fix(whatsapp): preserve transport envelope ownership (#116989)

Co-authored-by: Peter Steinberger <steipete@macos.shared>

extensions/whatsapp/src/inbound/extract.test.ts

@@ -117,6 +117,71 @@ describe("extractMentionedJids", () => {
         },
       },
     },
+    {
+      name: "shared contacts",
+      message: {
+        contactMessage: {
+          displayName: "Alice",
+          contextInfo: { mentionedJid: [botJid] },
+        },
+      },
+    },
+    {
+      name: "shared contact collections",
+      message: {
+        contactsArrayMessage: {
+          contacts: [{ displayName: "Alice" }],
+          contextInfo: { mentionedJid: [botJid] },
+        },
+      },
+    },
+    {
+      name: "location pins",
+      message: {
+        locationMessage: {
+          degreesLatitude: 1,
+          degreesLongitude: 2,
+          contextInfo: { mentionedJid: [botJid] },
+        },
+      },
+    },
+    {
+      name: "live locations",
+      message: {
+        liveLocationMessage: {
+          degreesLatitude: 1,
+          degreesLongitude: 2,
+          contextInfo: { mentionedJid: [botJid] },
+        },
+      },
+    },
+    {
+      name: "interactive button prompts",
+      message: {
+        buttonsMessage: {
+          contentText: "Choose one",
+          contextInfo: { mentionedJid: [botJid] },
+        },
+      },
+    },
+    {
+      name: "interactive lists",
+      message: {
+        listMessage: {
+          title: "Choose one",
+          contextInfo: { mentionedJid: [botJid] },
+        },
+      },
+    },
+    {
+      name: "native interactive prompts",
+      message: {
+        interactiveMessage: {
+          body: { text: "Choose one" },
+          contextInfo: { mentionedJid: [botJid] },
+        },
+      },
+    },
   ])("preserves direct bot mentions from $name", ({ message }) => {
     expect(extractMentionedJids(message as proto.IMessage)).toEqual([botJid]);
   });

extensions/whatsapp/src/inbound/extract.ts

@@ -113,34 +113,12 @@ export function extractContextInfo(
 }
 
 export function extractMentionedJids(rawMessage: proto.IMessage | undefined): string[] | undefined {
-  const message = unwrapMessage(rawMessage);
-  if (!message) {
+  // Context ownership already follows Baileys envelopes without entering quoted messages.
+  const mentionedJids = extractContextInfo(rawMessage)?.mentionedJid?.filter(Boolean);
+  if (!mentionedJids?.length) {
     return undefined;
   }
-
-  const candidates: Array<string[] | null | undefined> = [
-    message.extendedTextMessage?.contextInfo?.mentionedJid,
-    message.imageMessage?.contextInfo?.mentionedJid,
-    message.videoMessage?.contextInfo?.mentionedJid,
-    message.ptvMessage?.contextInfo?.mentionedJid,
-    message.documentMessage?.contextInfo?.mentionedJid,
-    message.audioMessage?.contextInfo?.mentionedJid,
-    message.stickerMessage?.contextInfo?.mentionedJid,
-    message.buttonsResponseMessage?.contextInfo?.mentionedJid,
-    message.listResponseMessage?.contextInfo?.mentionedJid,
-    message.templateButtonReplyMessage?.contextInfo?.mentionedJid,
-    message.interactiveResponseMessage?.contextInfo?.mentionedJid,
-    message.pollCreationMessage?.contextInfo?.mentionedJid,
-    message.pollCreationMessageV2?.contextInfo?.mentionedJid,
-    message.pollCreationMessageV3?.contextInfo?.mentionedJid,
-    message.pollCreationMessageV5?.contextInfo?.mentionedJid,
-  ];
-
-  const flattened = candidates.flatMap((arr) => arr ?? []).filter(Boolean);
-  if (flattened.length === 0) {
-    return undefined;
-  }
-  return uniqueStrings(flattened);
+  return uniqueStrings(mentionedJids);
 }
 
 function extractNativeFlowResponseText(

extensions/whatsapp/src/send.test.ts

@@ -656,22 +656,25 @@ describe("web outbound", () => {
     });
   });
 
-  it("keeps explicit document kind for prehydrated image payloads", async () => {
-    const buf = Buffer.from("image-as-document");
+  it.each([
+    { contentType: "image/png", fileName: "photo.png" },
+    { contentType: "video/mp4", fileName: "clip.mp4" },
+  ])("keeps explicit document delivery for prehydrated $contentType payloads", async (media) => {
+    const buf = Buffer.from("visual-as-document");
 
     await sendMessageWhatsApp("+1555", "doc", {
       verbose: false,
       cfg: WHATSAPP_TEST_CFG,
       mediaPayload: {
         buffer: buf,
-        contentType: "image/png",
+        ...media,
         kind: "document",
-        fileName: "photo.png",
       },
     });
 
-    expect(sendMessage).toHaveBeenLastCalledWith("+1555", "doc", buf, "image/png", {
-      fileName: "photo.png",
+    expect(sendMessage).toHaveBeenLastCalledWith("+1555", "doc", buf, media.contentType, {
+      asDocument: true,
+      fileName: media.fileName,
     });
   });
 
@@ -836,6 +839,32 @@ describe("web outbound", () => {
     });
   });
 
+  it("returns the actual outbound poll key when Baileys resolves a LID target", async () => {
+    sendPoll.mockResolvedValueOnce({
+      kind: "poll",
+      messageId: "poll-lid",
+      keys: [
+        {
+          id: "poll-lid",
+          remoteJid: "123456789@lid",
+          fromMe: true,
+        },
+      ],
+      providerAccepted: true,
+    });
+
+    await expect(
+      sendPollWhatsApp(
+        "+1555",
+        { question: "Lunch?", options: ["Pizza", "Sushi"] },
+        { verbose: false, cfg: WHATSAPP_TEST_CFG },
+      ),
+    ).resolves.toEqual({
+      messageId: "poll-lid",
+      toJid: "123456789@lid",
+    });
+  });
+
   it("checks send readiness before sending direct polls", async () => {
     const assertSendReady = vi.fn(async () => {
       throw new Error("WhatsApp reachout timelock is active");

extensions/whatsapp/src/send.ts

@@ -49,9 +49,10 @@ function buildWhatsAppMediaSendState(params: {
   forceDocument?: boolean;
 }): WhatsAppMediaSendState {
   const { media, caption } = params;
-  const forceDocumentDelivery = Boolean(
-    params.forceDocument && supportsForcedDocumentDelivery(media.kind),
-  );
+  const forceDocumentDelivery =
+    Boolean(params.forceDocument && supportsForcedDocumentDelivery(media.kind)) ||
+    (media.kind === "document" &&
+      (media.mimetype.startsWith("image/") || media.mimetype.startsWith("video/")));
   let text = caption ?? "";
   let documentFileName = media.kind === "document" ? media.fileName : undefined;
   let visibleTextAfterVoice: string | undefined;
@@ -405,7 +406,7 @@ export async function sendPollWhatsApp(
     const durationMs = Date.now() - startedAt;
     outboundLog.info(`Sent poll ${messageId} -> ${redactedJid} (${durationMs}ms)`);
     logger.info({ jid: redactedJid, messageId }, "sent poll");
-    return { messageId, toJid: jid };
+    return { messageId, toJid: resolveActualSentRemoteJid(result, jid) };
   } catch (err) {
     logger.error({ err: String(err), to: redactedTo }, "failed to send poll via web session");
     throw err;