Files
crossdesk/apps/mcp/src/sessions.ts
T

1149 lines
52 KiB
TypeScript

import { join } from "node:path";
import { mkdir } from "node:fs/promises";
import { setTimeout as delay } from "node:timers/promises";
import {
Failure,
failure,
token,
equal,
parse,
uiConnect,
unlockPassword,
type Data,
type Panel,
type VideoOptions,
} from "./common.js";
import { DeviceStore, type Device } from "./store.js";
import { nativeBinary, NativeClient } from "./native.js";
import { displayEnvironment } from "./environment.js";
// Keep the public error categories while exposing only known native reasons.
const connectionErrors = new Map<string, string>([
["incorrect_password", "AUTH_FAILED"],
["signaling_timeout", "CONNECTION_TIMEOUT"],
["connection_timeout", "CONNECTION_TIMEOUT"],
["signaling_login_failed", "CONNECTION_FAILED"],
["signaling_disconnected", "CONNECTION_FAILED"],
["tls_certificate_error", "CONNECTION_FAILED"],
["cli_identity_not_supported", "CONNECTION_FAILED"],
["join_request_failed", "CONNECTION_FAILED"],
["device_not_found", "CONNECTION_FAILED"],
["device_offline_or_unavailable", "CONNECTION_FAILED"],
["connection_lost", "CONNECTION_FAILED"],
["connection_failed", "CONNECTION_FAILED"],
["remote_disconnected", "CONNECTION_FAILED"],
["display_not_available", "CONNECTION_FAILED"],
["media_initialization_failed", "CONNECTION_FAILED"],
["control_event_overflow", "CONNECTION_FAILED"],
]);
export interface View {
id: string;
generation: number;
secret: string;
panel: Exclude<Panel, "auto">;
created: number;
ready: boolean;
lastSeen: number;
alias?: string;
session?: string;
disconnected?: boolean;
connectionError?: Data;
}
interface Session {
id: string;
device: Device;
native: NativeClient;
status: Data;
release: () => Promise<void>;
connecting: boolean;
connection: AbortController;
displayBatch?: boolean;
mode: string;
desiredControl: boolean;
audioEnabled?: boolean;
recoveryEnabled: boolean;
recoveryAttempts: number;
recoveryAttempt: number;
recovering?: Promise<void>;
recoveryAbort?: AbortController;
monitor?: ReturnType<typeof setInterval>;
recoveredFrom?: string;
statusRequest?: Promise<Data>;
closing?: Promise<void>;
unlock?: UnlockRequest;
unlocks: Map<string, UnlockRequest>;
}
interface UnlockRequest {
id: string;
operation: string;
observation: string;
useSaved: boolean;
state: "preparing" | "awaiting_password" | "submitting" | "verifying" | "verified" | "submitted" | "failed" | "cancelled";
verification?: "locked" | "unlocked" | "unknown";
expires: number;
generation?: number;
geometry?: number;
error?: Data;
result?: Data;
persistenceError?: string;
saved?: boolean;
controller: AbortController;
}
export class Sessions {
readonly views = new Map<string, View>();
private current?: Session;
private active = new Map<string, Session>();
private replacements = new Map<string, string>();
private captures = new Map<string, { session: string; native: NativeClient; created: number; displays: Data[] }>();
private drags = new Map<string, { arguments: string; result: Data }>();
private connecting = new Set<AbortController>();
private connectionDone = new Set<Promise<void>>();
private closing = false;
private binary?: Promise<string>;
constructor(
readonly store: DeviceStore,
readonly policy: Panel,
private nativePath?: string,
private videoOptions?: VideoOptions,
) {}
choose(apps: boolean): View["panel"] {
if (this.policy === "apps" && !apps) throw new Failure("PANEL_UNSUPPORTED");
if (this.policy !== "auto") return this.policy;
const environment = displayEnvironment();
return environment.gui ? "browser" : environment.terminal ? "terminal" : "off";
}
create(panel: View["panel"], alias?: string): View {
for (const [id, view] of this.views)
if (Date.now() - view.lastSeen > 30 * 60 * 1000) {
void this.revokeVideo(view);
this.views.delete(id);
}
if (this.views.size >= 32) throw new Failure("VIEW_LIMIT");
const selected = alias ? [...this.active.values()].find((s) => s.device.alias === alias && !s.closing)
: this.current && !this.current.closing ? this.current : [...this.active.values()].find((s) => !s.closing);
const view: View = {
id: token(),
generation: 1,
secret: token(),
panel,
created: Date.now(),
lastSeen: Date.now(),
ready: false,
alias: alias ?? selected?.device.alias,
session: selected?.id,
};
this.views.set(view.id, view);
return view;
}
view(id: string): View {
const v = this.views.get(id);
if (!v) throw new Failure("VIEW_NOT_FOUND");
return v;
}
authenticate(id: string, generation: number, secret: string): View {
const v = this.view(id);
if (v.generation !== generation || !equal(v.secret, secret))
throw new Failure("VIEW_UNAUTHORIZED");
return v;
}
describe(v: View): Data {
return {
view_id: v.id,
view_generation: v.generation,
panel: v.panel,
panel_status: v.ready
? "ready"
: Date.now() - v.created > 10000
? "not_confirmed"
: "waiting",
session_id: v.session ?? null,
state: v.session ? "session_available" : v.disconnected ? "closed" : v.connectionError ? "failed" : "awaiting_user",
...(v.connectionError ? { error: v.connectionError } : {}),
device_alias: v.alias ?? null,
next_step: v.session
? "Call crossdesk_get_session with view_id."
: v.disconnected || v.connectionError
? "The session was disconnected. Do not automatically reconnect."
: v.panel === "off"
? "Use setup to configure credentials and authorization, then crossdesk_connect with the saved device alias. Never send passwords in chat."
: "The user selects and authorizes a device in the panel. To continue a pending remote task, immediately call crossdesk_get_session with this view_id and wait_ms: 25000; repeat on wait_timed_out. Do not ask the user to send a connection confirmation message.",
};
}
private session(id: string): Session {
if (this.replacements.has(id)) throw new Failure("SESSION_REPLACED");
const session = this.active.get(id);
if (!session) throw new Failure("SESSION_NOT_FOUND");
return session;
}
async status(sessionId?: string, viewId?: string, signal?: AbortSignal): Promise<Data> {
const view = viewId ? this.view(viewId) : undefined;
const id = sessionId ?? view?.session;
if (!id && view) return this.describe(view);
const replacement = this.replacements.get(id!);
if (replacement) return { ...await this.status(replacement, undefined, signal), previous_session_id: id };
const s = this.session(id!);
if (s.recovering) return {
...s.status, error: "", frame_age_ms: null,
...(view ? this.describe(view) : {}), session_id: s.id, device_alias: s.device.alias,
state: "reconnecting", control: false, reconnect_enabled: s.recoveryEnabled,
recovery_attempt: s.recoveryAttempt,
next_step: "Authorized connection recovery is running. Wait with crossdesk_get_session. Use the returned session_id and a fresh screenshot after recovery. Query uncertain task outcomes without replaying input.",
};
// A panel connection may still be performing the native hello handshake.
// Reading native status here would race initialization and could close it.
if (s.connecting) return {
...(view ? this.describe(view) : {}),
session_id: s.id,
device_alias: s.device.alias,
state: "connecting",
control: false,
};
// Concurrent panel/model polls share one native read. A cancelled waiter
// must not cancel another caller's status request or disconnect the session.
if (signal?.aborted) throw new Failure("CANCELLED");
if (!["closed", "failed"].includes(s.status.state)) {
s.statusRequest ??= s.native.request("status", {}, 5000)
.then((status) => (s.status = status))
.catch((error) => {
if (!(error instanceof Failure) || error.code !== "SESSION_CLOSED") throw error;
return (s.status = { ...s.status, state: "closed", control: false, error: "native_process_closed" });
})
.finally(() => { s.statusRequest = undefined; });
await this.waitStatus(s.statusRequest, signal);
}
if (s.recovering) return this.status(sessionId, viewId, signal);
// The user may have disconnected or replaced the session during the read.
if (this.active.get(s.id) !== s || (view && view.session !== id))
return view ? this.describe(view) : { session_id: id, state: "closed", control: false };
if (["closed", "failed"].includes(s.status.state)) {
if (this.startRecovery(s)) return this.status(sessionId, viewId, signal);
await this.dispose(s);
}
this.checkUnlock(s);
return {
...s.status,
session_id: s.id,
device_alias: s.device.alias,
reconnect_enabled: s.recoveryEnabled,
recovery_attempt: s.recoveryAttempt,
...(s.recoveredFrom ? { recovered_from: s.recoveredFrom } : {}),
...(s.unlock ? { unlock: this.unlockStatus(s.unlock) } : {}),
...(view ? this.describe(view) : {}),
state: s.status.state,
...(s.status.state === "ready" ? {
next_step: "Connection ready. Continue the user's pending task within the current authorization. Obtain a screenshot before input; respect human_paused and control. No user confirmation message is needed.",
} : ["closed", "failed", "disconnected"].includes(s.status.state) ? {
next_step: "The connection ended. Stop waiting and do not automatically reconnect or replay actions.",
} : {}),
...(s.unlock && this.unlockPending(s.unlock) ? { next_step: this.unlockStatus(s.unlock).next_step } : {}),
};
}
private async waitStatus<T>(request: Promise<T>, signal?: AbortSignal) {
if (!signal) return request;
let abort: () => void = () => {};
try {
return await Promise.race([request, new Promise<never>((_, reject) => {
abort = () => reject(new Failure("CANCELLED"));
signal.addEventListener("abort", abort, { once: true });
if (signal.aborted) abort();
})]);
} finally {
signal.removeEventListener("abort", abort);
}
}
private dispose(s: Session): Promise<void> {
this.cancelUnlock(s);
clearInterval(s.monitor);
s.monitor = undefined;
return s.closing ??= s.native.close().finally(() => s.release());
}
private monitor(s: Session) {
clearInterval(s.monitor);
if (!s.recoveryEnabled || s.closing) return;
s.monitor = setInterval(() => {
if (s.recovering || s.connecting || this.closing) return;
void this.status(s.id).catch((error) => {
if (this.active.get(s.id) !== s || s.closing || s.recovering) return;
if (failure(error).code === "NATIVE_TIMEOUT") {
s.status = { ...s.status, state: "failed", error: "native_process_unresponsive" };
this.startRecovery(s);
}
});
}, 1000);
s.monitor.unref();
}
private startRecovery(s: Session): boolean {
if (!s.recoveryEnabled || s.closing || this.closing || s.status.human_paused ||
!["connection_lost", "signaling_disconnected", "native_process_closed", "native_process_unresponsive"].includes(s.status.error))
return false;
if (s.recovering) return true;
clearInterval(s.monitor);
const controller = s.recoveryAbort = new AbortController();
s.recovering = this.recover(s, controller.signal).finally(() => { s.recovering = undefined; });
return true;
}
private async recover(s: Session, signal: AbortSignal) {
const views = [...this.views.values()].filter((view) => view.session === s.id);
const display = s.status.display ?? 0;
try { await this.dispose(s); }
catch (error) {
s.recoveryEnabled = false;
s.status = { ...s.status, state: "failed", recovery_error: failure(error) };
return;
}
const deadline = Date.now() + 60000;
for (let attempt = 1; attempt <= s.recoveryAttempts && Date.now() < deadline; attempt++) {
s.recoveryAttempt = attempt;
try {
await delay(Math.min(500 * 3 ** (attempt - 1), 5000), undefined, { signal });
if (!s.recoveryEnabled || this.closing || signal.aborted) throw new Failure("CANCELLED");
const device = await this.store.device(s.device.alias);
if (device.device_id !== s.device.device_id || device.server !== s.device.server || device.port !== s.device.port)
throw new Failure("DEVICE_ALIAS_CONFLICT");
const timeout = AbortSignal.timeout(Math.max(1, deadline - Date.now()));
const result = await this.connectSaved(s.device.alias, s.mode, display,
AbortSignal.any([signal, timeout]), s.desiredControl);
const next = this.session(result.session_id);
if (signal.aborted || this.closing || !s.recoveryEnabled) {
await this.disconnect(next.id);
throw new Failure("CANCELLED");
}
if (!s.desiredControl) await next.native.request("control", { enabled: false });
next.desiredControl = s.desiredControl;
if (next.status.preview_supported && s.status.preview_supported) {
const settings: Data = { fps: s.status.preview_fps, quality: s.status.preview_quality };
if (next.status.preview_resolution_supported) Object.assign(settings, {
width: s.status.preview_width, height: s.status.preview_height });
if (next.status.preview_bitrate_supported) settings.max_bitrate_bps = s.status.preview_bitrate_bps;
await next.native.request("video_settings", settings).catch(() => {});
}
if (s.audioEnabled) {
await next.native.request("audio", { enabled: true });
next.audioEnabled = true;
}
next.recoveryEnabled = true;
next.recoveryAttempts = s.recoveryAttempts;
next.recoveryAttempt = attempt;
next.recoveredFrom = s.id;
for (const view of views) if (this.views.get(view.id) === view && view.session === s.id) {
view.session = next.id;
view.disconnected = false;
view.connectionError = undefined;
}
for (const [old, replacement] of this.replacements) if (replacement === s.id) this.replacements.set(old, next.id);
if (this.replacements.size >= 128) this.replacements.delete(this.replacements.keys().next().value!);
this.replacements.set(s.id, next.id);
this.active.delete(s.id);
this.monitor(next);
return;
} catch (error) {
s.status = { ...s.status, state: "failed", control: false, recovery_error: failure(error) };
if (signal.aborted || this.closing || !s.recoveryEnabled ||
["AUTH_REQUIRED", "AUTH_FAILED", "DEVICE_ALIAS_CONFLICT", "CANCELLED"].includes(failure(error).code) ||
failure(error).reason === "tls_certificate_error") break;
}
}
s.recoveryEnabled = false;
s.status = { ...s.status, state: signal.aborted ? "closed" : "failed", control: false, recovery_exhausted: !signal.aborted };
}
async setRecovery(sessionId: string, enabled: boolean, attempts = 3): Promise<Data> {
const s = this.session(sessionId);
if (enabled) {
if (s.closing) throw new Failure("SESSION_CLOSED");
const device = await this.store.device(s.device.alias);
if (device.device_id !== s.device.device_id || device.server !== s.device.server || device.port !== s.device.port)
throw new Failure("DEVICE_ALIAS_CONFLICT");
if (!device.authorization || (s.mode === "control" && device.authorization !== "control"))
throw new Failure("AUTH_REQUIRED");
await this.store.password(device);
if (s.closing || this.active.get(s.id) !== s) throw new Failure("SESSION_CLOSED");
}
s.recoveryEnabled = enabled;
s.recoveryAttempts = attempts;
if (!enabled) s.recoveryAbort?.abort();
this.monitor(s);
return { session_id: s.id, reconnect_enabled: enabled, max_attempts: attempts };
}
async waitForSession(sessionId: string, waitMs: number, signal?: AbortSignal): Promise<Data> {
const deadline = Date.now() + waitMs;
while (true) {
const result = await this.status(sessionId, undefined, signal);
if (!["connecting", "reconnecting"].includes(result.state)) return result;
if (Date.now() >= deadline) return { ...result, wait_timed_out: true };
await delay(Math.min(200, deadline - Date.now()), undefined, { signal });
}
}
async waitForConnection(viewId: string, waitMs: number, signal?: AbortSignal): Promise<Data> {
const deadline = Date.now() + waitMs;
let result = this.describe(this.view(viewId));
const timedOut = () => ({
...result,
wait_timed_out: true,
next_step: "Still waiting for the user to connect in the panel. Call crossdesk_get_session again with the same view_id and wait_ms: 25000 while the remote task is pending. Do not ask the user to type a connection confirmation.",
});
const timeout = new AbortController();
const timer = waitMs > 0 ? setTimeout(() => timeout.abort(), waitMs) : undefined;
const waitSignal = signal ? AbortSignal.any([signal, timeout.signal]) : timeout.signal;
try {
for (;;) {
if (this.closing || signal?.aborted) throw new Failure("CANCELLED");
const view = this.view(viewId);
view.lastSeen = Date.now();
result = await this.status(undefined, viewId, waitSignal);
if (waitMs === 0 || view.panel === "off" ||
["ready", "closed", "failed", "disconnected"].includes(result.state))
return result;
const remaining = deadline - Date.now();
if (remaining <= 0) return timedOut();
await delay(Math.min(250, remaining), undefined, { signal: waitSignal });
}
} catch (error) {
if (this.closing || signal?.aborted) throw new Failure("CANCELLED");
if (timeout.signal.aborted) return timedOut();
throw error;
} finally {
clearTimeout(timer);
}
}
async connectSaved(
name: string,
mode: string,
display: number,
signal?: AbortSignal,
initialControl = mode === "control",
): Promise<Data> {
const d = await this.store.device(name);
if (
!d.authorization ||
(mode === "control" && d.authorization !== "control")
)
throw new Failure("AUTH_REQUIRED");
return this.connect(
d,
await this.waitStatus(this.store.password(d), signal),
mode,
display,
undefined,
signal,
initialControl,
);
}
async connectUI(v: View, args: unknown, signal?: AbortSignal): Promise<Data> {
const p = parse(uiConnect, args);
if (!!p.device_alias === !!p.device_id)
throw new Failure("SELECT_ONE_DEVICE");
const config = await this.store.config();
const d: Device = p.device_alias
? { ...(await this.store.device(p.device_alias)) }
: {
alias: p.label ?? p.device_id!,
device_id: p.device_id!,
server: p.server ?? config.server,
port: p.port ?? config.port,
};
if (p.device_alias && (p.server || p.port || p.label))
throw new Failure("INVALID_ARGUMENT");
if ((p.remember_password || p.remember_unlock_password || p.persistent_authorization) && !p.save_device)
throw new Failure("SAVE_DEVICE_REQUIRED");
if (p.remember_unlock_password && !p.unlock_password)
throw new Failure("UNLOCK_CREDENTIAL_REQUIRED");
if (p.unlock_password && !p.remember_unlock_password)
throw new Failure("SAVE_UNLOCK_REQUIRED");
let password = p.password ?? (await this.store.password(d));
p.password = undefined;
// Persist only after successful authentication, avoiding poisoned saved passwords.
const result = await this.connect(
d,
password,
p.mode,
p.display_index,
v,
signal,
);
if (p.save_device) {
d.authorization = p.persistent_authorization ? p.mode : undefined;
delete d.unlock_credential;
delete d.unlock_environment_variable;
try {
await this.store.save(d, p.remember_password ? password : undefined, p.unlock_password);
} catch (error) {
result.persistence_error =
error instanceof Failure ? error.code : "CONFIG_WRITE_FAILED";
}
}
password = "";
p.unlock_password = undefined;
return result;
}
private async connect(
d: Device,
password: string,
mode: string,
display: number,
view?: View,
signal?: AbortSignal,
initialControl = mode === "control",
): Promise<Data> {
if (this.closing) throw new Failure("SESSION_CLOSED");
const established = [...this.active.values()].filter((s) => !s.connecting && !s.closing).length;
if (established + this.connecting.size >= 8) throw new Failure("SESSION_LIMIT");
const controller = new AbortController();
this.connecting.add(controller);
signal = signal ? AbortSignal.any([signal, controller.signal]) : controller.signal;
let complete!: () => void;
const done = new Promise<void>((resolve) => { complete = resolve; });
this.connectionDone.add(done);
let release: (() => Promise<void>) | undefined;
let native: NativeClient | undefined;
let session: Session | undefined;
try {
for (const s of this.active.values()) {
if (s.closing && !s.recovering) {
this.active.delete(s.id);
for (const v of this.views.values()) if (v.session === s.id) {
v.session = undefined;
v.disconnected = true;
}
}
}
if (view) {
await this.revokeVideo(view);
view.disconnected = false;
view.connectionError = undefined;
}
release = await this.store.lock(
`device:${d.server}:${d.port}:${d.device_id}`,
);
this.binary ??= nativeBinary(this.nativePath).catch((e) => {
this.binary = undefined;
throw e;
});
const binary = await this.binary;
if (this.closing || signal?.aborted) throw new Failure("CANCELLED");
const sessionId = token();
const logs = join(this.store.directory, "logs", sessionId);
await mkdir(logs, { recursive: true, mode: 0o700 });
native = new NativeClient(
binary,
d.server,
d.port,
logs,
this.videoOptions ?? (await this.store.config()).video,
);
const s: Session = session = {
id: sessionId,
device: d,
native,
status: {},
release,
connecting: true,
connection: controller,
mode,
desiredControl: initialControl,
recoveryEnabled: false,
recoveryAttempts: 3,
recoveryAttempt: 0,
unlocks: new Map(),
};
this.current = s;
this.active.set(s.id, s);
if (view) {
view.session = s.id;
view.alias = d.alias;
view.disconnected = false;
}
s.status = await native.connect(
d.device_id,
password,
mode,
display,
signal,
);
password = "";
if (mode === "control" && !initialControl)
await native.request("control", { enabled: false }, 5000, signal);
const deadline = Date.now() + 35000;
while (Date.now() < deadline) {
if (this.closing || signal?.aborted) throw new Failure("CANCELLED");
s.status = await native.request("status", {}, 5000, signal);
if (this.active.get(s.id) !== s || signal.aborted) throw new Failure("CANCELLED");
if (s.status.state === "ready" && typeof s.status.frame_age_ms === "number" && s.status.frame_age_ms <= 2000) {
s.connecting = false;
return { ...s.status, session_id: s.id, device_alias: s.device.alias };
}
if (["failed", "closed", "disconnected"].includes(s.status.state)) {
const code = connectionErrors.get(s.status.error);
throw new Failure(code ?? "CONNECTION_FAILED", false, false,
code ? s.status.error : undefined);
}
await delay(100, undefined, { signal });
}
throw new Failure("CONNECTION_TIMEOUT");
} catch (error) {
if (session) { await this.dispose(session); this.active.delete(session.id); }
else {
await native?.close();
await release?.();
}
if (this.current?.native === native) this.current = [...this.active.values()].find((s) => !s.closing);
const reason = signal.aborted ? new Failure("CANCELLED") : error;
if (view && (!view.session || view.session === session?.id)) {
view.session = undefined;
if (!view.disconnected) view.connectionError = failure(reason);
}
throw reason;
} finally {
password = "";
this.connecting.delete(controller);
this.connectionDone.delete(done);
complete();
}
}
async screenshot(p: Data, signal?: AbortSignal): Promise<Data> {
const { session_id, ...args } = p;
const s = this.session(session_id);
if (s.displayBatch) throw new Failure("BUSY");
if (args.after_frame_id !== undefined && args.generation === undefined)
throw new Failure("GENERATION_REQUIRED");
const result = await s.native.request(
"screenshot",
args,
(args.timeout_ms ?? 2000) + 3000,
signal,
);
return this.publicObservation(s, result);
}
private publicObservation(s: Session, observation: Data): Data {
return { ...observation, metadata: { ...observation.metadata,
observation_id: `${s.id}_${observation.metadata.observation_id}` } };
}
private nativeObservation(s: Session, observation: string): string {
const prefix = `${s.id}_`;
if (!observation.startsWith(prefix)) throw new Failure("STALE_OBSERVATION");
return observation.slice(prefix.length);
}
async action(p: Data, signal?: AbortSignal): Promise<Data> {
const { session_id, observe_after = true, ...args } = p;
const s = this.session(session_id);
if (s.displayBatch) throw new Failure("BUSY");
this.checkUnlock(s);
if (s.unlock && this.unlockPending(s.unlock)) throw new Failure("UNLOCK_PENDING");
args.observation_id = this.nativeObservation(s, args.observation_id);
const result = await s.native.request("action", args, (args.action.duration_ms ?? 0) + 5000, signal);
if (observe_after && result.status === "submitted") {
try {
result.observation = await this.screenshot(
{
session_id,
generation: result.generation,
after_frame_id: result.after_frame_id,
timeout_ms: 2000,
},
signal,
);
} catch (error) {
result.observation_error =
error instanceof Failure ? error.code : "OBSERVATION_FAILED";
}
}
return result;
}
async waitFor(p: Data, signal?: AbortSignal): Promise<Data> {
const { session_id, ...args } = p;
const s = this.session(session_id);
args.observation_id = this.nativeObservation(s, args.observation_id);
const result = await s.native.request(
"wait_frame", args, (args.timeout_ms ?? 5000) + 3000, signal);
if (result.matched) result.observation = await this.screenshot({ session_id }, signal);
return result;
}
async captureDisplays(p: Data, signal?: AbortSignal): Promise<Data> {
const s = this.session(p.session_id);
if (s.displayBatch) throw new Failure("BUSY");
const status = await this.status(s.id, undefined, signal);
if (s.displayBatch) throw new Failure("BUSY");
if (s.unlock && this.unlockPending(s.unlock)) throw new Failure("UNLOCK_PENDING");
if (status.held_key_codes?.length || status.held_mouse_flags?.length) throw new Failure("INPUT_HELD");
const indices: number[] = p.display_indices ?? status.displays.map((display: Data) => display.index);
if (!indices.length || indices.length > 8) throw new Failure("DISPLAY_LIMIT");
if (indices.some((index) => !status.displays.some((display: Data) => display.index === index)))
throw new Failure("DISPLAY_UNAVAILABLE");
const native = s.native;
const created = Date.now();
if (!["ready", "waiting_frame"].includes(status.state)) throw new Failure("NOT_READY");
s.displayBatch = true;
const observations: Data[] = [];
let error: unknown;
try {
for (const index of indices) {
await native.request("display", { index }, 5000, signal);
const observation = await native.request("screenshot", { max_edge: p.max_edge ?? 1024 }, 5000, signal);
observations.push({ ...this.publicObservation(s, observation), actionable: false });
}
} catch (failure) { error = failure; }
finally {
try {
await native.request("display", { index: status.display });
await native.request("screenshot", { max_edge: 64 });
} catch (restoreError) { error ??= restoreError; }
s.displayBatch = false;
}
if (error) throw error;
for (const [key, capture] of this.captures) if (Date.now() - capture.created > 30000) this.captures.delete(key);
if (this.captures.size >= 32) this.captures.delete(this.captures.keys().next().value!);
const capture_id = token();
this.captures.set(capture_id, { session: s.id, native, created, displays: observations.map((o) => o.metadata) });
return { session_id: s.id, capture_id, observations, restored_display: status.display };
}
async dragAcross(p: Data, signal?: AbortSignal): Promise<Data> {
const key = `${p.session_id}:${p.operation_id}`;
const args = JSON.stringify({ capture_id: p.capture_id, path: p.path, button: p.button ?? "left", duration_ms: p.duration_ms ?? 500 });
const previous = this.drags.get(key);
if (previous) {
if (previous.arguments !== args) throw new Failure("OPERATION_ID_CONFLICT");
return previous.result;
}
const s = this.session(p.session_id);
const capture = this.captures.get(p.capture_id);
if (!capture || capture.session !== s.id || capture.native !== s.native || Date.now() - capture.created > 30000)
throw new Failure("STALE_OBSERVATION");
const sources = new Map(capture.displays.map((display) => [display.display_index, display]));
for (const point of p.path) {
const source = sources.get(point.display_index);
if (!source || point.x < 0 || point.y < 0 || point.x >= source.image_size.width || point.y >= source.image_size.height)
throw new Failure("INVALID_COORDINATE");
}
if (s.displayBatch) throw new Failure("BUSY");
if (s.unlock && this.unlockPending(s.unlock)) throw new Failure("UNLOCK_PENDING");
s.displayBatch = true;
const native = s.native;
const record = { arguments: args, result: { operation_id: p.operation_id, status: "in_progress" } as Data };
if (this.drags.size >= 128) this.drags.delete(this.drags.keys().next().value!);
this.drags.set(key, record);
let submitted = false;
const deadline = Date.now() + 30000;
try {
const status = await native.request("status", {}, 5000, signal);
if (!status.control || status.human_paused) throw new Failure("CONTROL_DENIED");
if (status.held_key_codes?.length || status.held_mouse_flags?.length) throw new Failure("INPUT_HELD");
let display = -1, observation: Data = {};
const submit = async (type: string, point: Data) => {
if (Date.now() > deadline) throw new Failure("ACTION_TIMEOUT", submitted);
if (point.display_index !== display) {
await native.request("display", { index: point.display_index, preserve_input: submitted }, 5000, signal);
observation = await native.request("screenshot", { max_edge: 1024 }, 5000, signal);
display = point.display_index;
}
const source = sources.get(display)!;
const image = observation.metadata.image_size;
const action = { type, x: point.x * image.width / source.image_size.width,
y: point.y * image.height / source.image_size.height, ...(type === "move" ? {} : { button: p.button ?? "left" }) };
submitted = true;
const result = await native.request("action", { operation_id: token(),
observation_id: observation.metadata.observation_id, action }, 5000, signal);
if (result.status !== "submitted") throw new Failure("ACTION_OUTCOME_UNKNOWN", true);
};
await submit("mouse_down", p.path[0]);
for (const point of p.path.slice(1)) {
await delay((p.duration_ms ?? 500) / (p.path.length - 1), undefined, { signal });
await submit("move", point);
}
await submit("mouse_up", p.path.at(-1));
record.result = { operation_id: p.operation_id, status: "submitted", side_effect_possible: true,
observation: this.publicObservation(s, await native.request("screenshot", {}, 5000, signal)) };
} catch (error) {
record.result = { operation_id: p.operation_id, status: submitted ? "outcome_unknown" : "failed_before_submit",
error: failure(error), side_effect_possible: submitted };
} finally {
await native.request("release_inputs").catch(() => {});
s.displayBatch = false;
}
return record.result;
}
async audio(p: Data, capture: boolean, signal?: AbortSignal): Promise<Data> {
const { session_id, ...args } = p;
const s = this.session(session_id);
const result = await s.native.request(capture ? "audio_capture" : "audio", args,
capture ? (args.duration_ms ?? 3000) + 5000 : 5000, signal);
if (!capture) { s.status = result; s.audioEnabled = args.enabled; }
return result;
}
async videoSettings(p: Data, signal?: AbortSignal): Promise<Data> {
const { session_id, ...args } = p;
const s = this.session(session_id);
if (s.displayBatch) throw new Failure("BUSY");
if (s.unlock && this.unlockPending(s.unlock)) throw new Failure("UNLOCK_PENDING");
const requested = await s.native.request("video_settings", args, 5000, signal);
const deadline = Date.now() + 2500;
do {
const status = await this.status(s.id, undefined, signal);
if (status.preview_ack === requested.preview_request) return status;
await delay(100, undefined, { signal });
} while (Date.now() < deadline);
return { ...requested, wait_timed_out: true };
}
async clipboard(p: Data, signal?: AbortSignal): Promise<Data> {
const { session_id, ...args } = p;
const s = this.session(session_id);
this.checkUnlock(s);
if (s.unlock && this.unlockPending(s.unlock)) throw new Failure("UNLOCK_PENDING");
return s.native.request("clipboard", args, 7000, signal);
}
async files(p: Data, signal?: AbortSignal): Promise<Data> {
const { session_id, ...args } = p;
const s = this.session(session_id);
this.checkUnlock(s);
if (s.unlock && this.unlockPending(s.unlock) && !["status", "cancel"].includes(args.operation))
throw new Failure("UNLOCK_PENDING");
if (args.operation !== "list") return s.native.request("files", args, 7000, signal);
const requestId = args.request_id ?? (await s.native.request("files",
{ operation: "list", path: args.path ?? "" }, 7000, signal)).request_id;
try {
const deadline = Date.now() + 25000;
while (true) {
const result = await s.native.request("files", {
operation: "list_result", request_id: requestId, offset: args.offset ?? 0,
}, 7000, signal);
if (result.state !== "pending") {
if (result.next_offset == null)
await s.native.request("files", { operation: "list_cancel", request_id: requestId });
return result;
}
if (Date.now() >= deadline) return { request_id: requestId, state: "pending", wait_timed_out: true };
await delay(100, undefined, { signal });
}
} catch (error) {
await s.native.request("files", { operation: "list_cancel", request_id: requestId }).catch(() => {});
throw error;
}
}
async desktop(p: Data, signal?: AbortSignal): Promise<Data> {
const { session_id, ...args } = p;
const s = this.session(session_id);
if (args.action) { Object.assign(args, args.action); delete args.action; }
const write = !["capabilities", "windows", "tree"].includes(args.operation);
if (write && s.displayBatch) throw new Failure("BUSY");
if (write && s.unlock && this.unlockPending(s.unlock)) throw new Failure("UNLOCK_PENDING");
if (args.snapshot_id) {
const prefix = `${s.id}_`;
if (!args.snapshot_id.startsWith(prefix)) throw new Failure("STALE_UI_SNAPSHOT");
args.snapshot_id = args.snapshot_id.slice(prefix.length);
}
const result = await s.native.request("desktop", args, 7000, signal);
if (result.snapshot_id) result.snapshot_id = `${s.id}_${result.snapshot_id}`;
return result;
}
async waitUI(p: Data, signal?: AbortSignal): Promise<Data> {
const deadline = Date.now() + (p.timeout_ms ?? 5000);
const normalized = (value: unknown) => p.case_sensitive ? String(value ?? "") : String(value ?? "").toLocaleLowerCase();
let incomplete = false;
do {
const result = await this.desktop({ session_id: p.session_id, operation: p.scope ?? "tree",
...(p.pid !== undefined ? { pid: p.pid } : {}), max_nodes: 128, max_depth: 16 }, signal);
if (result.error) return result;
const matches = (result.elements ?? []).filter((element: Data) =>
(!p.text || normalized(`${element.name ?? ""} ${element.value ?? ""}`).includes(normalized(p.text))) &&
(!p.role || normalized(element.role) === normalized(p.role)));
incomplete = result.truncated === true;
if (p.condition === "absent" ? !matches.length && !incomplete : matches.length > 0)
return { ...result, matched: true, matches };
if (Date.now() >= deadline) break;
await delay(Math.min(250, deadline - Date.now()), undefined, { signal });
} while (Date.now() <= deadline);
return { matched: false, wait_timed_out: true, search_incomplete: incomplete };
}
async system(p: Data, signal?: AbortSignal): Promise<Data> {
const { session_id, ...args } = p;
const s = this.session(session_id);
if (args.command !== "status" && s.unlock && this.unlockPending(s.unlock))
throw new Failure("UNLOCK_PENDING");
return s.native.request("system", args, 7000, signal);
}
async releaseInputs(sessionId: string) {
const s = this.session(sessionId);
this.cancelUnlock(s);
return s.native.request("release_inputs");
}
async control(id: string, enabled: boolean) {
const s = this.session(id);
if (!enabled) { this.cancelUnlock(s); s.desiredControl = false; }
const result = await s.native.request("control", { enabled });
s.desiredControl = enabled;
return result;
}
async display(id: string, index: number) {
const s = this.session(id);
if (s.displayBatch) throw new Failure("BUSY");
this.checkUnlock(s);
if (s.unlock && this.unlockPending(s.unlock)) throw new Failure("UNLOCK_PENDING");
return s.native.request("display", { index });
}
async disconnect(id: string) {
id = this.replacements.get(id) ?? id;
const s = this.session(id);
s.recoveryEnabled = false;
s.recoveryAbort?.abort();
clearInterval(s.monitor);
if (s.connecting) s.connection.abort();
for (const v of this.views.values())
if (v.session === id) {
v.session = undefined;
v.disconnected = true;
v.connectionError = undefined;
}
await this.dispose(s);
await s.recovering?.catch(() => {});
this.active.delete(s.id);
if (this.current === s) this.current = [...this.active.values()].find((other) => !other.closing);
return { session_id: id, state: "closed" };
}
async viewerControl(v: View, operation: string) {
if (!v.session) throw new Failure("NOT_READY");
if (operation === "disconnect") return this.disconnect(v.session);
if (operation === "take_over" && !this.session(v.session).native.viewerInputSupported)
throw new Failure("NATIVE_INPUT_UNSUPPORTED");
if (operation === "pause" || operation === "take_over") {
const s = this.session(v.session);
this.cancelUnlock(s);
s.status.human_paused = true;
s.desiredControl = false;
s.recoveryAbort?.abort();
}
const s = this.session(v.session);
const result = await s.native.request("viewer_control", {
operation, owner: v.id,
});
if (operation === "resume") s.desiredControl = true;
s.status = result;
return result;
}
async viewerInput(view: View, args: Data, signal?: AbortSignal): Promise<Data> {
if (!view.session) throw new Failure("NOT_READY");
const s = this.session(view.session);
if (s.displayBatch) throw new Failure("BUSY");
return s.native.request("viewer_input", { ...args, owner: view.id }, 5000, signal);
}
private unlockPending(request: UnlockRequest) {
return ["preparing", "awaiting_password", "submitting", "verifying"].includes(request.state);
}
private unlockStatus(request: UnlockRequest): Data {
return {
request_id: request.id,
state: request.state,
...(request.verification ? { verification: request.verification } : {}),
expires_at: request.expires,
...(request.error ? { error: request.error } : {}),
...(request.result ? { result: request.result } : {}),
...(request.persistenceError ? { persistence_error: request.persistenceError } : {}),
...(request.saved ? { credential_saved: true } : {}),
next_step: request.state === "verifying"
? "The remote system is verifying unlock completion. Wait with crossdesk_get_session and unlock_request_id."
: this.unlockPending(request)
? "The user enters the system lock-screen password in the local panel. Never request it in chat. Wait with crossdesk_get_session and unlock_request_id."
: ["submitted", "verified"].includes(request.state)
? "Take a fresh screenshot to verify that the desktop is unlocked, then continue the pending task. If the password was rejected, request user entry with use_saved: false; never automatically replay an unlock."
: "Unlock stopped. Inspect the error and current screen before starting another request. Never automatically replay an uncertain submission.",
};
}
private cancelUnlock(s: Session, code = "CANCELLED") {
const request = s.unlock;
if (!request || !this.unlockPending(request)) return;
request.error = failure(new Failure(code, ["submitting", "verifying"].includes(request.state)));
request.state = "cancelled";
request.controller.abort();
}
private checkUnlock(s: Session) {
const request = s.unlock;
if (!request || !this.unlockPending(request)) return;
if (Date.now() > request.expires) this.cancelUnlock(s, "UNLOCK_EXPIRED");
else if (s.closing || !s.status.control || s.status.human_paused ||
(request.generation !== undefined && (request.generation !== s.status.generation ||
request.geometry !== s.status.geometry_epoch))) this.cancelUnlock(s, "CONTROL_DENIED");
}
async requestUnlock(p: Data, signal?: AbortSignal): Promise<Data> {
const s = this.session(p.session_id);
if (s.displayBatch) throw new Failure("BUSY");
if (!s.native.unlockSupported) throw new Failure("NATIVE_UNLOCK_UNSUPPORTED");
const previous = s.unlocks.get(p.operation_id);
if (previous) {
if (previous.observation !== p.observation_id || previous.useSaved !== p.use_saved)
throw new Failure("OPERATION_ID_CONFLICT");
this.checkUnlock(s);
return { session_id: s.id, unlock: this.unlockStatus(previous) };
}
this.checkUnlock(s);
if (s.unlock && this.unlockPending(s.unlock)) throw new Failure("UNLOCK_PENDING");
const request: UnlockRequest = {
id: p.operation_id, operation: token(), observation: p.observation_id, useSaved: p.use_saved,
state: "preparing", expires: Date.now() + 300000, controller: new AbortController(),
};
s.unlock = request;
s.unlocks.set(request.id, request);
if (s.unlocks.size > 128) s.unlocks.delete(s.unlocks.keys().next().value!);
try {
const prepared = await s.native.request("unlock_prepare", { observation_id: this.nativeObservation(s, p.observation_id) }, 5000, signal);
request.generation = prepared.generation;
request.geometry = prepared.geometry_epoch;
let password: string | undefined;
if (p.use_saved) {
try {
const d = await this.store.device(s.device.alias);
if (d.device_id !== s.device.device_id || d.server !== s.device.server || d.port !== s.device.port)
throw new Failure("DEVICE_ALIAS_CONFLICT");
password = await this.store.password(d, "unlock");
} catch (error) {
if (!["DEVICE_NOT_FOUND", "UNLOCK_CREDENTIAL_REQUIRED", "CREDENTIAL_STORE_UNAVAILABLE"].includes(failure(error).code)) throw error;
}
}
if (!this.unlockPending(request) || signal?.aborted) throw new Failure("CANCELLED");
request.state = "awaiting_password";
if (password !== undefined) {
try { await this.submitUnlock(s, request, password, false, signal); }
finally { password = undefined; }
} else if (this.policy === "off") throw new Failure("UNLOCK_CREDENTIAL_REQUIRED");
} catch (error) {
if (request.state !== "cancelled") {
request.state = "failed";
request.error = failure(error);
}
}
return { session_id: s.id, unlock: this.unlockStatus(request) };
}
private async submitUnlock(s: Session, request: UnlockRequest, password: string, remember: boolean, signal?: AbortSignal) {
parse(unlockPassword, password);
await this.status(s.id, undefined, signal);
if (s.unlock !== request || request.state !== "awaiting_password") throw new Failure("UNLOCK_EXPIRED");
request.state = "submitting";
const combined = signal ? AbortSignal.any([signal, request.controller.signal]) : request.controller.signal;
try {
const verify = s.status.capabilities?.system === true;
if (verify) {
const before = await this.system({ session_id: s.id, command: "status" }, combined);
if (combined.aborted) return;
if (before.lock_state === "unlocked") {
request.state = "verified";
request.verification = "unlocked";
request.result = { status: "already_unlocked" };
return;
}
}
// User input may take longer than an observation's lifetime. Re-observe the
// same display generation before sending the secret to its focused field.
const observation = await this.screenshot({ session_id: s.id }, combined);
const metadata = observation.metadata;
if (metadata.generation !== request.generation || metadata.geometry_epoch !== request.geometry)
throw new Failure("STALE_OBSERVATION");
const result = await s.native.request("unlock", {
operation_id: request.operation,
observation_id: this.nativeObservation(s, metadata.observation_id),
password,
}, 5000, combined);
request.result = result;
if (request.controller.signal.aborted) return;
request.state = result.status === "submitted" ? "submitted" : "failed";
if (verify && request.state === "submitted") {
request.state = "verifying";
const deadline = Date.now() + 8000;
do {
const state = await this.system({ session_id: s.id, command: "status" }, combined);
request.verification = ["locked", "unlocked"].includes(state.lock_state) ? state.lock_state : "unknown";
if (request.verification !== "locked") break;
await delay(250, undefined, { signal: combined });
} while (Date.now() < deadline);
if (request.controller.signal.aborted) return;
request.state = request.verification === "unlocked" ? "verified" : "submitted";
if (request.verification === "locked") request.error = failure(new Failure("UNLOCK_NOT_CONFIRMED", true));
}
if (remember && (request.state === "verified" || (!verify && request.state === "submitted"))) {
try { await this.store.saveUnlock(s.device, password); request.saved = true; }
catch (error) { request.persistenceError = failure(error).code; }
} else if (remember && request.state === "submitted") {
request.persistenceError = "UNLOCK_NOT_CONFIRMED";
}
} catch (error) {
if (!request.controller.signal.aborted) {
request.state = "failed";
request.error = failure(error);
if (request.result?.status === "submitted") request.error.side_effect_possible = true;
}
} finally { password = ""; }
}
async unlockUI(v: View, p: Data, signal?: AbortSignal): Promise<Data> {
const s = this.session(v.session!);
const request = s.unlock;
if (!request || request.id !== p.request_id) throw new Failure("UNLOCK_NOT_FOUND");
if (p.cancel) this.cancelUnlock(s);
else if (request.state === "awaiting_password") {
const password = p.password;
p.password = undefined;
await this.submitUnlock(s, request, password, p.remember_password, signal);
}
return { unlock: this.unlockStatus(request) };
}
async waitForUnlock(sessionId: string | undefined, viewId: string | undefined, requestId: string, waitMs: number, signal?: AbortSignal): Promise<Data> {
const deadline = Date.now() + waitMs;
while (true) {
const result = await this.status(sessionId, viewId, signal);
const s = this.active.get(result.session_id);
if (!s || ["closed", "failed"].includes(result.state)) return result;
const request = s.unlocks.get(requestId);
if (!request) throw new Failure("UNLOCK_NOT_FOUND");
result.unlock = this.unlockStatus(request);
if (!this.unlockPending(request)) return result;
if (Date.now() >= deadline) return { ...result, wait_timed_out: true };
await delay(Math.min(200, deadline - Date.now()), undefined, { signal });
}
}
async terminalFrame(v: View, args: Data, signal?: AbortSignal) {
if (v.panel !== "terminal" || !v.session)
throw new Failure("NOT_READY");
const s = this.session(v.session);
return s.native.request("terminal_frame", args, 5000, signal);
}
async openVideo(v: View, origin: string): Promise<Data> {
if (!v.session || this.views.get(v.id) !== v) throw new Failure("NOT_READY");
const s = this.session(v.session);
if (this.closing || s.connecting || ["closed", "failed"].includes(s.status.state))
throw new Failure("NOT_READY");
const generation = v.generation, session = v.session;
const ticket = token();
const result = await s.native.request("video_open", {
ticket, view: v.id, view_generation: generation, session, origin, apps: v.panel === "apps",
});
if (this.active.get(s.id) !== s || this.views.get(v.id) !== v ||
v.generation !== generation || v.session !== session) {
await s.native.request("video_revoke", { view: v.id }).catch(() => {});
throw new Failure("SESSION_CHANGED");
}
if (result.transport !== "loopback-websocket/1" ||
typeof result.url !== "string" || !/^ws:\/\/127\.0\.0\.1:\d+\/api\/video$/.test(result.url) ||
result.protocol !== (s.native.videoOptions.decode_path === "minirtc" ? "nv12/1" : "h264-annexb/1") ||
result.decoder !== s.native.videoOptions.decoder)
throw new Failure("NATIVE_PROTOCOL_MISMATCH");
return { ...result, ticket };
}
async listSessions(): Promise<Data> {
const sessions = await Promise.all([...this.active.keys()].map((id) => this.status(id)));
return { sessions };
}
async revokeVideo(view?: View) {
const targets = view ? [this.active.get(view.session ?? "")].filter((s): s is Session => !!s)
: [...this.active.values()];
await Promise.all(targets.map(async (s) => {
try { await s.native.request("video_revoke", view ? { view: view.id } : {}); }
catch { if (this.active.get(s.id) === s) await this.disconnect(s.id); }
}));
}
async close() {
this.closing = true;
for (const controller of this.connecting) controller.abort();
const pending = [...this.connectionDone];
const results = await Promise.allSettled([...this.active.keys()].map((id) => this.disconnect(id)));
await Promise.allSettled(pending);
this.views.clear();
const error = results.find((result) => result.status === "rejected");
if (error?.status === "rejected") throw error.reason;
}
}