@openclaw/whatsapp 2026.5.2-beta.2 → 2026.5.3-beta.1

package/src/inbound/extract.ts

@@ -1,488 +0,0 @@
-import type { proto } from "@whiskeysockets/baileys";
-import {
-  extractMessageContent,
-  getContentType,
-  normalizeMessageContent,
-} from "@whiskeysockets/baileys";
-import { formatLocationText, type NormalizedLocation } from "openclaw/plugin-sdk/channel-inbound";
-import { logVerbose } from "openclaw/plugin-sdk/runtime-env";
-import { resolveComparableIdentity, type WhatsAppReplyContext } from "../identity.js";
-import { jidToE164 } from "../text-runtime.js";
-import { parseVcard } from "../vcard.js";
-import type { WhatsAppStructuredContactContext } from "./types.js";
-
-const MESSAGE_WRAPPER_KEYS = [
-  "botInvokeMessage",
-  "ephemeralMessage",
-  "viewOnceMessage",
-  "viewOnceMessageV2",
-  "viewOnceMessageV2Extension",
-  "documentWithCaptionMessage",
-  "groupMentionedMessage",
-] as const;
-
-const MESSAGE_CONTENT_KEYS = [
-  "conversation",
-  "extendedTextMessage",
-  "imageMessage",
-  "videoMessage",
-  "audioMessage",
-  "documentMessage",
-  "stickerMessage",
-  "locationMessage",
-  "liveLocationMessage",
-  "contactMessage",
-  "contactsArrayMessage",
-  "buttonsResponseMessage",
-  "listResponseMessage",
-  "templateButtonReplyMessage",
-  "interactiveResponseMessage",
-  "buttonsMessage",
-  "listMessage",
-] as const;
-
-function fallbackNormalizeMessageContent(
-  message: proto.IMessage | undefined,
-): proto.IMessage | undefined {
-  let current = message as unknown;
-  while (current && typeof current === "object") {
-    let unwrapped = false;
-    for (const key of MESSAGE_WRAPPER_KEYS) {
-      const candidate = (current as Record<string, unknown>)[key];
-      if (
-        candidate &&
-        typeof candidate === "object" &&
-        "message" in (candidate as Record<string, unknown>) &&
-        (candidate as { message?: unknown }).message
-      ) {
-        current = (candidate as { message: unknown }).message;
-        unwrapped = true;
-        break;
-      }
-    }
-    if (!unwrapped) {
-      break;
-    }
-  }
-  return current as proto.IMessage | undefined;
-}
-
-function normalizeMessage(message: proto.IMessage | undefined): proto.IMessage | undefined {
-  if (typeof normalizeMessageContent === "function") {
-    return normalizeMessageContent(message);
-  }
-  return fallbackNormalizeMessageContent(message);
-}
-
-function fallbackGetContentType(
-  message: proto.IMessage | undefined,
-): keyof proto.IMessage | undefined {
-  const normalized = fallbackNormalizeMessageContent(message);
-  if (!normalized || typeof normalized !== "object") {
-    return undefined;
-  }
-  for (const key of MESSAGE_CONTENT_KEYS) {
-    if ((normalized as Record<string, unknown>)[key] != null) {
-      return key as keyof proto.IMessage;
-    }
-  }
-  return undefined;
-}
-
-function getMessageContentType(
-  message: proto.IMessage | undefined,
-): keyof proto.IMessage | undefined {
-  if (typeof getContentType === "function") {
-    return getContentType(message);
-  }
-  return fallbackGetContentType(message);
-}
-
-function extractMessage(message: proto.IMessage | undefined): proto.IMessage | undefined {
-  if (typeof extractMessageContent === "function") {
-    return extractMessageContent(message);
-  }
-  const normalized = fallbackNormalizeMessageContent(message);
-  const contentType = fallbackGetContentType(normalized);
-  if (!normalized || !contentType || contentType === "conversation") {
-    return normalized;
-  }
-  const candidate = (normalized as Record<string, unknown>)[contentType];
-  return candidate && typeof candidate === "object" ? (candidate as proto.IMessage) : normalized;
-}
-
-function getFutureProofInnerMessage(message: proto.IMessage): proto.IMessage | undefined {
-  const contentType = getMessageContentType(message);
-  const candidate = contentType ? (message as Record<string, unknown>)[contentType] : undefined;
-  if (
-    candidate &&
-    typeof candidate === "object" &&
-    "message" in candidate &&
-    (candidate as { message?: unknown }).message &&
-    typeof (candidate as { message: unknown }).message === "object"
-  ) {
-    const inner = normalizeMessage((candidate as { message: proto.IMessage }).message);
-    if (inner) {
-      const innerType = getMessageContentType(inner);
-      if (innerType && innerType !== contentType) {
-        return inner;
-      }
-    }
-  }
-  return undefined;
-}
-
-function buildMessageChain(message: proto.IMessage | undefined): proto.IMessage[] {
-  const chain: proto.IMessage[] = [];
-  let current = normalizeMessage(message);
-  while (current && chain.length < 4) {
-    chain.push(current);
-    current = getFutureProofInnerMessage(current);
-  }
-  return chain;
-}
-
-function unwrapMessage(message: proto.IMessage | undefined): proto.IMessage | undefined {
-  const chain = buildMessageChain(message);
-  return chain.at(-1);
-}
-
-function extractContextInfoFromMessage(message: proto.IMessage): proto.IContextInfo | undefined {
-  const contentType = getMessageContentType(message);
-  const candidate = contentType ? (message as Record<string, unknown>)[contentType] : undefined;
-  const contextInfo =
-    candidate && typeof candidate === "object" && "contextInfo" in candidate
-      ? (candidate as { contextInfo?: proto.IContextInfo }).contextInfo
-      : undefined;
-  if (contextInfo) {
-    return contextInfo;
-  }
-  const fallback =
-    message.extendedTextMessage?.contextInfo ??
-    message.imageMessage?.contextInfo ??
-    message.videoMessage?.contextInfo ??
-    message.documentMessage?.contextInfo ??
-    message.audioMessage?.contextInfo ??
-    message.stickerMessage?.contextInfo ??
-    message.buttonsResponseMessage?.contextInfo ??
-    message.listResponseMessage?.contextInfo ??
-    message.templateButtonReplyMessage?.contextInfo ??
-    message.interactiveResponseMessage?.contextInfo ??
-    message.buttonsMessage?.contextInfo ??
-    message.listMessage?.contextInfo;
-  if (fallback) {
-    return fallback;
-  }
-  for (const value of Object.values(message)) {
-    if (!value || typeof value !== "object") {
-      continue;
-    }
-    if ("contextInfo" in value) {
-      const candidateContext = (value as { contextInfo?: proto.IContextInfo }).contextInfo;
-      if (candidateContext) {
-        return candidateContext;
-      }
-    }
-    // FutureProofMessage wrapper: dig into .message to find contextInfo
-    if ("message" in value) {
-      const inner = (value as { message?: proto.IMessage }).message;
-      if (inner) {
-        const innerCtx = extractContextInfo(inner);
-        if (innerCtx) {
-          return innerCtx;
-        }
-      }
-    }
-  }
-  return undefined;
-}
-
-export function extractContextInfo(
-  message: proto.IMessage | undefined,
-): proto.IContextInfo | undefined {
-  for (const candidate of buildMessageChain(message)) {
-    const contextInfo = extractContextInfoFromMessage(candidate);
-    if (contextInfo) {
-      return contextInfo;
-    }
-  }
-  return undefined;
-}
-
-export function extractMentionedJids(rawMessage: proto.IMessage | undefined): string[] | undefined {
-  const message = unwrapMessage(rawMessage);
-  if (!message) {
-    return undefined;
-  }
-
-  const candidates: Array<string[] | null | undefined> = [
-    message.extendedTextMessage?.contextInfo?.mentionedJid,
-    message.imageMessage?.contextInfo?.mentionedJid,
-    message.videoMessage?.contextInfo?.mentionedJid,
-    message.documentMessage?.contextInfo?.mentionedJid,
-    message.audioMessage?.contextInfo?.mentionedJid,
-    message.stickerMessage?.contextInfo?.mentionedJid,
-    message.buttonsResponseMessage?.contextInfo?.mentionedJid,
-    message.listResponseMessage?.contextInfo?.mentionedJid,
-  ];
-
-  const flattened = candidates.flatMap((arr) => arr ?? []).filter(Boolean);
-  if (flattened.length === 0) {
-    return undefined;
-  }
-  return Array.from(new Set(flattened));
-}
-
-export function extractText(rawMessage: proto.IMessage | undefined): string | undefined {
-  const message = unwrapMessage(rawMessage);
-  if (!message) {
-    return undefined;
-  }
-  const extracted = extractMessage(message);
-  const candidates = [message, extracted && extracted !== message ? extracted : undefined];
-  for (const candidate of candidates) {
-    if (!candidate) {
-      continue;
-    }
-    if (typeof candidate.conversation === "string" && candidate.conversation.trim()) {
-      return candidate.conversation.trim();
-    }
-    const extended = candidate.extendedTextMessage?.text;
-    if (extended?.trim()) {
-      return extended.trim();
-    }
-    const caption =
-      candidate.imageMessage?.caption ??
-      candidate.videoMessage?.caption ??
-      candidate.documentMessage?.caption;
-    if (caption?.trim()) {
-      return caption.trim();
-    }
-  }
-  const contactPlaceholder =
-    extractContactPlaceholder(message) ??
-    (extracted && extracted !== message
-      ? extractContactPlaceholder(extracted as proto.IMessage | undefined)
-      : undefined);
-  if (contactPlaceholder) {
-    return contactPlaceholder;
-  }
-  return undefined;
-}
-
-export function extractMediaPlaceholder(
-  rawMessage: proto.IMessage | undefined,
-): string | undefined {
-  const message = unwrapMessage(rawMessage);
-  if (!message) {
-    return undefined;
-  }
-  if (message.imageMessage) {
-    return "<media:image>";
-  }
-  if (message.videoMessage) {
-    return "<media:video>";
-  }
-  if (message.audioMessage) {
-    return "<media:audio>";
-  }
-  if (message.documentMessage) {
-    return "<media:document>";
-  }
-  if (message.stickerMessage) {
-    return "<media:sticker>";
-  }
-  return undefined;
-}
-
-function extractContactPlaceholder(rawMessage: proto.IMessage | undefined): string | undefined {
-  const contactContext = extractContactContext(rawMessage);
-  if (!contactContext) {
-    return undefined;
-  }
-  if (contactContext.kind === "contact") {
-    return "<contact>";
-  }
-  const suffix = contactContext.total === 1 ? "contact" : "contacts";
-  return `<contacts: ${contactContext.total} ${suffix}>`;
-}
-
-export function extractContactContext(
-  rawMessage: proto.IMessage | undefined,
-): WhatsAppStructuredContactContext | undefined {
-  const message = unwrapMessage(rawMessage);
-  if (!message) {
-    return undefined;
-  }
-  const contact = message.contactMessage ?? undefined;
-  if (contact) {
-    const { name, phones } = describeContact({
-      displayName: contact.displayName,
-      vcard: contact.vcard,
-    });
-    return {
-      kind: "contact",
-      total: 1,
-      contacts: [{ name, phones }],
-    };
-  }
-  const contactsArray = message.contactsArrayMessage?.contacts ?? undefined;
-  if (!contactsArray || contactsArray.length === 0) {
-    return undefined;
-  }
-  return {
-    kind: "contacts",
-    total: contactsArray.length,
-    contacts: contactsArray.map((entry) =>
-      describeContact({ displayName: entry.displayName, vcard: entry.vcard }),
-    ),
-  };
-}
-
-function describeContact(input: { displayName?: string | null; vcard?: string | null }): {
-  name?: string;
-  phones: string[];
-} {
-  const displayName = (input.displayName ?? "").trim();
-  const parsed = parseVcard(input.vcard ?? undefined);
-  const name = displayName || parsed.name;
-  return { name, phones: parsed.phones };
-}
-
-export function extractLocationData(
-  rawMessage: proto.IMessage | undefined,
-): NormalizedLocation | null {
-  const message = unwrapMessage(rawMessage);
-  if (!message) {
-    return null;
-  }
-
-  const live = message.liveLocationMessage ?? undefined;
-  if (live) {
-    const latitudeRaw = live.degreesLatitude;
-    const longitudeRaw = live.degreesLongitude;
-    if (latitudeRaw != null && longitudeRaw != null) {
-      const latitude = latitudeRaw;
-      const longitude = longitudeRaw;
-      if (Number.isFinite(latitude) && Number.isFinite(longitude)) {
-        return {
-          latitude,
-          longitude,
-          accuracy: live.accuracyInMeters ?? undefined,
-          caption: live.caption ?? undefined,
-          source: "live",
-          isLive: true,
-        };
-      }
-    }
-  }
-
-  const location = message.locationMessage ?? undefined;
-  if (location) {
-    const latitudeRaw = location.degreesLatitude;
-    const longitudeRaw = location.degreesLongitude;
-    if (latitudeRaw != null && longitudeRaw != null) {
-      const latitude = latitudeRaw;
-      const longitude = longitudeRaw;
-      if (Number.isFinite(latitude) && Number.isFinite(longitude)) {
-        const isLive = Boolean(location.isLive);
-        return {
-          latitude,
-          longitude,
-          accuracy: location.accuracyInMeters ?? undefined,
-          name: location.name ?? undefined,
-          address: location.address ?? undefined,
-          caption: location.comment ?? undefined,
-          source: isLive ? "live" : location.name || location.address ? "place" : "pin",
-          isLive,
-        };
-      }
-    }
-  }
-
-  return null;
-}
-
-export function describeReplyContext(
-  rawMessage: proto.IMessage | undefined,
-): WhatsAppReplyContext | null {
-  const message = unwrapMessage(rawMessage);
-  if (!message) {
-    return null;
-  }
-  const contextInfo = extractContextInfo(message);
-  const quoted = normalizeMessage(contextInfo?.quotedMessage as proto.IMessage | undefined);
-  if (!quoted) {
-    return null;
-  }
-  const location = extractLocationData(quoted);
-  const locationText = location ? formatLocationText(location) : undefined;
-  const text = extractText(quoted);
-  let body: string | undefined = [text, locationText].filter(Boolean).join("\n").trim();
-  if (!body) {
-    body = extractMediaPlaceholder(quoted);
-  }
-  if (!body) {
-    const quotedType = quoted ? getMessageContentType(quoted) : undefined;
-    logVerbose(
-      `Quoted message missing extractable body${quotedType ? ` (type ${quotedType})` : ""}`,
-    );
-    return null;
-  }
-  const senderJid = contextInfo?.participant ?? undefined;
-  const sender = resolveComparableIdentity({
-    jid: senderJid,
-    label: senderJid ? (jidToE164(senderJid) ?? senderJid) : "unknown sender",
-  });
-  return {
-    id: contextInfo?.stanzaId || undefined,
-    body,
-    sender,
-  };
-}
-
-function hasInteractiveResponseContent(message: proto.IMessage | undefined): boolean {
-  if (!message) {
-    return false;
-  }
-  // Button/list/template/interactive selections that the existing four
-  // extractors do not cover. Treat any presence of these keys as user
-  // content — Baileys never delivers these as receipts or protocol
-  // envelopes, only as explicit user choices.
-  return Boolean(
-    message.buttonsResponseMessage ||
-    message.listResponseMessage ||
-    message.templateButtonReplyMessage ||
-    message.interactiveResponseMessage,
-  );
-}
-
-/**
- * Fast check that a Baileys message carries user-visible inbound content
- * (text, media, contact, location, button/list selection). Returns false for
- * protocol/receipt/typing notifications that arrive on the same
- * `messages.upsert` stream as real messages but should not trigger pairing
- * access-control side effects.
- */
-export function hasInboundUserContent(rawMessage: proto.IMessage | undefined): boolean {
-  if (!rawMessage) {
-    return false;
-  }
-  if (extractText(rawMessage)) {
-    return true;
-  }
-  if (extractMediaPlaceholder(rawMessage)) {
-    return true;
-  }
-  if (extractLocationData(rawMessage)) {
-    return true;
-  }
-  // Walk wrappers (ephemeral, viewOnce, etc.) — interactive responses
-  // can arrive nested.
-  for (const candidate of buildMessageChain(rawMessage)) {
-    if (hasInteractiveResponseContent(candidate)) {
-      return true;
-    }
-  }
-  return false;
-}

package/src/inbound/lifecycle.ts

@@ -1,39 +0,0 @@
-type Listener = (...args: unknown[]) => void;
-
-type OffCapableEmitter = {
-  on: (event: string, listener: Listener) => void;
-  off?: (event: string, listener: Listener) => void;
-  removeListener?: (event: string, listener: Listener) => void;
-};
-
-type ClosableSocket = {
-  end?: (error: Error | undefined) => void;
-  ws?: {
-    close?: () => void;
-  };
-};
-
-export function attachEmitterListener(
-  emitter: OffCapableEmitter,
-  event: string,
-  listener: Listener,
-): () => void {
-  emitter.on(event, listener);
-  return () => {
-    if (typeof emitter.off === "function") {
-      emitter.off(event, listener);
-      return;
-    }
-    if (typeof emitter.removeListener === "function") {
-      emitter.removeListener(event, listener);
-    }
-  };
-}
-
-export function closeInboundMonitorSocket(sock: ClosableSocket): void {
-  if (typeof sock.end === "function") {
-    sock.end(new Error("OpenClaw WhatsApp listener close"));
-    return;
-  }
-  sock.ws?.close?.();
-}

package/src/inbound/media.node.test.ts

@@ -1,83 +0,0 @@
-import { beforeAll, beforeEach, describe, expect, it, vi } from "vitest";
-import { mockNormalizeMessageContent } from "../../../../test/mocks/baileys.js";
-
-type MockMessageInput = Parameters<typeof mockNormalizeMessageContent>[0];
-
-const { normalizeMessageContent, downloadMediaMessage } = vi.hoisted(() => ({
-  normalizeMessageContent: vi.fn((msg: MockMessageInput) => mockNormalizeMessageContent(msg)),
-  downloadMediaMessage: vi.fn().mockResolvedValue(Buffer.from("fake-media-data")),
-}));
-
-vi.mock("@whiskeysockets/baileys", async () => {
-  return {
-    DisconnectReason: { loggedOut: 401 },
-    normalizeMessageContent,
-    downloadMediaMessage,
-  };
-});
-
-let downloadInboundMedia: typeof import("./media.js").downloadInboundMedia;
-
-const mockSock = {
-  updateMediaMessage: vi.fn(),
-  logger: { child: () => ({}) },
-};
-
-async function expectMimetype(message: Record<string, unknown>, expected: string) {
-  const result = await downloadInboundMedia({ message } as never, mockSock as never);
-  expect(result).toBeDefined();
-  expect(result?.mimetype).toBe(expected);
-}
-
-describe("downloadInboundMedia", () => {
-  beforeAll(async () => {
-    ({ downloadInboundMedia } = await import("./media.js"));
-  });
-
-  beforeEach(() => {
-    normalizeMessageContent.mockClear();
-    downloadMediaMessage.mockClear();
-    mockSock.updateMediaMessage.mockClear();
-  });
-
-  it("returns undefined for messages without media", async () => {
-    const msg = { message: { conversation: "hello" } } as never;
-    const result = await downloadInboundMedia(msg, mockSock as never);
-    expect(result).toBeUndefined();
-  });
-
-  it("uses explicit mimetype from audioMessage when present", async () => {
-    await expectMimetype({ audioMessage: { mimetype: "audio/mp4", ptt: true } }, "audio/mp4");
-  });
-
-  it.each([
-    { name: "voice messages without explicit MIME", audioMessage: { ptt: true } },
-    { name: "audio messages without MIME or ptt flag", audioMessage: {} },
-  ])("defaults to audio/ogg for $name", async ({ audioMessage }) => {
-    await expectMimetype({ audioMessage }, "audio/ogg; codecs=opus");
-  });
-
-  it("uses explicit mimetype from imageMessage when present", async () => {
-    await expectMimetype({ imageMessage: { mimetype: "image/png" } }, "image/png");
-  });
-
-  it.each([
-    { name: "image", message: { imageMessage: {} }, mimetype: "image/jpeg" },
-    { name: "video", message: { videoMessage: {} }, mimetype: "video/mp4" },
-    { name: "sticker", message: { stickerMessage: {} }, mimetype: "image/webp" },
-  ])("defaults MIME for $name messages without explicit MIME", async ({ message, mimetype }) => {
-    await expectMimetype(message, mimetype);
-  });
-
-  it("preserves fileName from document messages", async () => {
-    const msg = {
-      message: {
-        documentMessage: { mimetype: "application/pdf", fileName: "report.pdf" },
-      },
-    } as never;
-    const result = await downloadInboundMedia(msg, mockSock as never);
-    expect(result).toBeDefined();
-    expect(result?.mimetype).toBe("application/pdf");
-    expect(result?.fileName).toBe("report.pdf");
-  });
-});

package/src/inbound/media.ts

@@ -1,102 +0,0 @@
-import type { proto, WAMessage } from "@whiskeysockets/baileys";
-import { logVerbose } from "openclaw/plugin-sdk/runtime-env";
-import type { createWaSocket } from "../session.js";
-import { extractContextInfo } from "./extract.js";
-import { downloadMediaMessage, normalizeMessageContent } from "./runtime-api.js";
-
-function unwrapMessage(message: proto.IMessage | undefined): proto.IMessage | undefined {
-  const normalized = normalizeMessageContent(message);
-  return normalized;
-}
-
-/**
- * Resolve the MIME type for an inbound media message.
- * Falls back to WhatsApp's standard formats when Baileys omits the MIME.
- */
-function resolveMediaMimetype(message: proto.IMessage): string | undefined {
-  const explicit =
-    message.imageMessage?.mimetype ??
-    message.videoMessage?.mimetype ??
-    message.documentMessage?.mimetype ??
-    message.audioMessage?.mimetype ??
-    message.stickerMessage?.mimetype ??
-    undefined;
-  if (explicit) {
-    return explicit;
-  }
-  // WhatsApp voice messages (PTT) and audio use OGG Opus by default
-  if (message.audioMessage) {
-    return "audio/ogg; codecs=opus";
-  }
-  if (message.imageMessage) {
-    return "image/jpeg";
-  }
-  if (message.videoMessage) {
-    return "video/mp4";
-  }
-  if (message.stickerMessage) {
-    return "image/webp";
-  }
-  return undefined;
-}
-
-export async function downloadInboundMedia(
-  msg: proto.IWebMessageInfo,
-  sock: Awaited<ReturnType<typeof createWaSocket>>,
-): Promise<{ buffer: Buffer; mimetype?: string; fileName?: string } | undefined> {
-  const message = unwrapMessage(msg.message as proto.IMessage | undefined);
-  if (!message) {
-    return undefined;
-  }
-  const mimetype = resolveMediaMimetype(message);
-  const fileName = message.documentMessage?.fileName ?? undefined;
-  if (
-    !message.imageMessage &&
-    !message.videoMessage &&
-    !message.documentMessage &&
-    !message.audioMessage &&
-    !message.stickerMessage
-  ) {
-    return undefined;
-  }
-  try {
-    const buffer = await downloadMediaMessage(
-      msg as WAMessage,
-      "buffer",
-      {},
-      {
-        reuploadRequest: sock.updateMediaMessage,
-        logger: sock.logger,
-      },
-    );
-    return { buffer, mimetype, fileName };
-  } catch (err) {
-    logVerbose(`downloadMediaMessage failed: ${String(err)}`);
-    return undefined;
-  }
-}
-
-export async function downloadQuotedInboundMedia(
-  msg: proto.IWebMessageInfo,
-  sock: Awaited<ReturnType<typeof createWaSocket>>,
-): Promise<{ buffer: Buffer; mimetype?: string; fileName?: string } | undefined> {
-  const message = unwrapMessage(msg.message as proto.IMessage | undefined);
-  const contextInfo = extractContextInfo(message);
-  if (!contextInfo?.quotedMessage) {
-    return undefined;
-  }
-  const quotedMessage = contextInfo.quotedMessage;
-  return downloadInboundMedia(
-    {
-      key: {
-        id: contextInfo?.stanzaId || undefined,
-        remoteJid: contextInfo.remoteJid ?? msg.key?.remoteJid ?? undefined,
-        participant: contextInfo?.participant ?? undefined,
-        fromMe: false,
-      },
-      message: quotedMessage,
-      messageTimestamp: msg.messageTimestamp,
-    },
-    sock,
-  );
-}