from __future__ import annotations import os from pathlib import Path from dotenv import load_dotenv # Explicit runtime environment variables take precedence over .env. load_dotenv(override=False) PACKAGE_DIR = Path(__file__).resolve().parent PROJECT_ROOT = PACKAGE_DIR.parent DEFAULT_CSV = PACKAGE_DIR / "data" / "papers.csv" # In the HF Search Space, metadata_search.config can be imported before # metadata_search.app executes its Dataset bootstrap (because Python imports the # package before running `python -m metadata_search.app`). Therefore the cloud # paths must be resolvable here, not only later inside app.py. USE_DATASET_BOOTSTRAP = os.getenv("WATERSCOPE_DATA_BOOTSTRAP", "1").lower() in { "1", "true", "yes", } DATA_ROOT = Path( os.getenv("WATERSCOPE_DATA_ROOT", "/tmp/waterscope-data") ).resolve() if USE_DATASET_BOOTSTRAP: default_paper_csv = DATA_ROOT / "metadata" / "papers.csv" default_qdrant_path = DATA_ROOT / "metadata_search_data" / "qdrant" default_qdrant_manifest = DATA_ROOT / "metadata_search_data" / "qdrant_manifest.json" else: default_paper_csv = DEFAULT_CSV default_qdrant_path = PROJECT_ROOT / "metadata_search_data" / "qdrant" default_qdrant_manifest = PROJECT_ROOT / "metadata_search_data" / "qdrant_manifest.json" PAPER_CSV = Path(os.getenv("PAPER_CSV", str(default_paper_csv))) EMBEDDING_MODEL = os.getenv( "METADATA_EMBEDDING_MODEL", "sentence-transformers/all-MiniLM-L6-v2", ) METADATA_QDRANT_PATH = Path( os.getenv("METADATA_QDRANT_PATH", str(default_qdrant_path)) ) METADATA_QDRANT_COLLECTION = os.getenv( "METADATA_QDRANT_COLLECTION", "papers_metadata", ) METADATA_QDRANT_MANIFEST = Path( os.getenv("METADATA_QDRANT_MANIFEST", str(default_qdrant_manifest)) ) METADATA_SEARCH_PORT = int( os.getenv("PORT", os.getenv("METADATA_SEARCH_PORT", "7860")) ) PAPER_FIGURES_DIR = Path( os.getenv( "PAPER_FIGURES_DIR", str(PROJECT_ROOT / "Media"), ) ) TITLE_COL = os.getenv("METADATA_TITLE_COL", "TI") ABSTRACT_COL = os.getenv("METADATA_ABSTRACT_COL", "AB") KEYWORD_COL = os.getenv("METADATA_KEYWORD_COL", "DE")