fix: convert telegram audio files into a Record component

astrbot/core/platform/sources/telegram/tg_adapter.py

@@ -588,6 +588,25 @@ def _apply_caption() -> None:
             record.path = path_wav
             message.message = [record]
 
+        elif update.message.audio:
+            # Audio files use their own Bot API field and do not fall back to document.
+            file = await update.message.audio.get_file()
+
+            file_basename = os.path.basename(cast(str, file.file_path))
+            temp_dir = get_astrbot_temp_path()
+            temp_path = os.path.join(temp_dir, file_basename)
+            await download_file(cast(str, file.file_path), path=temp_path)
+            path_wav = await MediaResolver(
+                temp_path,
+                media_type="audio",
+                default_suffix=".wav",
+            ).to_path(target_format="wav")
+
+            record = Comp.Record(file=path_wav, url=path_wav)
+            record.path = path_wav
+            message.message.append(record)
+            _apply_caption()
+
         elif update.message.photo:
             photo = update.message.photo[-1]  # get the largest photo
             file = await photo.get_file()

tests/fixtures/helpers.py

@@ -98,6 +98,7 @@ def create_mock_update(
     video: MagicMock | None = None,
     document: MagicMock | None = None,
     voice: MagicMock | None = None,
+    audio: MagicMock | None = None,
     sticker: MagicMock | None = None,
     video_note: MagicMock | None = None,
     reply_to_message: MagicMock | None = None,
@@ -122,6 +123,7 @@ def create_mock_update(
         video: 视频对象
         document: 文档对象
         voice: 语音对象
+        audio: 音频文件对象
         sticker: 贴纸对象
         video_note: 圆形视频消息对象
         reply_to_message: 回复的消息
@@ -160,6 +162,7 @@ def create_mock_update(
     message.video = video
     message.document = document
     message.voice = voice
+    message.audio = audio
     message.sticker = sticker
     message.video_note = video_note
     message.reply_to_message = reply_to_message

tests/test_telegram_adapter.py

@@ -349,6 +349,49 @@ async def test_telegram_voice_message_creates_record_component(tmp_path):
     assert result.message[0].url == str(wav_path)
 
 
+@pytest.mark.asyncio
+async def test_telegram_audio_caption_populates_message_text_and_plain(tmp_path):
+    TelegramPlatformAdapter = _load_telegram_adapter()
+    adapter = TelegramPlatformAdapter(
+        make_platform_config("telegram"),
+        {},
+        asyncio.Queue(),
+    )
+    audio = create_mock_file("https://api.telegram.org/file/test/song.mp3")
+    update = create_mock_update(
+        message_text=None,
+        audio=audio,
+        caption="这首歌是什么",
+    )
+    wav_path = tmp_path / "song.mp3.wav"
+    convert_message_globals = adapter.convert_message.__func__.__globals__
+
+    with (
+        patch.dict(
+            convert_message_globals,
+            {
+                "get_astrbot_temp_path": MagicMock(return_value=str(tmp_path)),
+                "download_file": AsyncMock(),
+            },
+        ),
+        patch(
+            "astrbot.core.utils.media_utils.ensure_wav",
+            AsyncMock(return_value=str(wav_path)),
+        ),
+    ):
+        result = await adapter.convert_message(update, _build_context())
+
+    assert result is not None
+    assert result.message_str == "这首歌是什么"
+    assert len(result.message) == 2
+    assert isinstance(result.message[0], Comp.Record)
+    assert result.message[0].file == str(wav_path)
+    assert result.message[0].path == str(wav_path)
+    assert result.message[0].url == str(wav_path)
+    assert isinstance(result.message[1], Comp.Plain)
+    assert result.message[1].text == "这首歌是什么"
+
+
 @pytest.mark.asyncio
 async def test_telegram_final_segment_splits_long_markdown_messages():
     TelegramPlatformEvent = _load_telegram_platform_event()