add whatsapp and telegram support

This commit is contained in:
Arjun 2026-07-03 19:25:06 +05:30
parent 3ba94402d3
commit fc9a76e1cb
15 changed files with 2244 additions and 9 deletions

View file

@ -30,6 +30,7 @@
"@x/shared": "workspace:*",
"ai": "^5.0.133",
"awilix": "^12.0.5",
"baileys": "7.0.0-rc13",
"chokidar": "^4.0.3",
"cors": "^2.8.6",
"cron-parser": "^5.5.0",
@ -45,6 +46,7 @@
"papaparse": "^5.5.3",
"pdf-parse": "^2.4.5",
"posthog-node": "^4.18.0",
"qrcode": "^1.5.4",
"react": "^19.2.3",
"xlsx": "^0.18.5",
"yaml": "^2.8.2",
@ -56,6 +58,7 @@
"@types/node": "^25.0.3",
"@types/papaparse": "^5.5.2",
"@types/pdf-parse": "^1.1.5",
"@types/qrcode": "^1.5.6",
"vitest": "catalog:"
}
}

View file

@ -0,0 +1,396 @@
import type { SessionIndexEntry } from "@x/shared/dist/sessions.js";
import { reduceTurn, type TurnStreamEvent } from "@x/shared/dist/turns.js";
import { assistantText, lastAssistantText } from "../agents/headless.js";
import { TurnNotSettledError, type ISessions } from "../sessions/api.js";
import type { EmitterSessionBus } from "../sessions/bus.js";
// Transport-agnostic command layer: inbound texts from a messaging channel
// are parsed into commands (list / resume / new / stop / status) or forwarded
// into a regular chat session; the turn's final assistant text is sent back
// through the transport's reply callback. Turns run with autoPermission and
// show up live in the desktop UI like any other session.
const AGENT_ID = "copilot";
const TURN_TIMEOUT_MS = 30 * 60 * 1000;
const LIST_LIMIT = 10;
// Telegram caps messages at 4096 chars; WhatsApp is far higher. Long replies
// are chunked, then truncated — the desktop app has the full text.
const REPLY_CHUNK_SIZE = 3500;
const MAX_REPLY_CHUNKS = 3;
const HELP_TEXT = [
"🤖 Rowboat commands:",
"• list — recent chats",
"• resume N — continue chat N from the list",
"• new [message] — start a fresh chat",
"• status — current chat and what it's doing",
"• stop — cancel the running task",
"",
"Anything else is sent to the current chat.",
].join("\n");
export type ReplyFn = (text: string) => Promise<void>;
interface SenderState {
activeSessionId: string | null;
// sessionIds as last shown by `list` (1-based indexing for `resume N`).
lastList: string[];
pendingAsk: { turnId: string; toolCallId: string } | null;
busy: boolean;
}
type Settled =
| { kind: "completed"; text: string | null }
| { kind: "failed"; error: string }
| { kind: "cancelled" }
| { kind: "ask_human"; toolCallId: string; query: string; options?: string[] }
| { kind: "suspended" }
| { kind: "timeout" };
function settleOf(event: TurnStreamEvent): Settled | null {
switch (event.type) {
case "turn_completed":
return { kind: "completed", text: assistantText(event.output) };
case "turn_failed":
return { kind: "failed", error: event.error };
case "turn_cancelled":
return { kind: "cancelled" };
case "turn_suspended": {
const ask = event.pendingAsyncTools.find(
(t) => t.toolId === "builtin:ask-human" || t.toolName === "ask-human",
);
if (ask) {
const input = ask.input as { query?: unknown; options?: unknown } | null;
const query =
typeof input?.query === "string" && input.query
? input.query
: "The agent needs your input.";
const options = Array.isArray(input?.options)
? input.options.filter((o): o is string => typeof o === "string")
: undefined;
return { kind: "ask_human", toolCallId: ask.toolCallId, query, options };
}
// Other async tools settle on their own and the turn resumes;
// keep waiting. Pending permissions need the desktop.
if (event.pendingAsyncTools.length === 0 && event.pendingPermissions.length > 0) {
return { kind: "suspended" };
}
return null;
}
default:
return null;
}
}
function relativeTime(iso: string): string {
const then = Date.parse(iso);
if (!Number.isFinite(then)) return "";
const diffSec = Math.round((Date.now() - then) / 1000);
if (diffSec < 60) return "just now";
const diffMin = Math.round(diffSec / 60);
if (diffMin < 60) return `${diffMin}m ago`;
const diffHr = Math.round(diffMin / 60);
if (diffHr < 24) return `${diffHr}h ago`;
return `${Math.round(diffHr / 24)}d ago`;
}
function chunkReply(text: string): string[] {
if (text.length <= REPLY_CHUNK_SIZE) return [text];
const parts: string[] = [];
let rest = text;
while (rest.length > 0 && parts.length < MAX_REPLY_CHUNKS) {
parts.push(rest.slice(0, REPLY_CHUNK_SIZE));
rest = rest.slice(REPLY_CHUNK_SIZE);
}
if (rest.length > 0) {
parts[parts.length - 1] += "\n… (truncated — open Rowboat for the full reply)";
}
return parts;
}
export class ChannelBridge {
private senders = new Map<string, SenderState>();
constructor(
private readonly deps: {
sessions: ISessions;
sessionBus: EmitterSessionBus;
},
) {}
async handleInbound(senderKey: string, text: string, reply: ReplyFn): Promise<void> {
const trimmed = text.trim();
if (!trimmed) return;
const state = this.senderState(senderKey);
const lower = trimmed.toLowerCase();
try {
if (lower === "help" || lower === "?") {
await reply(HELP_TEXT);
return;
}
if (lower === "list" || lower === "chats") {
await reply(this.renderList(state));
return;
}
const resume = /^(?:resume|open)\s+(\d+)$/.exec(lower);
if (resume) {
await reply(this.resumeSession(state, Number(resume[1])));
return;
}
if (lower === "status") {
await reply(this.renderStatus(state));
return;
}
if (lower === "stop") {
await reply(await this.stopActive(state));
return;
}
if (lower === "new") {
state.activeSessionId = null;
state.pendingAsk = null;
await reply("🆕 Fresh chat — send your first message.");
return;
}
const newWithText = /^new\s+([\s\S]+)$/i.exec(trimmed);
if (newWithText) {
state.activeSessionId = null;
state.pendingAsk = null;
await this.runMessage(state, newWithText[1].trim(), reply);
return;
}
await this.runMessage(state, trimmed, reply);
} catch (error) {
const message = error instanceof Error ? error.message : String(error);
await reply(`${message}`).catch(() => undefined);
}
}
private senderState(senderKey: string): SenderState {
let state = this.senders.get(senderKey);
if (!state) {
state = { activeSessionId: null, lastList: [], pendingAsk: null, busy: false };
this.senders.set(senderKey, state);
}
return state;
}
private sortedSessions(): SessionIndexEntry[] {
return this.deps.sessions
.listSessions()
.filter((e) => !e.error)
.sort((a, b) => b.updatedAt.localeCompare(a.updatedAt));
}
private renderList(state: SenderState): string {
const entries = this.sortedSessions().slice(0, LIST_LIMIT);
if (entries.length === 0) {
return "No chats yet — just send a message to start one.";
}
state.lastList = entries.map((e) => e.sessionId);
const lines = entries.map((e, i) => {
const marker =
e.latestTurnStatus === "suspended" ? " ⚠️" :
e.latestTurnStatus === "idle" ? " ⏳" : "";
const active = e.sessionId === state.activeSessionId ? " ← current" : "";
return `${i + 1}. ${e.title ?? "Untitled"}${marker} (${relativeTime(e.updatedAt)})${active}`;
});
return [
"Recent chats:",
...lines,
"",
`Reply "resume N" to continue one.`,
].join("\n");
}
private resumeSession(state: SenderState, index: number): string {
if (state.lastList.length === 0) {
state.lastList = this.sortedSessions()
.slice(0, LIST_LIMIT)
.map((e) => e.sessionId);
}
const sessionId = state.lastList[index - 1];
if (!sessionId) {
return `No chat #${index} — send "list" to see recent chats.`;
}
state.activeSessionId = sessionId;
state.pendingAsk = null;
const entry = this.sortedSessions().find((e) => e.sessionId === sessionId);
return `▶️ Resumed "${entry?.title ?? "Untitled"}" — send a message to continue.`;
}
private renderStatus(state: SenderState): string {
if (!state.activeSessionId) {
return "No current chat — your next message starts a new one.";
}
const entry = this.sortedSessions().find(
(e) => e.sessionId === state.activeSessionId,
);
if (!entry) return "Current chat no longer exists — send a message to start fresh.";
const status = state.busy
? "working"
: entry.latestTurnStatus === "suspended"
? "waiting on input"
: entry.latestTurnStatus;
return `Current chat: "${entry.title ?? "Untitled"}" — ${status}.`;
}
private async stopActive(state: SenderState): Promise<string> {
state.pendingAsk = null;
if (!state.activeSessionId) return "Nothing to stop.";
const entry = this.deps.sessions
.listSessions()
.find((e) => e.sessionId === state.activeSessionId);
if (!entry?.latestTurnId) return "Nothing to stop.";
if (
entry.latestTurnStatus === "completed" ||
entry.latestTurnStatus === "failed" ||
entry.latestTurnStatus === "cancelled"
) {
return "Nothing running in the current chat.";
}
await this.deps.sessions.stopTurn(entry.latestTurnId, "stopped from mobile channel");
return "🛑 Stop requested.";
}
private async runMessage(state: SenderState, text: string, reply: ReplyFn): Promise<void> {
if (state.busy) {
await reply('⏳ Still working on the previous message — send "stop" to cancel it.');
return;
}
state.busy = true;
// Subscribe before advancing so no settle event can slip past.
const watcher = this.watchBus();
try {
let turnId: string;
if (state.pendingAsk) {
const ask = state.pendingAsk;
state.pendingAsk = null;
turnId = ask.turnId;
await reply("⏳ Working on it…");
await this.deps.sessions.respondToAskHuman(ask.turnId, ask.toolCallId, text);
} else {
if (!state.activeSessionId) {
state.activeSessionId = await this.deps.sessions.createSession();
}
await reply("⏳ Working on it…");
const sent = await this.deps.sessions.sendMessage(
state.activeSessionId,
{ role: "user", content: text },
{ agent: { agentId: AGENT_ID }, autoPermission: true },
);
turnId = sent.turnId;
}
const settled = await watcher.waitFor(turnId, TURN_TIMEOUT_MS);
await this.deliverSettled(state, turnId, settled, reply);
} catch (error) {
if (error instanceof TurnNotSettledError) {
await reply(
'⏳ That chat is still working on something — send "stop" to cancel it, or "new" to start a fresh chat.',
);
return;
}
throw error;
} finally {
watcher.dispose();
state.busy = false;
}
}
private async deliverSettled(
state: SenderState,
turnId: string,
settled: Settled,
reply: ReplyFn,
): Promise<void> {
switch (settled.kind) {
case "completed": {
let text = settled.text;
if (!text) {
// Rare: final message had no text parts; recover the last
// assistant text from the persisted turn.
try {
const turn = await this.deps.sessions.getTurn(turnId);
text = lastAssistantText(reduceTurn(turn.events));
} catch {
text = null;
}
}
for (const chunk of chunkReply(text ?? "✅ Done (no text reply).")) {
await reply(chunk);
}
return;
}
case "failed":
await reply(`❌ Task failed: ${settled.error}`);
return;
case "cancelled":
await reply("🛑 Stopped.");
return;
case "ask_human": {
state.pendingAsk = { turnId, toolCallId: settled.toolCallId };
const lines = [`${settled.query}`];
if (settled.options?.length) {
lines.push(...settled.options.map((o, i) => `${i + 1}. ${o}`));
}
lines.push("", "Reply with your answer.");
await reply(lines.join("\n"));
return;
}
case "suspended":
await reply(
"⚠️ The agent is waiting for a permission approval — open Rowboat on your desktop to continue.",
);
return;
case "timeout":
await reply(
"⏱️ Still running after 30 minutes — check the desktop app for progress.",
);
return;
}
}
// Buffers turn events from the moment of subscription so a settle that
// fires between sendMessage() returning and waitFor() attaching is never
// lost. One watcher per in-flight message; disposed in runMessage.
private watchBus() {
const buffered: Array<{ turnId: string; event: TurnStreamEvent }> = [];
let waiter: { turnId: string; resolve: (settled: Settled) => void } | null = null;
const unsubscribe = this.deps.sessionBus.subscribe((event) => {
if (event.kind !== "turn-event") return;
if (waiter) {
if (event.turnId !== waiter.turnId) return;
const settled = settleOf(event.event);
if (settled) waiter.resolve(settled);
return;
}
buffered.push({ turnId: event.turnId, event: event.event });
});
return {
waitFor: (turnId: string, timeoutMs: number): Promise<Settled> =>
new Promise<Settled>((resolve) => {
for (const b of buffered) {
if (b.turnId !== turnId) continue;
const settled = settleOf(b.event);
if (settled) {
resolve(settled);
return;
}
}
buffered.length = 0;
const timer = setTimeout(
() => resolve({ kind: "timeout" }),
timeoutMs,
);
waiter = {
turnId,
resolve: (settled) => {
clearTimeout(timer);
resolve(settled);
},
};
}),
dispose: unsubscribe,
};
}
}

View file

@ -0,0 +1,40 @@
import fs from 'node:fs/promises';
import path from 'node:path';
import type { z } from 'zod';
import { WorkDir } from '../config/config.js';
import { ChannelsConfig, DEFAULT_CHANNELS_CONFIG } from '@x/shared/dist/channels.js';
export interface IChannelsConfigRepo {
getConfig(): Promise<z.infer<typeof ChannelsConfig>>;
setConfig(config: z.infer<typeof ChannelsConfig>): Promise<void>;
}
export class FSChannelsConfigRepo implements IChannelsConfigRepo {
private readonly configPath = path.join(WorkDir, 'config', 'channels.json');
constructor() {
this.ensureConfigFile();
}
private async ensureConfigFile(): Promise<void> {
try {
await fs.access(this.configPath);
} catch {
await fs.writeFile(this.configPath, JSON.stringify(DEFAULT_CHANNELS_CONFIG, null, 2));
}
}
async getConfig(): Promise<z.infer<typeof ChannelsConfig>> {
try {
const content = await fs.readFile(this.configPath, 'utf8');
return ChannelsConfig.parse(JSON.parse(content));
} catch {
return DEFAULT_CHANNELS_CONFIG;
}
}
async setConfig(config: z.infer<typeof ChannelsConfig>): Promise<void> {
const validated = ChannelsConfig.parse(config);
await fs.writeFile(this.configPath, JSON.stringify(validated, null, 2));
}
}

View file

@ -0,0 +1,183 @@
import path from "node:path";
import fs from "node:fs/promises";
import QRCode from "qrcode";
import type { z } from "zod";
import type { ChannelsConfig, ChannelsStatus } from "@x/shared/dist/channels.js";
import container from "../di/container.js";
import { WorkDir } from "../config/config.js";
import type { ISessions } from "../sessions/api.js";
import type { EmitterSessionBus } from "../sessions/bus.js";
import { ChannelBridge } from "./bridge.js";
import type { IChannelsConfigRepo } from "./repo.js";
import { TelegramTransport } from "./transports/telegram.js";
import { WhatsAppTransport } from "./transports/whatsapp.js";
// Lifecycle owner for the mobile channels: reads config, runs the enabled
// transports against one shared ChannelBridge, and fans status out to the
// renderer (QR pairing, connection state). init() from main after
// sessions.initialize(); applyChannelsConfig() on every settings save.
type Config = z.infer<typeof ChannelsConfig>;
type Status = z.infer<typeof ChannelsStatus>;
const WHATSAPP_AUTH_DIR = path.join(WorkDir, "channels", "whatsapp-auth");
let bridge: ChannelBridge | null = null;
let whatsapp: WhatsAppTransport | null = null;
let telegram: TelegramTransport | null = null;
const status: Status = {
whatsapp: { state: "disabled" },
telegram: { state: "disabled" },
};
const statusListeners = new Set<(status: Status) => void>();
// Serializes apply/logout so a fast settings double-save can't interleave
// transport teardown and startup.
let lifecycle: Promise<void> = Promise.resolve();
function notifyStatus(): void {
for (const listener of statusListeners) {
try {
listener(structuredClone(status));
} catch {
// observers must never affect the channels
}
}
}
function setWhatsAppStatus(next: Status["whatsapp"]): void {
status.whatsapp = next;
notifyStatus();
}
function setTelegramStatus(next: Status["telegram"]): void {
status.telegram = next;
notifyStatus();
}
export function getChannelsStatus(): Status {
return structuredClone(status);
}
export function subscribeChannelsStatus(listener: (status: Status) => void): () => void {
statusListeners.add(listener);
return () => statusListeners.delete(listener);
}
function ensureBridge(): ChannelBridge {
if (!bridge) {
bridge = new ChannelBridge({
sessions: container.resolve<ISessions>("sessions"),
sessionBus: container.resolve<EmitterSessionBus>("sessionBus"),
});
}
return bridge;
}
async function stopWhatsApp(): Promise<void> {
if (!whatsapp) return;
await whatsapp.stop().catch(() => undefined);
whatsapp = null;
}
function stopTelegram(): void {
if (!telegram) return;
telegram.stop();
telegram = null;
}
function startWhatsApp(config: Config["whatsapp"]): void {
if (!config.enabled) {
setWhatsAppStatus({ state: "disabled" });
return;
}
const channelBridge = ensureBridge();
const transport = new WhatsAppTransport({
authDir: WHATSAPP_AUTH_DIR,
allowFrom: config.allowFrom,
onInbound: (senderKey, text, reply) => {
void channelBridge.handleInbound(senderKey, text, reply);
},
onStatus: (update) => {
if (update.state === "qr" && update.qr) {
// Render the pairing QR main-side so the renderer just shows
// an <img>; the raw pairing string never leaves core.
QRCode.toDataURL(update.qr, { margin: 1, width: 256 })
.then((qrDataUrl) => setWhatsAppStatus({ state: "qr", qrDataUrl }))
.catch(() => setWhatsAppStatus({ state: "error", error: "Failed to render pairing QR" }));
return;
}
setWhatsAppStatus({
state: update.state,
...(update.self ? { self: update.self } : {}),
...(update.error ? { error: update.error } : {}),
});
},
});
whatsapp = transport;
transport.start().catch((error) => {
setWhatsAppStatus({
state: "error",
error: error instanceof Error ? error.message : String(error),
});
});
}
function startTelegram(config: Config["telegram"]): void {
if (!config.enabled) {
setTelegramStatus({ state: "disabled" });
return;
}
if (!config.botToken) {
setTelegramStatus({ state: "error", error: "Bot token missing — create one with @BotFather" });
return;
}
const channelBridge = ensureBridge();
const transport = new TelegramTransport({
botToken: config.botToken,
allowFrom: config.allowFrom,
onInbound: (senderKey, text, reply) => {
void channelBridge.handleInbound(senderKey, text, reply);
},
onStatus: setTelegramStatus,
});
telegram = transport;
void transport.start();
}
export function applyChannelsConfig(config: Config): Promise<void> {
lifecycle = lifecycle.then(async () => {
await stopWhatsApp();
stopTelegram();
startWhatsApp(config.whatsapp);
startTelegram(config.telegram);
});
return lifecycle;
}
// Unlink the WhatsApp device and, if the channel is still enabled, restart it
// so a fresh pairing QR appears. Telegram is left untouched.
export function logoutWhatsApp(): Promise<void> {
lifecycle = lifecycle.then(async () => {
if (whatsapp) {
await whatsapp.logout().catch(() => undefined);
whatsapp = null;
} else {
await fs.rm(WHATSAPP_AUTH_DIR, { recursive: true, force: true });
}
const config = await container
.resolve<IChannelsConfigRepo>("channelsConfigRepo")
.getConfig();
startWhatsApp(config.whatsapp);
});
return lifecycle;
}
export async function init(): Promise<void> {
const config = await container
.resolve<IChannelsConfigRepo>("channelsConfigRepo")
.getConfig();
await applyChannelsConfig(config);
}

View file

@ -0,0 +1,133 @@
import type { z } from "zod";
import type { TelegramChannelStatus } from "@x/shared/dist/channels.js";
import type { ReplyFn } from "../bridge.js";
// Telegram Bot API transport. Deliberately dependency-free: the Bot API is
// plain HTTPS — getUpdates long polling (outbound connection, works behind
// NAT) plus sendMessage. The user supplies their own bot token (@BotFather).
const POLL_TIMEOUT_S = 50;
const RETRY_DELAY_MS = 5000;
type Status = z.infer<typeof TelegramChannelStatus>;
interface TelegramUpdate {
update_id: number;
message?: {
message_id: number;
text?: string;
chat: { id: number; type: string };
from?: { id: number; is_bot?: boolean };
};
}
export interface TelegramTransportOptions {
botToken: string;
allowFrom: string[];
onInbound: (senderKey: string, text: string, reply: ReplyFn) => void;
onStatus: (status: Status) => void;
}
export class TelegramTransport {
private abort: AbortController | null = null;
private stopped = false;
private offset = 0;
constructor(private readonly opts: TelegramTransportOptions) {}
async start(): Promise<void> {
this.stopped = false;
this.opts.onStatus({ state: "starting" });
void this.run();
}
stop(): void {
this.stopped = true;
this.abort?.abort();
this.opts.onStatus({ state: "disabled" });
}
private api(method: string): string {
return `https://api.telegram.org/bot${this.opts.botToken}/${method}`;
}
private async call(method: string, body?: unknown, signal?: AbortSignal): Promise<unknown> {
const res = await fetch(this.api(method), {
method: "POST",
headers: { "content-type": "application/json" },
body: body === undefined ? undefined : JSON.stringify(body),
...(signal ? { signal } : {}),
});
const payload = (await res.json()) as { ok: boolean; result?: unknown; description?: string };
if (!payload.ok) {
throw new Error(payload.description ?? `Telegram API error (${method})`);
}
return payload.result;
}
private async run(): Promise<void> {
// Validate the token up front so a typo surfaces in settings
// immediately instead of as a silent poll loop failure.
try {
const me = (await this.call("getMe")) as { username?: string };
if (this.stopped) return;
this.opts.onStatus({ state: "polling", botUsername: me.username });
} catch (error) {
if (this.stopped) return;
this.opts.onStatus({
state: "error",
error: error instanceof Error ? error.message : String(error),
});
return;
}
while (!this.stopped) {
this.abort = new AbortController();
try {
const updates = (await this.call(
"getUpdates",
{
timeout: POLL_TIMEOUT_S,
offset: this.offset,
allowed_updates: ["message"],
},
this.abort.signal,
)) as TelegramUpdate[];
for (const update of updates) {
this.offset = Math.max(this.offset, update.update_id + 1);
this.handleUpdate(update);
}
} catch (error) {
if (this.stopped) return;
this.opts.onStatus({
state: "error",
error: error instanceof Error ? error.message : String(error),
});
await new Promise((resolve) => setTimeout(resolve, RETRY_DELAY_MS));
if (this.stopped) return;
this.opts.onStatus({ state: "polling" });
}
}
}
private handleUpdate(update: TelegramUpdate): void {
const message = update.message;
if (!message?.text || message.from?.is_bot) return;
// DMs only: group chats would let any member drive the bridge.
if (message.chat.type !== "private") return;
const chatId = String(message.chat.id);
const reply: ReplyFn = (text) => this.send(chatId, text);
if (!this.opts.allowFrom.includes(chatId)) {
void this.send(
chatId,
`⛔ Not authorized. Your chat ID is ${chatId} — add it under Rowboat → Settings → Mobile to pair this chat.`,
).catch(() => undefined);
return;
}
this.opts.onInbound(`telegram:${chatId}`, message.text, reply);
}
private async send(chatId: string, text: string): Promise<void> {
await this.call("sendMessage", { chat_id: chatId, text });
}
}

View file

@ -0,0 +1,191 @@
import fs from "node:fs/promises";
import makeWASocket, {
DisconnectReason,
areJidsSameUser,
isJidGroup,
jidDecode,
useMultiFileAuthState,
} from "baileys";
import type { ReplyFn } from "../bridge.js";
// WhatsApp transport via Baileys: the app links to the user's own WhatsApp
// account as a linked device (QR pairing, same as WhatsApp Web) over an
// outbound WebSocket — no server, no port forwarding.
//
// Access model: the linked account's own self-chat ("message yourself") is
// always allowed; other senders must be explicitly allowlisted by phone
// number. Group chats are ignored entirely.
type WASocket = ReturnType<typeof makeWASocket>;
const RECONNECT_DELAY_MS = 3000;
// Marks bridge-sent messages. In the self-chat our own replies come back on
// messages.upsert like any other message; the marker (plus sent-id tracking)
// keeps the bridge from answering itself in a loop.
const REPLY_MARKER = "🤖 ";
export interface WhatsAppTransportStatus {
state: "starting" | "qr" | "connected" | "error" | "disabled";
qr?: string;
self?: string;
error?: string;
}
export interface WhatsAppTransportOptions {
authDir: string;
allowFrom: string[];
onInbound: (senderKey: string, text: string, reply: ReplyFn) => void;
onStatus: (status: WhatsAppTransportStatus) => void;
}
interface TextishMessage {
conversation?: unknown;
extendedTextMessage?: { text?: unknown };
ephemeralMessage?: { message?: TextishMessage };
}
interface InboundWAMessage {
key?: { remoteJid?: string | null; fromMe?: boolean | null; id?: string | null };
message?: unknown;
}
function messageText(message: unknown): string | null {
if (!message || typeof message !== "object") return null;
const m = message as TextishMessage;
const unwrapped = m.ephemeralMessage?.message ?? m;
const text: unknown = unwrapped.conversation ?? unwrapped.extendedTextMessage?.text;
return typeof text === "string" && text ? text : null;
}
export class WhatsAppTransport {
private sock: WASocket | null = null;
private stopped = false;
private sentIds = new Set<string>();
constructor(private readonly opts: WhatsAppTransportOptions) {}
async start(): Promise<void> {
this.stopped = false;
this.opts.onStatus({ state: "starting" });
await this.connect();
}
async stop(): Promise<void> {
this.stopped = true;
try {
this.sock?.end(undefined);
} catch {
// already closed
}
this.sock = null;
this.opts.onStatus({ state: "disabled" });
}
// Unlink this device: invalidates the pairing on the phone and clears
// local credentials so the next start shows a fresh QR.
async logout(): Promise<void> {
this.stopped = true;
try {
await this.sock?.logout();
} catch {
// best effort — clearing creds below is what actually unpairs us
}
this.sock = null;
await fs.rm(this.opts.authDir, { recursive: true, force: true });
this.opts.onStatus({ state: "disabled" });
}
private async connect(): Promise<void> {
if (this.stopped) return;
const { state, saveCreds } = await useMultiFileAuthState(this.opts.authDir);
const sock = makeWASocket({
auth: state,
syncFullHistory: false,
markOnlineOnConnect: false,
});
this.sock = sock;
sock.ev.on("creds.update", saveCreds);
sock.ev.on("connection.update", (update) => {
if (this.stopped) return;
if (update.qr) {
this.opts.onStatus({ state: "qr", qr: update.qr });
}
if (update.connection === "open") {
const self = jidDecode(sock.user?.id ?? "")?.user;
this.opts.onStatus({ state: "connected", ...(self ? { self } : {}) });
}
if (update.connection === "close") {
const statusCode = (update.lastDisconnect?.error as { output?: { statusCode?: number } } | undefined)
?.output?.statusCode;
if (statusCode === DisconnectReason.loggedOut) {
// Unlinked from the phone; stale creds would loop forever.
void fs.rm(this.opts.authDir, { recursive: true, force: true });
this.opts.onStatus({
state: "error",
error: "Logged out from the phone — toggle WhatsApp off and on to pair again.",
});
return;
}
setTimeout(() => {
this.connect().catch((error) => {
this.opts.onStatus({
state: "error",
error: error instanceof Error ? error.message : String(error),
});
});
}, RECONNECT_DELAY_MS);
}
});
sock.ev.on("messages.upsert", ({ messages, type }) => {
if (type !== "notify") return;
for (const msg of messages) {
this.handleMessage(sock, msg);
}
});
}
private handleMessage(sock: WASocket, msg: InboundWAMessage): void {
const jid: string | undefined = msg.key?.remoteJid ?? undefined;
if (!jid || isJidGroup(jid) || jid === "status@broadcast") return;
const messageId: string | undefined = msg.key?.id ?? undefined;
if (messageId && this.sentIds.has(messageId)) return;
const text = messageText(msg.message);
if (!text || text.startsWith(REPLY_MARKER)) return;
const selfIds = [sock.user?.id, (sock.user as { lid?: string } | undefined)?.lid].filter(
(v): v is string => Boolean(v),
);
const isSelfChat = selfIds.some((selfId) => areJidsSameUser(jid, selfId));
const senderNumber = jidDecode(jid)?.user ?? "";
// Self-chat is the owner by definition. Anyone else must be
// allowlisted — this bridge is remote control over the desktop agent.
if (!isSelfChat) {
if (msg.key?.fromMe) return;
if (!this.opts.allowFrom.includes(senderNumber)) return;
}
const reply: ReplyFn = (replyText) => this.send(jid, replyText);
this.opts.onInbound(`whatsapp:${senderNumber || jid}`, text, reply);
}
private async send(jid: string, text: string): Promise<void> {
const sock = this.sock;
if (!sock) throw new Error("WhatsApp is not connected");
const sent = await sock.sendMessage(jid, { text: `${REPLY_MARKER}${text}` });
const id = sent?.key?.id;
if (id) {
this.sentIds.add(id);
if (this.sentIds.size > 500) {
// Trim oldest; Set iteration order is insertion order.
for (const old of this.sentIds) {
this.sentIds.delete(old);
if (this.sentIds.size <= 250) break;
}
}
}
}
}

View file

@ -18,6 +18,7 @@ import { IAbortRegistry, InMemoryAbortRegistry } from "../runs/abort-registry.js
import { FSAgentScheduleRepo, IAgentScheduleRepo } from "../agent-schedule/repo.js";
import { FSAgentScheduleStateRepo, IAgentScheduleStateRepo } from "../agent-schedule/state-repo.js";
import { FSSlackConfigRepo, ISlackConfigRepo } from "../slack/repo.js";
import { FSChannelsConfigRepo, IChannelsConfigRepo } from "../channels/repo.js";
import { CodeModeManager } from "../code-mode/acp/manager.js";
import { CodePermissionRegistry } from "../code-mode/acp/permission-registry.js";
import { FSCodeProjectsRepo, ICodeProjectsRepo } from "../code-mode/projects/repo.js";
@ -90,6 +91,7 @@ container.register({
agentScheduleRepo: asClass<IAgentScheduleRepo>(FSAgentScheduleRepo).singleton(),
agentScheduleStateRepo: asClass<IAgentScheduleStateRepo>(FSAgentScheduleStateRepo).singleton(),
slackConfigRepo: asClass<ISlackConfigRepo>(FSSlackConfigRepo).singleton(),
channelsConfigRepo: asClass<IChannelsConfigRepo>(FSChannelsConfigRepo).singleton(),
// ACP code-mode engine: the manager holds a live agent connection per chat only
// around an active turn (torn down after a short idle grace; resumed via

View file

@ -0,0 +1,54 @@
import { z } from "zod";
// ---------------------------------------------------------------------------
// Mobile messaging channels (WhatsApp / Telegram bridges).
//
// Each channel links the user's OWN account/bot to the local app: WhatsApp
// connects as a linked device (QR pairing), Telegram long-polls the user's
// own bot token. There is no central relay — the desktop app must be running.
// ---------------------------------------------------------------------------
export const WhatsAppChannelConfig = z.object({
enabled: z.boolean(),
// Extra sender phone numbers (digits only, no '+') allowed to drive the
// bridge. The linked account itself (self-chat) is always allowed.
allowFrom: z.array(z.string()),
});
export const TelegramChannelConfig = z.object({
enabled: z.boolean(),
botToken: z.string(),
// Telegram chat ids (numeric strings) allowed to drive the bridge.
// Unknown senders are told their chat id so it can be added here.
allowFrom: z.array(z.string()),
});
export const ChannelsConfig = z.object({
whatsapp: WhatsAppChannelConfig,
telegram: TelegramChannelConfig,
});
export const DEFAULT_CHANNELS_CONFIG: z.infer<typeof ChannelsConfig> = {
whatsapp: { enabled: false, allowFrom: [] },
telegram: { enabled: false, botToken: "", allowFrom: [] },
};
export const WhatsAppChannelStatus = z.object({
state: z.enum(["disabled", "starting", "qr", "connected", "error"]),
// PNG data URL of the pairing QR (present while state === "qr").
qrDataUrl: z.string().optional(),
// Linked account phone number (digits) once connected.
self: z.string().optional(),
error: z.string().optional(),
});
export const TelegramChannelStatus = z.object({
state: z.enum(["disabled", "starting", "polling", "error"]),
botUsername: z.string().optional(),
error: z.string().optional(),
});
export const ChannelsStatus = z.object({
whatsapp: WhatsAppChannelStatus,
telegram: TelegramChannelStatus,
});

View file

@ -19,4 +19,5 @@ export * as browserControl from './browser-control.js';
export * as billing from './billing.js';
export * as notificationSettings from './notification-settings.js';
export * as codeSessions from './code-sessions.js';
export * as channels from './channels.js';
export { PrefixLogger };

View file

@ -24,6 +24,7 @@ import { EmailBlockSchema, GmailThreadSchema } from './blocks.js';
import { PermissionDecision, ApprovalPolicy, CodingAgent } from './code-mode.js';
import { NotificationSettingsSchema } from './notification-settings.js';
import { CodeProject, CodeSession, CodeSessionMode, CodeSessionStatus, GitRepoInfo, GitStatusFile, CodeAgentModelOptions } from './code-sessions.js';
import { ChannelsConfig, ChannelsStatus } from './channels.js';
// ============================================================================
// Runtime Validation Schemas (Single Source of Truth)
@ -950,6 +951,28 @@ const ipcSchemas = {
success: z.literal(true),
}),
},
// ── Mobile channels (WhatsApp / Telegram bridge) ─────────────
'channels:getConfig': {
req: z.null(),
res: ChannelsConfig,
},
'channels:setConfig': {
req: ChannelsConfig,
res: z.object({ success: z.literal(true) }),
},
'channels:getStatus': {
req: z.null(),
res: ChannelsStatus,
},
'channels:whatsappLogout': {
req: z.null(),
res: z.object({ success: z.literal(true) }),
},
// Push: main → renderer status updates (QR rotation, connect/disconnect).
'channels:status': {
req: ChannelsStatus,
res: z.null(),
},
'slack:getConfig': {
req: z.null(),
res: z.object({