AgentHub/server/lib/native-dialog.js

/**
 * @file native-dialog.js
 * @description Opens the operating system's own folder/file picker ON THE
 * SERVER'S DESKTOP and returns the absolute path the user chose. A browser's
 * native picker never hands a page an absolute path, and the dashboard is
 * local-first, so the server asks the desktop itself: zenity or kdialog on
 * Linux (with the graphical session's variables from `desktop-env.js`, which a
 * systemd user service lacks), `osascript` on macOS, PowerShell's WinForms
 * dialogs on Windows. Runs with `execFile` (never a shell, never sync — a
 * dialog waits on a human for minutes), one dialog at a time. Nothing from the
 * request reaches argv except the starting folder, resolved to an absolute
 * path first. macOS and Windows are code-only, untested on those systems.
 * @author Nguyễn Ngọc Trí Vĩ <vinnt@smartgift.vn>
 */

const { execFile } = require("node:child_process");
const fs = require("node:fs");
const os = require("node:os");
const path = require("node:path");
const { desktopEnv } = require("./desktop-env");

const TITLES = { dir: "Select a folder", file: "Select a file" };
const TIMEOUT_MS = 10 * 60 * 1000;

const defaultExec = (cmd, args, env, signal) =>
  new Promise((resolve) => {
    execFile(
      cmd,
      args,
      { env, timeout: TIMEOUT_MS, windowsHide: true, encoding: "utf8", signal },
      (err, stdout, stderr) =>
        resolve({
          missing: err?.code === "ENOENT",
          code: err ? (typeof err.code === "number" ? err.code : -1) : 0,
          stdout: String(stdout || ""),
          stderr: String(stderr || ""),
        })
    );
  });

let execImpl = defaultExec;

/** Test seam — same pattern as desktop-env.js; `null` restores the real one. */
function __setExecImpl(fn) {
  execImpl = fn || defaultExec;
}

/** An existing folder to open the dialog in: `start` (or its parent when it is a file), else home. */
function startDir(start) {
  if (typeof start === "string" && start.trim()) {
    const raw = start.trim().replace(/^~(?=$|[\\/])/, os.homedir());
    const abs = path.resolve(raw);
    for (const p of [abs, path.dirname(abs)]) {
      try {
        if (fs.statSync(p).isDirectory()) return p;
      } catch {
        // missing — try the parent, then home
      }
    }
  }
  return os.homedir();
}

// user32 calls PowerShell has no cmdlet for: `Front` makes a window the
// foreground one from a background process (see dialogCommands).
const WIN_FRONT = [
  '[DllImport("user32.dll")] public static extern bool ShowWindow(IntPtr h, int n);',
  '[DllImport("user32.dll")] public static extern bool SetForegroundWindow(IntPtr h);',
  '[DllImport("user32.dll")] public static extern bool BringWindowToTop(IntPtr h);',
  '[DllImport("user32.dll")] public static extern IntPtr GetForegroundWindow();',
  '[DllImport("user32.dll")] public static extern uint GetWindowThreadProcessId(IntPtr h, IntPtr p);',
  '[DllImport("user32.dll")] public static extern bool AttachThreadInput(uint a, uint b, bool on);',
  '[DllImport("kernel32.dll")] public static extern uint GetCurrentThreadId();',
  "public static void Front(IntPtr h) {",
  "  uint fg = GetWindowThreadProcessId(GetForegroundWindow(), IntPtr.Zero);",
  "  uint me = GetCurrentThreadId();",
  "  bool joined = fg != 0 && fg != me && AttachThreadInput(me, fg, true);",
  "  BringWindowToTop(h); SetForegroundWindow(h);",
  "  if (joined) AttachThreadInput(me, fg, false);",
  "}",
].join("\n");

/** The candidate dialog commands for a platform, in order of preference. */
function dialogCommands(kind, dir, platform = process.platform) {
  const title = TITLES[kind];
  if (platform === "darwin") {
    const what = kind === "dir" ? "folder" : "file";
    // osascript started by a background server is not the frontmost app, so
    // its dialog opened BEHIND the browser; `activate` brings it forward.
    return [
      [
        "osascript",
        [
          "-e",
          "on run argv",
          "-e",
          "activate",
          "-e",
          `POSIX path of (choose ${what} with prompt (item 1 of argv) default location (POSIX file (item 2 of argv)))`,
          "-e",
          "end run",
          title,
          dir,
        ],
      ],
    ];
  }
  if (platform === "win32") {
    const dialog =
      kind === "dir"
        ? "$d = New-Object System.Windows.Forms.FolderBrowserDialog; $d.SelectedPath = $env:AGENTHUB_PICK_START; $d.Description = $env:AGENTHUB_PICK_TITLE"
        : "$d = New-Object System.Windows.Forms.OpenFileDialog; $d.InitialDirectory = $env:AGENTHUB_PICK_START; $d.Title = $env:AGENTHUB_PICK_TITLE";
    const result = kind === "dir" ? "$d.SelectedPath" : "$d.FileName";
    // The server is a background process, and Windows keeps one from bringing
    // a window forward: an ownerless dialog opened at the BOTTOM of the
    // z-order, behind the browser. So the dialog gets a TopMost owner, made
    // the foreground window by joining the current foreground thread's input
    // (AttachThreadInput — the lock exempts it). The first ShowWindow of a
    // process started with windowsHide is turned into a hide; a throwaway call
    // spends it, or the owner stays hidden and takes the dialog down with it.
    const script = [
      "$ProgressPreference = 'SilentlyContinue'",
      "[Console]::OutputEncoding = [Text.Encoding]::UTF8",
      "Add-Type -AssemblyName System.Windows.Forms",
      `Add-Type -Namespace AgentHub -Name Win -MemberDefinition '${WIN_FRONT}'`,
      "$o = New-Object System.Windows.Forms.Form",
      "$o.TopMost = $true; $o.ShowInTaskbar = $false; $o.FormBorderStyle = 'None'; $o.Opacity = 0; $o.Width = 1; $o.Height = 1; $o.StartPosition = 'CenterScreen'",
      "[void][AgentHub.Win]::ShowWindow($o.Handle, 1)",
      "$o.Show()",
      "[AgentHub.Win]::Front($o.Handle)",
      dialog,
      `if ($d.ShowDialog($o) -eq 'OK') { [Console]::Out.Write(${result}) } else { exit 1 }`,
    ].join("\n");
    // -EncodedCommand: the C# above carries double quotes, which Windows
    // PowerShell's own command-line parsing mangles inside -Command.
    return [
      [
        "powershell.exe",
        [
          "-NoProfile",
          "-STA",
          "-EncodedCommand",
          Buffer.from(script, "utf16le").toString("base64"),
        ],
      ],
    ];
  }
  const sep = dir.endsWith("/") ? "" : "/";
  return [
    [
      "zenity",
      [
        "--file-selection",
        `--title=${title}`,
        `--filename=${dir}${sep}`,
        ...(kind === "dir" ? ["--directory"] : []),
      ],
    ],
    [
      "kdialog",
      [kind === "dir" ? "--getexistingdirectory" : "--getopenfilename", dir, "--title", title],
    ],
  ];
}

// A dialog that could not reach the desktop exits 1 too — a stale DISPLAY
// handed over by `systemctl --user show-environment` after a logout, say.
const NO_DISPLAY_RE =
  /cannot open display|could not connect to display|failed to open display|cannot connect to x server|unable to init server/i;

/**
 * Whether an exit status of 1 means the user pressed Cancel. zenity/kdialog
 * may print GTK/Qt warnings on a normal Cancel, so only a display failure
 * counts against it; osascript says Cancel as error -128 and nothing else.
 */
function isCancel(cmd, stderr = "") {
  if (cmd === "osascript") return /-128/.test(stderr);
  if (cmd === "powershell.exe") return !stderr.trim();
  return !NO_DISPLAY_RE.test(stderr);
}

function fail(code, message) {
  return Object.assign(new Error(message), { code });
}

let busy = false;

/**
 * Show the native picker and wait for the user.
 * @param {"dir"|"file"} kind
 * @param {string} [start] Folder (or file) to open the dialog at.
 * @param {AbortSignal} [signal] Closes the dialog — the browser went away.
 * @returns {Promise<string|null>} The chosen absolute path, or null when cancelled.
 * @throws `ENODIALOG` (no desktop or no dialog program), `EBUSY` (one is already open), `EDIALOG`.
 */
async function pickPath(kind, start, signal) {
  if (busy) throw fail("EBUSY", "a picker is already open on the desktop");
  const env = { ...process.env, ...desktopEnv() };
  if (process.platform === "linux" && !env.DISPLAY && !env.WAYLAND_DISPLAY) {
    throw fail("ENODIALOG", "no graphical desktop session to show a picker on");
  }
  const dir = startDir(start);
  env.AGENTHUB_PICK_START = dir;
  env.AGENTHUB_PICK_TITLE = TITLES[kind];
  busy = true;
  try {
    for (const [cmd, args] of dialogCommands(kind, dir)) {
      const r = await execImpl(cmd, args, env, signal);
      if (r.missing) continue;
      if (r.code === 1) {
        if (isCancel(cmd, r.stderr)) return null; // every one of these exits 1 on Cancel
        throw fail("ENODIALOG", `${cmd} could not open a picker on the desktop`);
      }
      const out = r.stdout.replace(/[\r\n]+$/, "");
      if (r.code !== 0 || !out) throw fail("EDIALOG", `${cmd} exited with ${r.code}`);
      // osascript ends a folder with "/"; "/" itself stays.
      return out.replace(/\/+$/, "") || "/";
    }
    throw fail("ENODIALOG", "no native picker found (install zenity or kdialog)");
  } finally {
    busy = false;
  }
}

module.exports = { pickPath, dialogCommands, startDir, isCancel, __setExecImpl };