telegram-user: support voice-note media

extensions/telegram-user/src/monitor/handler.ts

@@ -663,6 +663,7 @@ export function createTelegramUserMessageHandler(params: TelegramUserHandlerPara
                   replyToId,
                   threadId,
                   mediaUrl,
+                  audioAsVoice: payload.audioAsVoice === true,
                   maxBytes: mediaMaxMb * 1024 * 1024,
                 });
               } catch (err) {

extensions/telegram-user/src/send.test.ts

@@ -0,0 +1,110 @@
+import { beforeEach, describe, expect, it, vi } from "vitest";
+
+const loadWebMedia = vi.fn();
+
+vi.mock("./runtime.js", () => {
+  return {
+    getTelegramUserRuntime: () => ({
+      config: { loadConfig: () => ({}) },
+      media: {
+        loadWebMedia: (...args: unknown[]) => loadWebMedia(...args),
+      },
+    }),
+  };
+});
+
+const inputMediaAuto = vi.fn((file: unknown, params: unknown) => ({
+  type: "auto",
+  file,
+  ...(params && typeof params === "object" ? params : {}),
+}));
+const inputMediaVoice = vi.fn((file: unknown, params: unknown) => ({
+  type: "voice",
+  file,
+  ...(params && typeof params === "object" ? params : {}),
+}));
+
+vi.mock("@mtcute/core", () => {
+  return {
+    InputMedia: {
+      auto: (...args: unknown[]) => inputMediaAuto(...args),
+      voice: (...args: unknown[]) => inputMediaVoice(...args),
+      poll: () => ({ type: "poll" }),
+    },
+  };
+});
+
+describe("telegram-user send", () => {
+  beforeEach(() => {
+    loadWebMedia.mockReset();
+    inputMediaAuto.mockClear();
+    inputMediaVoice.mockClear();
+  });
+
+  it("sends audio media as voice note when audioAsVoice is set", async () => {
+    loadWebMedia.mockResolvedValue({
+      buffer: Buffer.from("voice"),
+      contentType: "audio/ogg",
+      fileName: "note.ogg",
+    });
+
+    const sendMedia = vi.fn(async () => ({ id: 123 }));
+    const { sendMediaTelegramUser } = await import("./send.js");
+    await sendMediaTelegramUser("telegram-user:123", "hi", {
+      client: { sendMedia } as unknown as import("@mtcute/node").TelegramClient,
+      mediaUrl: "https://example.com/note.ogg",
+      audioAsVoice: true,
+    });
+
+    expect(inputMediaVoice).toHaveBeenCalledTimes(1);
+    expect(sendMedia).toHaveBeenCalledTimes(1);
+    const [, media] = sendMedia.mock.calls[0] ?? [];
+    expect(media).toMatchObject({ type: "voice" });
+  });
+
+  it("falls back to normal media when audioAsVoice is set but media is not voice-compatible", async () => {
+    loadWebMedia.mockResolvedValue({
+      buffer: Buffer.from("img"),
+      contentType: "image/png",
+      fileName: "image.png",
+    });
+
+    const sendMedia = vi.fn(async () => ({ id: 123 }));
+    const { sendMediaTelegramUser } = await import("./send.js");
+    await sendMediaTelegramUser("telegram-user:123", "hi", {
+      client: { sendMedia } as unknown as import("@mtcute/node").TelegramClient,
+      mediaUrl: "https://example.com/image.png",
+      audioAsVoice: true,
+    });
+
+    expect(inputMediaVoice).toHaveBeenCalledTimes(0);
+    expect(inputMediaAuto).toHaveBeenCalledTimes(1);
+  });
+
+  it("falls back to auto when voice messages are forbidden", async () => {
+    loadWebMedia.mockResolvedValue({
+      buffer: Buffer.from("voice"),
+      contentType: "audio/ogg",
+      fileName: "note.ogg",
+    });
+
+    const sendMedia = vi.fn(async (_to: unknown, media: unknown) => {
+      if (media && typeof media === "object" && (media as { type?: string }).type === "voice") {
+        throw new Error("VOICE_MESSAGES_FORBIDDEN");
+      }
+      return { id: 123 };
+    });
+
+    const { sendMediaTelegramUser } = await import("./send.js");
+    await sendMediaTelegramUser("telegram-user:123", "hi", {
+      client: { sendMedia } as unknown as import("@mtcute/node").TelegramClient,
+      mediaUrl: "https://example.com/note.ogg",
+      audioAsVoice: true,
+    });
+
+    expect(inputMediaVoice).toHaveBeenCalledTimes(1);
+    expect(inputMediaAuto).toHaveBeenCalledTimes(1);
+    expect(sendMedia).toHaveBeenCalledTimes(2);
+  });
+});
+

extensions/telegram-user/src/send.ts

@@ -39,6 +39,7 @@ export type TelegramUserSendOpts = {
   replyToId?: number;
   threadId?: string | number | null;
   mediaUrl?: string;
+  audioAsVoice?: boolean;
 };
 
 function normalizeTarget(raw: string): string {
@@ -69,6 +70,24 @@ function resolveTargetAndThread(raw: string, threadId?: string | number | null)
   return { target, threadId: parsedThreadId };
 }
 
+function isVoiceMessagesForbidden(err: unknown): boolean {
+  const message = err instanceof Error ? err.message : String(err);
+  return /VOICE_MESSAGES_FORBIDDEN/i.test(message);
+}
+
+function shouldSendAsVoice(params: {
+  wantsVoice: boolean;
+  contentType?: string | null;
+  fileName?: string | null;
+}): boolean {
+  if (!params.wantsVoice) return false;
+  const contentType = params.contentType?.toLowerCase() ?? "";
+  const fileName = params.fileName?.toLowerCase() ?? "";
+  if (/(^|\/)(ogg|opus)(;|$)/.test(contentType)) return true;
+  if (/\.(ogg|opus|oga)$/.test(fileName)) return true;
+  return false;
+}
+
 export function normalizeTelegramUserMessagingTarget(raw: string): string {
   return normalizeTarget(raw);
 }
@@ -192,11 +211,24 @@ export async function sendMediaTelegramUser(
     const resolved = resolveTargetAndThread(to, opts.threadId);
     const target = resolveTelegramUserPeer(resolved.target);
     const media = await getTelegramUserRuntime().media.loadWebMedia(opts.mediaUrl, opts.maxBytes);
-    const input = InputMedia.auto(media.buffer, {
-      fileName: media.fileName ?? undefined,
-      fileMime: media.contentType,
-      caption: text,
+    const wantsVoice = shouldSendAsVoice({
+      wantsVoice: opts.audioAsVoice === true,
+      contentType: media.contentType,
+      fileName: media.fileName,
     });
+    const buildAuto = () =>
+      InputMedia.auto(media.buffer, {
+        fileName: media.fileName ?? undefined,
+        fileMime: media.contentType,
+        caption: text,
+      });
+    const buildVoice = () =>
+      InputMedia.voice(media.buffer, {
+        fileName: media.fileName ?? undefined,
+        fileMime: media.contentType,
+        caption: text,
+      });
+    const input = wantsVoice ? buildVoice() : buildAuto();
     let message: Awaited<ReturnType<TelegramClient["sendMedia"]>> | null = null;
     try {
       message = await client.sendMedia(target, input, {
@@ -204,7 +236,14 @@ export async function sendMediaTelegramUser(
         ...(resolved.threadId ? { threadId: resolved.threadId } : {}),
       });
     } catch (err) {
-      if (!isDestroyedClientError(err)) throw err;
+      if (wantsVoice && isVoiceMessagesForbidden(err)) {
+        message = await client.sendMedia(target, buildAuto(), {
+          ...(opts.replyToId ? { replyTo: opts.replyToId } : {}),
+          ...(resolved.threadId ? { threadId: resolved.threadId } : {}),
+        });
+      } else if (!isDestroyedClientError(err)) {
+        throw err;
+      }
     }
     if (!message) {
       return { messageId: "", chatId: String(target) };