`);\n  }\n  if (input.action.action === \"sequence\") {\n    if (\n      !Array.isArray(input.action.actions) ||\n      input.action.actions.length < 1 ||\n      input.action.actions.length > 30\n    ) {\n      throw new Error(\"sequence must contain 1 to 30 actions\");\n    }\n    for (const action of input.action.actions) validateNestedSequenceAction(action);\n  }\n  if (input.action.action === \"arrange_windows\") validateArrangeWindowsAction(input.action);\n  if (input.action.action === \"record_start\" && !isArtifactPath(input.action.path, \".mp4\")) {\n    throw new Error(\"recording path must be an MP4 under the computer-use workspace\");\n  }\n  if (\n    [\"click\", \"double_click\", \"triple_click\", \"drag\", \"move\"].includes(input.action.action) &&\n    \"durationMs\" in input.action &&\n    (!Number.isInteger(input.action.durationMs) ||\n      input.action.durationMs < 0 ||\n      input.action.durationMs > 2000)\n  ) {\n    throw new Error(\"pointer durationMs must be an integer from 0 to 2000\");\n  }\n  validateMovementStyle(input.action, \"\");\n  if (\n    input.action.action === \"wait\" &&\n    (!Number.isInteger(input.action.durationMs) ||\n      input.action.durationMs < 100 ||\n      input.action.durationMs > 10_000)\n  ) {\n    throw new Error(\"wait durationMs must be an integer from 100 to 10000\");\n  }\n  validateClipboardOptions(input.action);\n  validateTypingOptions(input.action, \"\");\n  validateTypingBudget(input.action.action === \"sequence\" ? input.action.actions : [input.action]);\n}\n\nfunction validateNestedSequenceAction(action) {\n  if (typeof action !== \"object\" || action === null)\n    throw new Error(\"sequence action must be an object\");\n  const allowed = new Set([\n    \"arrange_windows\",\n    \"click\",\n    \"double_click\",\n    \"drag\",\n    \"keypress\",\n    \"move\",\n    \"record_start\",\n    \"record_stop\",\n    \"triple_click\",\n    \"type\",\n    \"wait\",\n  ]);\n  if (!allowed.has(action.action))\n    throw new Error(`unsupported sequence action: ${String(action.action)}`);\n  if (action.action === \"arrange_windows\") validateArrangeWindowsAction(action);\n  if (action.action === \"record_start\" && !isArtifactPath(action.path, \".mp4\")) {\n    throw new Error(\"sequence recording path must be an MP4 under the computer-use workspace\");\n  }\n  if (\n    [\"click\", \"double_click\", \"triple_click\", \"drag\", \"move\"].includes(action.action) &&\n    \"durationMs\" in action &&\n    (!Number.isInteger(action.durationMs) || action.durationMs < 0 || action.durationMs > 2000)\n  ) {\n    throw new Error(\"sequence pointer durationMs must be an integer from 0 to 2000\");\n  }\n  validateMovementStyle(action, \"sequence \");\n  if (\n    action.action === \"wait\" &&\n    (!Number.isInteger(action.durationMs) || action.durationMs < 100 || action.durationMs > 10_000)\n  ) {\n    throw new Error(\"sequence wait durationMs must be an integer from 100 to 10000\");\n  }\n  if (action.action === \"type\") validateTypingOptions(action, \"sequence \");\n}\n\nfunction validateMovementStyle(action, prefix) {\n  if (\n    \"movementStyle\" in action &&\n    action.movementStyle !== \"natural\" &&\n    action.movementStyle !== \"precision\"\n  ) {\n    throw new Error(`${prefix}movementStyle must be natural or precision`);\n  }\n  if (\n    action.action === \"drag\" &&\n    \"movementStyle\" in action &&\n    action.movementStyle !== \"precision\"\n  ) {\n    throw new Error(`${prefix}drag movementStyle must be precision`);\n  }\n}\n\nfunction validateClipboardOptions(action) {\n  if (\n    [\"clipboard_read\", \"clipboard_write\"].includes(action.action) &&\n    action.selection !== \"clipboard\" &&\n    action.selection !== \"primary\"\n  ) {\n    throw new Error(\"clipboard selection must be clipboard or primary\");\n  }\n  if (\n    action.action === \"clipboard_write\" &&\n    (typeof action.text !== \"string\" || Buffer.byteLength(action.text) > maxClipboardBytes)\n  ) {\n    throw new Error(`clipboard text must be at most ${maxClipboardBytes} bytes`);\n  }\n}\n\nfunction validateTypingOptions(action, prefix) {\n  if (\n    \"typingDelayMs\" in action &&\n    (!Number.isInteger(action.typingDelayMs) ||\n      action.typingDelayMs < 0 ||\n      action.typingDelayMs > 200)\n  ) {\n    throw new Error(`${prefix}typingDelayMs must be an integer from 0 to 200`);\n  }\n  if (\"typingStyle\" in action && action.typingStyle !== \"natural\") {\n    throw new Error(`${prefix}typingStyle must be natural`);\n  }\n  if (\"typingStyle\" in action && \"typingDelayMs\" in action) {\n    throw new Error(`${prefix}type action cannot set both typingStyle and typingDelayMs`);\n  }\n}\n\nfunction validateArrangeWindowsAction(action) {\n  if (!Number.isInteger(action.waitForMs) || action.waitForMs < 0 || action.waitForMs > 10_000) {\n    throw new Error(\"arrange_windows waitForMs must be an integer from 0 to 10000\");\n  }\n  validateWindowScene(action.windows);\n}\n\nfunction isArtifactPath(value, extension) {\n  if (typeof value !== \"string\" || !value.endsWith(extension)) return false;\n  const path = resolve(value);\n  return dirname(path) === workspaceRoot;\n}\n\nasync function readBody(request) {\n  let body = \"\";\n  for await (const chunk of request) {\n    body += chunk;\n    if (Buffer.byteLength(body) > maxRequestBytes) {\n      throw new Error(\"request exceeds 32 KiB\");\n    }\n  }\n  return body;\n}\n\nfunction sendJson(response, statusCode, body) {\n  if (response.destroyed) return;\n  const json = JSON.stringify(body);\n  response.writeHead(statusCode, {\n    \"content-length\": Buffer.byteLength(json),\n    \"content-type\": \"application/json\",\n  });\n  response.end(json);\n}\n";
+    readonly "client.mjs": "import http from \"node:http\";\nimport { env } from \"node:process\";\n\nconst socketPath = env.COMPUTER_USE_SOCKET_PATH;\nconst body = env.COMPUTER_USE_REQUEST;\n\nif (body === undefined) {\n  throw new Error(\"COMPUTER_USE_REQUEST is required\");\n}\n\nconst deadline = Date.now() + 120_000;\nconst response = await new Promise((resolve, reject) => {\n  const request = http.request(\n    {\n      socketPath,\n      method: \"POST\",\n      path: \"/call\",\n      headers: {\n        \"content-length\": Buffer.byteLength(body),\n        \"content-type\": \"application/json\",\n        \"x-computer-use-deadline\": String(deadline),\n      },\n    },\n    (incoming) => {\n      incoming.on(\"error\", reject);\n      incoming.setEncoding(\"utf8\");\n      let responseBody = \"\";\n      incoming.on(\"data\", (chunk) => {\n        responseBody += chunk;\n      });\n      incoming.on(\"end\", () => {\n        resolve({ body: responseBody, statusCode: incoming.statusCode ?? 500 });\n      });\n    },\n  );\n  const timer = setTimeout(\n    () => request.destroy(new Error(\"computer-use request deadline exceeded\")),\n    Math.max(0, deadline - Date.now()),\n  );\n  request.once(\"close\", () => clearTimeout(timer));\n  request.on(\"error\", reject);\n  request.end(body);\n});\n\nif (response.statusCode !== 200) {\n  process.stderr.write(response.body);\n  process.exitCode = 1;\n} else {\n  process.stdout.write(response.body);\n}\n";
+    readonly "state.mjs": "const MAX_GROUNDING_ELEMENTS = 80;\n\n/**\n * @typedef {{\n *   element_index?: number;\n *   frame?: { h?: number; w?: number; x?: number; y?: number };\n *   label?: unknown;\n *   [key: string]: unknown;\n * }} GroundingElement\n */\n\n/**\n * Keep the model-facing state aligned with the visible desktop screenshot.\n *\n * @param {unknown[]} elements\n * @returns {GroundingElement[]}\n */\nexport function selectGroundingElements(elements, screen = { width: 1920, height: 1080 }) {\n  return elements\n    .filter((element) => isVisibleLabeledElement(element, screen))\n    .slice(0, MAX_GROUNDING_ELEMENTS);\n}\n\n/**\n * @param {unknown} value\n * @returns {value is GroundingElement}\n */\nfunction isVisibleLabeledElement(value, screen) {\n  if (typeof value !== \"object\" || value === null) return false;\n  const element = /** @type {GroundingElement} */ (value);\n  const frame = element.frame;\n  if (typeof element.label !== \"string\" || element.label.trim().length === 0) return false;\n  if (typeof frame !== \"object\" || frame === null) return false;\n  if (\n    !Number.isFinite(frame.x) ||\n    !Number.isFinite(frame.y) ||\n    !Number.isFinite(frame.w) ||\n    !Number.isFinite(frame.h)\n  ) {\n    return false;\n  }\n\n  const x = /** @type {number} */ (frame.x);\n  const y = /** @type {number} */ (frame.y);\n  const width = /** @type {number} */ (frame.w);\n  const height = /** @type {number} */ (frame.h);\n  return (\n    width > 0 &&\n    height > 0 &&\n    x < screen.width &&\n    y < screen.height &&\n    x + width > 0 &&\n    y + height > 0\n  );\n}\n";
+    readonly "typing.mjs": "const DEMO_TYPING_SPEED = 1.2;\nconst MIN_DELAY_MS = 35;\nconst MAX_DELAY_MS = 220;\n\nexport function naturalTypingPlan(text) {\n  const graphemes = [...new Intl.Segmenter().segment(text)].map(({ segment }) => segment);\n  let state = hashText(text) || 1;\n  return graphemes.map((grapheme, index) => {\n    state = nextState(state);\n    return {\n      grapheme,\n      delayMs: index === graphemes.length - 1 ? 0 : naturalDelay(grapheme, state),\n    };\n  });\n}\n\nfunction naturalDelay(grapheme, state) {\n  const jitter = state / 0x1_0000_0000;\n  const range = delayRange(grapheme);\n  return Math.round((range.min + (range.max - range.min) * jitter) / DEMO_TYPING_SPEED);\n}\n\nfunction delayRange(grapheme) {\n  if (/\\s/u.test(grapheme)) return { min: 75, max: 125 };\n  if (/[.!?]/u.test(grapheme)) return { min: 145, max: MAX_DELAY_MS };\n  if (/[/|;&]/u.test(grapheme)) return { min: 105, max: 175 };\n  if (/[-_=,:]/u.test(grapheme)) return { min: 70, max: 125 };\n  return { min: MIN_DELAY_MS, max: 85 };\n}\n\nfunction hashText(text) {\n  let hash = 2166136261;\n  for (const character of text) {\n    hash ^= character.codePointAt(0);\n    hash = Math.imul(hash, 16777619);\n  }\n  return hash >>> 0;\n}\n\nfunction nextState(state) {\n  let value = state;\n  value ^= value << 13;\n  value ^= value >>> 17;\n  value ^= value << 5;\n  return value >>> 0;\n}\n";
+    readonly "pointer-motion.mjs": "const FRAME_MS = 16;\n\nexport function pointerMotionPlan(start, end, options = {}) {\n  const style = options.style ?? \"natural\";\n  const distance = Math.hypot(end.x - start.x, end.y - start.y);\n  const durationMs = options.durationMs ?? naturalDuration(distance, style);\n  if (distance < 2 || durationMs === 0) {\n    return { phases: [[{ ...end, delayMs: 0 }]], settleMs: 0 };\n  }\n\n  if (style === \"precision\" && distance >= 120) {\n    const ratio = 0.92 + deterministicUnit(start, end, 7) * 0.04;\n    const undershoot = {\n      x: Math.round(start.x + (end.x - start.x) * ratio),\n      y: Math.round(start.y + (end.y - start.y) * ratio),\n    };\n    // Cover most of the distance quickly, then spend a disproportionate\n    // share of the duration continuously correcting the final 4–8%.\n    const firstDuration = Math.round(durationMs * 0.62);\n    return {\n      phases: [\n        bezierPhase(start, undershoot, firstDuration, \"natural\", 11),\n        bezierPhase(undershoot, end, durationMs - firstDuration, \"precision\", 23),\n      ],\n      settleMs: 0,\n    };\n  }\n\n  return {\n    phases: [bezierPhase(start, end, durationMs, style, 17)],\n    settleMs: style === \"natural\" ? 15 + Math.round(deterministicUnit(start, end, 31) * 15) : 0,\n  };\n}\n\nfunction bezierPhase(start, end, durationMs, style, salt) {\n  const distance = Math.hypot(end.x - start.x, end.y - start.y);\n  const steps = Math.max(2, Math.min(60, Math.round(durationMs / FRAME_MS)));\n  const perpendicular = unitPerpendicular(start, end);\n  const bendLimit = style === \"precision\" ? 8 : Math.min(70, distance * 0.09);\n  const bend = (deterministicUnit(start, end, salt) * 2 - 1) * bendLimit;\n  const c1 = {\n    x: start.x + (end.x - start.x) * 0.28 + perpendicular.x * bend,\n    y: start.y + (end.y - start.y) * 0.28 + perpendicular.y * bend,\n  };\n  const c2 = {\n    x: start.x + (end.x - start.x) * 0.76 + perpendicular.x * bend * 0.65,\n    y: start.y + (end.y - start.y) * 0.76 + perpendicular.y * bend * 0.65,\n  };\n  const delayMs = durationMs / steps;\n  return Array.from({ length: steps }, (_, index) => {\n    // This asymmetric timing curve accelerates quickly and eases into the\n    // target without the perfectly mirrored feel of smoothstep.\n    const progress = (index + 1) / steps;\n    const t = 1 - (1 - progress) ** 2.35;\n    const point = cubicBezier(start, c1, c2, end, t);\n    return {\n      x: index === steps - 1 ? end.x : Math.round(point.x),\n      y: index === steps - 1 ? end.y : Math.round(point.y),\n      delayMs: index === steps - 1 ? 0 : delayMs,\n    };\n  });\n}\n\nfunction cubicBezier(start, c1, c2, end, t) {\n  const inverse = 1 - t;\n  return {\n    x:\n      inverse ** 3 * start.x +\n      3 * inverse ** 2 * t * c1.x +\n      3 * inverse * t ** 2 * c2.x +\n      t ** 3 * end.x,\n    y:\n      inverse ** 3 * start.y +\n      3 * inverse ** 2 * t * c1.y +\n      3 * inverse * t ** 2 * c2.y +\n      t ** 3 * end.y,\n  };\n}\n\nfunction naturalDuration(distance, style) {\n  const base = style === \"precision\" ? 150 : 110;\n  const scale = style === \"precision\" ? 0.35 : 0.28;\n  const minimum = style === \"precision\" ? 180 : 140;\n  const maximum = style === \"precision\" ? 650 : 550;\n  return Math.round(Math.min(maximum, Math.max(minimum, base + distance * scale)));\n}\n\nfunction unitPerpendicular(start, end) {\n  const dx = end.x - start.x;\n  const dy = end.y - start.y;\n  const length = Math.hypot(dx, dy) || 1;\n  return { x: -dy / length, y: dx / length };\n}\n\nfunction deterministicUnit(start, end, salt) {\n  let value =\n    (Math.imul(Math.round(start.x), 73_856_093) ^\n      Math.imul(Math.round(start.y), 19_349_663) ^\n      Math.imul(Math.round(end.x), 83_492_791) ^\n      Math.imul(Math.round(end.y), 2_654_435_761) ^\n      salt) >>>\n    0;\n  value ^= value << 13;\n  value ^= value >>> 17;\n  value ^= value << 5;\n  return (value >>> 0) / 0x1_0000_0000;\n}\n";
+    readonly "package.json": "{\n  \"name\": \"computer-use-driver\",\n  \"private\": true,\n  \"type\": \"module\",\n  \"dependencies\": {\n    \"@trycua/cua-driver\": \"0.12.5\"\n  }\n}\n";
+};