feat: transcribe Telegram voice messages via APInex

bot.py

@@ -1,7 +1,10 @@
 import asyncio
+import io
 import logging
 import os
+from typing import Any
 
+import aiohttp
 from aiogram import Bot, Dispatcher, F, Router
 from aiogram.client.session.aiohttp import AiohttpSession
 from aiogram.client.telegram import TEST
@@ -9,13 +12,105 @@ from aiogram.filters import CommandStart
 from aiogram.types import Message
 
 logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
 
 router = Router()
+APINEX_STT_URL = "https://api.apinex.bond/v1/tools/voice/stt?language=ru"
+MAX_AUDIO_BYTES = 20 * 1024 * 1024
+
+
+def extract_transcript(payload: dict[str, Any]) -> str:
+    """Return the first APInex voice/STT transcript, or an empty string."""
+    channels = payload.get("results", {}).get("channels", [])
+    for channel in channels:
+        alternatives = channel.get("alternatives", [])
+        if alternatives:
+            transcript = alternatives[0].get("transcript", "")
+            if transcript and transcript.strip():
+                return transcript.strip()
+    return ""
+
+
+def split_text(text: str, limit: int = 4000) -> list[str]:
+    """Split a long transcript into Telegram-safe chunks, preferring word boundaries."""
+    chunks: list[str] = []
+    remaining = text.strip()
+    while len(remaining) > limit:
+        split_at = remaining.rfind(" ", 0, limit + 1)
+        if split_at < limit // 2:
+            split_at = limit
+        chunks.append(remaining[:split_at].strip())
+        remaining = remaining[split_at:].strip()
+    if remaining:
+        chunks.append(remaining)
+    return chunks
+
+
+async def transcribe_audio(audio: bytes, content_type: str, api_key: str) -> str:
+    timeout = aiohttp.ClientTimeout(total=120)
+    headers = {
+        "Authorization": f"Bearer {api_key}",
+        "Content-Type": content_type,
+        "Accept": "application/json",
+    }
+    async with aiohttp.ClientSession(timeout=timeout) as session:
+        async with session.post(APINEX_STT_URL, data=audio, headers=headers) as response:
+            if response.status != 200:
+                logger.warning("APInex STT returned HTTP %s", response.status)
+                raise RuntimeError("APInex STT request failed")
+            try:
+                payload = await response.json(content_type=None)
+            except (aiohttp.ContentTypeError, ValueError) as exc:
+                raise RuntimeError("APInex STT returned invalid JSON") from exc
+    return extract_transcript(payload)
 
 
 @router.message(CommandStart())
 async def start(message: Message) -> None:
-    await message.answer("Тестовый эхо-бот запущен. Отправь мне текст — я повторю его.")
+    await message.answer("Тестовый бот запущен. Отправь текст — я повторю его, или голосовое — расшифрую.")
+
+
+@router.message(F.voice | F.audio)
+async def transcribe_voice(message: Message, bot: Bot) -> None:
+    api_key = os.environ.get("APINEX_API_KEY", "").strip()
+    if not api_key:
+        await message.answer("Расшифровка пока не настроена.")
+        return
+
+    audio_file = message.voice or message.audio
+    if audio_file is None:
+        return
+    if audio_file.file_size and audio_file.file_size > MAX_AUDIO_BYTES:
+        await message.answer("Аудиофайл слишком большой. Максимальный размер — 20 МБ.")
+        return
+
+    content_type = "audio/ogg" if message.voice else (message.audio.mime_type or "audio/mpeg")
+    await bot.send_chat_action(message.chat.id, "typing")
+    try:
+        audio_buffer = io.BytesIO()
+        await bot.download(audio_file, destination=audio_buffer)
+        audio = audio_buffer.getvalue()
+        if not audio:
+            await message.answer("Не удалось получить аудиофайл.")
+            return
+        if len(audio) > MAX_AUDIO_BYTES:
+            await message.answer("Аудиофайл слишком большой. Максимальный размер — 20 МБ.")
+            return
+        transcript = await transcribe_audio(audio, content_type, api_key)
+    except (aiohttp.ClientError, asyncio.TimeoutError, RuntimeError) as exc:
+        logger.warning("Voice transcription failed: %s", type(exc).__name__)
+        await message.answer("Не удалось расшифровать аудио. Попробуй ещё раз позже.")
+        return
+    except Exception:
+        logger.exception("Unexpected voice transcription failure")
+        await message.answer("Произошла ошибка при обработке аудио.")
+        return
+
+    if not transcript:
+        await message.answer("Не удалось распознать речь в аудио.")
+        return
+    for chunk in split_text(transcript):
+        await message.answer(chunk)
 
 
 @router.message(F.text & ~F.text.startswith("/"))