FinRAG-MCP

# vectorstore/qdrant_init.py from qdrant_client import QdrantClient from qdrant_client.http.models import Distance, VectorParams import os def connect_qdrant(host: str | None = None, port: int | None = None) -> QdrantClient: """ Prefer local mode if QDRANT_LOCAL_PATH is set; otherwise use host/port. """ local_path = os.getenv("QDRANT_LOCAL_PATH") if local_path: # Persistent local storage in the given folder (no Docker needed) return QdrantClient(path=local_path) return QdrantClient(host=host or "localhost", port=port or 6333) def ensure_collection(client: QdrantClient, name: str, dim: int = 3072): existing = [c.name for c in client.get_collections().collections] if name not in existing: client.create_collection( collection_name=name, vectors_config=VectorParams(size=dim, distance=Distance.COSINE), )