File size: 1,713 Bytes
1b5b7a4 dbf2148 759c130 2298900 673edb9 759c130 1b5b7a4 673edb9 759c130 1b5b7a4 759c130 1b5b7a4 8289f9b 673edb9 be8cb43 7775d9b be8cb43 f16065f 55d88b0 759c130 1b5b7a4 6323ac8 dbf2148 f16065f 759c130 8289f9b 3aba5ae f16065f 8289f9b f16065f 2302a4d f16065f a440ca3 9c19512 a440ca3 759c130 dbf2148 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 |
import os
from dotenv import load_dotenv
load_dotenv()
class Settings:
GROQ_API_KEY = os.getenv("GROQ_API_KEY")
# Multilingual Model Settings
VIETNAMESE_EMBEDDING_MODEL = 'keepitreal/vietnamese-sbert'
VIETNAMESE_LLM_MODEL = "llama-3.1-8b-instant"
MULTILINGUAL_EMBEDDING_MODEL = 'sentence-transformers/all-MiniLM-L6-v2'
MULTILINGUAL_LLM_MODEL = "llama-3.1-8b-instant"
# Fallback models
FALLBACK_MULTILINGUAL_EMBEDDING_MODEL = 'sentence-transformers/all-MiniLM-L6-v2'
# Default models
DEFAULT_EMBEDDING_MODEL = 'sentence-transformers/all-MiniLM-L6-v2'
DEFAULT_LLM_MODEL = "llama-3.1-8b-instant"
OCR_MODEL = "kha-white/manga-ocr-base"
EASYOCR_LANGUAGES = ['vi', 'en']
# Whisper Settings
WHISPER_MODEL = "whisper-large-v3"
# TTS Settings
MAX_CHUNK_LENGTH = 200
SUPPORTED_LANGUAGES = {
'vi': 'vi', 'en': 'en', 'fr': 'fr', 'es': 'es',
'de': 'de', 'ja': 'ja', 'ko': 'ko', 'zh': 'zh'
}
# RAG Settings
EMBEDDING_DIMENSION = 768
TOP_K_RESULTS = 5
# Audio Processing Settings
SAMPLE_RATE = 16000
AUDIO_CHUNK_SIZE = 1024
AUDIO_SILENCE_THRESHOLD = 0.0005
MIN_AUDIO_DURATION = 0.8
MAX_AUDIO_DURATION = 15.0
# VAD Settings
VAD_MODEL = "snakers4/silero-vad"
VAD_THRESHOLD = 0.3
VAD_MIN_SPEECH_DURATION = 1.0
VAD_MIN_SILENCE_DURATION = 2.0
VAD_SPEECH_PAD_DURATION = 0.5
VAD_PRE_SPEECH_BUFFER = 0.3
# VOSK Settings
VOSK_MODEL_PATH = "models/vosk-model-small-vn-0.4"
VOSK_SAMPLE_RATE = 16000
VOSK_SILENCE_TIMEOUT = 2.0
# Tắt Whisper nếu dùng VOSK
USE_VOSK_ASR = True
settings = Settings() |