Initial commit: UMB Telegram Bot

2026-07-07 18:30:17 +07:00
commit 76e5701eba
31 changed files with 2524 additions and 0 deletions
@@ -0,0 +1,112 @@
+import asyncio
+import logging
+import os
+import re
+from pathlib import Path
+
+from faster_whisper import WhisperModel
+from aiogram import Bot
+
+from config import PROXY_ENABLED, PROXY_URL
+
+logger = logging.getLogger(__name__)
+
+_model = None
+
+BASE_DIR = Path(__file__).resolve().parent.parent.parent
+MODELS_DIR = BASE_DIR / "models" / "faster-whisper"
+VOICE_DIR = BASE_DIR / "bot" / "data" / "voice"
+
+
+def _setup_proxy():
+    if PROXY_ENABLED and PROXY_URL:
+        os.environ["ALL_PROXY"] = PROXY_URL
+        logger.info("Proxy set for model download: %s", PROXY_URL[:30] + "...")
+
+
+def _get_model() -> WhisperModel:
+    global _model
+    if _model is None:
+        MODELS_DIR.mkdir(parents=True, exist_ok=True)
+        VOICE_DIR.mkdir(parents=True, exist_ok=True)
+        model_path = str(MODELS_DIR / "base")
+        if not os.path.isdir(model_path) or not os.path.isfile(os.path.join(model_path, "model.bin")):
+            _setup_proxy()
+            logger.info("Model not found locally, downloading to %s...", MODELS_DIR)
+            _model = WhisperModel(
+                "base",
+                device="cpu",
+                cpu_threads=4,
+                compute_type="int8",
+                download_root=str(MODELS_DIR),
+            )
+        else:
+            logger.info("Loading model from %s...", model_path)
+            _model = WhisperModel(
+                model_path,
+                device="cpu",
+                cpu_threads=4,
+                compute_type="int8",
+            )
+        logger.info("Whisper model loaded")
+    return _model
+
+
+async def download_voice(bot: Bot, file_id: str) -> str:
+    VOICE_DIR.mkdir(parents=True, exist_ok=True)
+    file = await bot.get_file(file_id)
+    path = str(VOICE_DIR / f"{file_id}.ogg")
+    await bot.download_file(file.file_path, destination=path)
+    logger.info("Voice downloaded: %s -> %s", file_id, path)
+    return path
+
+
+async def convert_to_wav(input_path: str) -> str:
+    output_path = input_path.replace(".ogg", ".wav")
+    proc = await asyncio.create_subprocess_exec(
+        "ffmpeg", "-i", input_path, "-ar", "16000", "-ac", "1",
+        "-c:a", "pcm_s16le", output_path, "-y",
+        stdout=asyncio.subprocess.PIPE,
+        stderr=asyncio.subprocess.PIPE,
+    )
+    _, stderr = await proc.communicate()
+    if proc.returncode != 0:
+        logger.error(f"ffmpeg error: {stderr.decode(errors='replace')}")
+        raise RuntimeError("ffmpeg conversion failed")
+    logger.info(f"Converted to WAV: {output_path}")
+    return output_path
+
+
+async def transcribe_audio(file_path: str) -> str:
+    model = _get_model()
+    loop = asyncio.get_running_loop()
+
+    def _transcribe():
+        segments, info = model.transcribe(file_path, language="ru", beam_size=5)
+        text = " ".join(seg.text for seg in segments)
+        return text.strip()
+
+    text = await loop.run_in_executor(None, _transcribe)
+    logger.info(f"Transcription result ({len(text)} chars): {text[:100]}...")
+    return text
+
+
+def normalize_text(text: str) -> str:
+    text = text.strip()
+    if not text:
+        return text
+    text = text[0].upper() + text[1:]
+    if text[-1] not in ".!?…":
+        text += "."
+    text = re.sub(r"\s+", " ", text)
+    return text
+
+
+async def cleanup_files(*paths: str):
+    for path in paths:
+        if path and os.path.exists(path):
+            try:
+                os.unlink(path)
+                logger.debug(f"Cleaned up: {path}")
+            except Exception as e:
+                logger.warning(f"Cleanup failed for {path}: {e}")