fix(telegram): guard voice note sends

src/telegram/bot.ts

@@ -60,6 +60,7 @@ import { resolveTelegramAccount } from "./accounts.js";
 import { createTelegramDraftStream } from "./draft-stream.js";
 import { resolveTelegramFetch } from "./fetch.js";
 import { markdownToTelegramHtml } from "./format.js";
+import { isTelegramVoiceCompatible } from "./voice.js";
 import {
   readTelegramAllowFromStore,
   upsertTelegramPairingRequest,
@@ -1387,7 +1388,19 @@ async function deliverReplies(params: {
           ...mediaParams,
         });
       } else if (kind === "audio") {
-        const useVoice = reply.audioAsVoice === true; // default false (backward compatible)
+        const wantsVoice = reply.audioAsVoice === true; // default false (backward compatible)
+        const canVoice = wantsVoice
+          ? isTelegramVoiceCompatible({
+              contentType: media.contentType,
+              fileName,
+            })
+          : false;
+        const useVoice = wantsVoice && canVoice;
+        if (wantsVoice && !canVoice) {
+          logVerbose(
+            `Telegram voice requested but media is ${media.contentType ?? "unknown"} (${fileName}); sending as audio file instead.`,
+          );
+        }
         if (useVoice) {
           // Voice message - displays as round playable bubble (opt-in via [[audio_as_voice]])
           await bot.api.sendVoice(chatId, file, {

src/telegram/send.test.ts

@@ -324,6 +324,40 @@ describe("sendMessageTelegram", () => {
     expect(sendAudio).not.toHaveBeenCalled();
   });
 
+  it("falls back to audio when asVoice is true but media is not voice compatible", async () => {
+    const chatId = "123";
+    const sendAudio = vi.fn().mockResolvedValue({
+      message_id: 14,
+      chat: { id: chatId },
+    });
+    const sendVoice = vi.fn().mockResolvedValue({
+      message_id: 15,
+      chat: { id: chatId },
+    });
+    const api = { sendAudio, sendVoice } as unknown as {
+      sendAudio: typeof sendAudio;
+      sendVoice: typeof sendVoice;
+    };
+
+    loadWebMedia.mockResolvedValueOnce({
+      buffer: Buffer.from("audio"),
+      contentType: "audio/mpeg",
+      fileName: "clip.mp3",
+    });
+
+    await sendMessageTelegram(chatId, "caption", {
+      token: "tok",
+      api,
+      mediaUrl: "https://example.com/clip.mp3",
+      asVoice: true,
+    });
+
+    expect(sendAudio).toHaveBeenCalledWith(chatId, expect.anything(), {
+      caption: "caption",
+    });
+    expect(sendVoice).not.toHaveBeenCalled();
+  });
+
   it("includes message_thread_id for forum topic messages", async () => {
     const chatId = "-1001234567890";
     const sendMessage = vi.fn().mockResolvedValue({

src/telegram/send.ts

@@ -10,6 +10,7 @@ import { formatErrorMessage } from "../infra/errors.js";
 import { recordProviderActivity } from "../infra/provider-activity.js";
 import type { RetryConfig } from "../infra/retry.js";
 import { createTelegramRetryRunner } from "../infra/retry-policy.js";
+import { logVerbose } from "../globals.js";
 import { mediaKindFromMime } from "../media/constants.js";
 import { isGifMedia } from "../media/mime.js";
 import { loadWebMedia } from "../web/media.js";
@@ -20,6 +21,7 @@ import {
   parseTelegramTarget,
   stripTelegramInternalPrefixes,
 } from "./targets.js";
+import { resolveTelegramVoiceDecision } from "./voice.js";
 
 type TelegramSendOpts = {
   token?: string;
@@ -237,7 +239,16 @@ export async function sendMessageTelegram(
         throw wrapChatNotFound(err);
       });
     } else if (kind === "audio") {
-      const useVoice = opts.asVoice === true; // default false (backward compatible)
+      const { useVoice, reason } = resolveTelegramVoiceDecision({
+        wantsVoice: opts.asVoice === true, // default false (backward compatible)
+        contentType: media.contentType,
+        fileName,
+      });
+      if (reason) {
+        logVerbose(
+          `Telegram voice requested but ${reason}; sending as audio file instead.`,
+        );
+      }
       if (useVoice) {
         result = await request(
           () => api.sendVoice(chatId, file, mediaParams),

src/telegram/voice.ts

@@ -0,0 +1,15 @@
+import path from "node:path";
+
+export function isTelegramVoiceCompatible(opts: {
+  contentType?: string | null;
+  fileName?: string | null;
+}): boolean {
+  const mime = opts.contentType?.toLowerCase();
+  if (mime && (mime.includes("ogg") || mime.includes("opus"))) {
+    return true;
+  }
+  const fileName = opts.fileName?.trim();
+  if (!fileName) return false;
+  const ext = path.extname(fileName).toLowerCase();
+  return ext === ".ogg" || ext === ".opus" || ext === ".oga";
+}