fix(discord): keep audio voice replies threaded

nxmxbbd · vincentkoc · commit f02406c9dcba · 2026-06-25T16:22:01.000+08:00
diff --git a/extensions/discord/src/outbound-adapter.test.ts b/extensions/discord/src/outbound-adapter.test.ts
@@ -345,22 +345,47 @@ describe("discordOutbound", () => {
       2,
     );
     expect(messageOptions.accountId).toBe("default");
-    expect(messageOptions.replyTo).toBeUndefined();
+    expect(messageOptions.replyTo).toBe("reply-1");
 
     const mediaCall = mockCall(hoisted.sendMessageDiscordMock, "sendMessageDiscord", 1);
     expect(mediaCall[0]).toBe("channel:123456");
     expect(mediaCall[1]).toBe("");
     const mediaOptions = mockObjectArg(hoisted.sendMessageDiscordMock, "sendMessageDiscord", 1, 2);
     expect(mediaOptions.accountId).toBe("default");
     expect(mediaOptions.mediaUrl).toBe("https://example.com/extra.png");
-    expect(mediaOptions.replyTo).toBeUndefined();
+    expect(mediaOptions.replyTo).toBe("reply-1");
     expect(result).toEqual({
       channel: "discord",
       messageId: "msg-1",
       channelId: "ch-1",
     });
   });
 
+  it("keeps captured replyTo on audioAsVoice sends when replyToMode is batched", async () => {
+    await discordOutbound.sendPayload?.({
+      cfg: {},
+      to: "channel:123456",
+      text: "",
+      payload: {
+        text: "voice note",
+        mediaUrls: ["https://example.com/voice.ogg", "https://example.com/extra.png"],
+        audioAsVoice: true,
+      },
+      accountId: "default",
+      replyToId: "reply-1",
+      replyToMode: "batched",
+    });
+
+    expect(
+      mockObjectArg(hoisted.sendVoiceMessageDiscordMock, "sendVoiceMessageDiscord", 0, 2).replyTo,
+    ).toBe("reply-1");
+    expect(
+      hoisted.sendMessageDiscordMock.mock.calls.map(
+        (call) => (call[2] as { replyTo?: unknown } | undefined)?.replyTo,
+      ),
+    ).toEqual(["reply-1", "reply-1"]);
+  });
+
   it("keeps replyToId on every internal audioAsVoice send when replyToMode is all", async () => {
     await discordOutbound.sendPayload?.({
       cfg: {},
diff --git a/extensions/discord/src/outbound-payload.ts b/extensions/discord/src/outbound-payload.ts
@@ -84,20 +84,23 @@ export async function sendDiscordOutboundPayload(params: {
   const sendContext = await createDiscordPayloadSendContext(ctx);
 
   if (payload.audioAsVoice && mediaUrls.length > 0) {
+    // audioAsVoice emits one logical Discord reply across voice/text/media sends.
+    // Capture before helper calls consume implicit single-use reply targets.
+    const voiceReplyTo = sendContext.resolveReplyTo();
     let lastResult = await sendContext.withRetry(
       async () =>
-        await sendContext.sendVoice(
-          sendContext.target,
-          mediaUrls[0],
-          resolveDiscordDeliveryOptions(ctx, sendContext),
-        ),
+        await sendContext.sendVoice(sendContext.target, mediaUrls[0], {
+          ...resolveDiscordDeliveryOptions(ctx, sendContext),
+          replyTo: voiceReplyTo,
+        }),
     );
     if (payload.text?.trim()) {
       lastResult = await sendContext.withRetry(
         async () =>
           await sendContext.send(sendContext.target, payload.text, {
             verbose: false,
             ...resolveDiscordFormattedDeliveryOptions(ctx, sendContext),
+            replyTo: voiceReplyTo,
           }),
       );
     }
@@ -107,6 +110,7 @@ export async function sendDiscordOutboundPayload(params: {
           await sendContext.send(sendContext.target, "", {
             verbose: false,
             ...resolveDiscordMediaDeliveryOptions(ctx, sendContext, mediaUrl),
+            replyTo: voiceReplyTo,
           }),
       );
     }
