Production_Rag / config.py
TharaKavin's picture
Upload 17 files
6f94597 verified
Raw History Blame Contribute Delete
1.18 kB
import os
from dotenv import load_dotenv
load_dotenv()
os.environ["HF_HUB_DISABLE_SYMLINKS_WARNING"] = "1"
BASE_DIR = os.path.dirname(os.path.abspath(__file__))
DATA_PATH = os.path.join(BASE_DIR, "DeepSeek_v4.json")
CHROMA_DB_PATH = os.path.join(BASE_DIR, "chroma_db")
EMBEDDING_MODEL = "BAAI/bge-small-en-v1.5"
EMBEDDING_DIM = 384
EMBEDDING_DEVICE = "cpu"
RERANKER_MODEL = "cross-encoder/ms-marco-MiniLM-L-6-v2"
RERANKER_DEVICE = "cpu"
CHUNK_SIZE = 512
CHUNK_OVERLAP = 64
BM25_TOP_K = 15
DENSE_TOP_K = 15
HYBRID_TOP_K = 15
RERANK_CANDIDATES = 10
RERANK_TOP_K = 7
FINAL_TOP_K = 5
HYBRID_ALPHA = 0.5
HYBRID_FUSION_METHOD = "normalized"
CACHE_MAX_SIZE = 200
CACHE_TTL_SECONDS = 3600
HNSW_SPACE = "cosine"
HNSW_EF_CONSTRUCTION = 200
HNSW_M = 32
HNSW_EF_SEARCH = 384
DUP_SIM_THRESHOLD = 0.92
DUP_NGRAM_SIZE = 5
DUP_NUM_PERM = 128
GROQ_API_KEY = os.getenv("GROQ_API_KEY", "")
GROQ_MODEL = "llama-3.1-8b-instant"
GROQ_TEMPERATURE = 0.1
GROQ_MAX_TOKENS = 2048
GROQ_TIMEOUT = 10
MAX_CONTEXT_TOKENS = 8192
SYSTEM_PROMPT_RATIO = 0.15
CHUNK_BUDGET_RATIO = 0.70
GENERATION_RATIO = 0.15
EVAL_NUM_QUESTIONS = 20
EVAL_RETRIEVAL_K = [1, 5, 10]
GRADIO_PORT = 7860
GRADIO_SHARE = False