Qdrant Client

Python applications usually reach a Qdrant database through this official SDK, installed with pip install qdrant-client. It wraps every server endpoint with type hints, talks REST or gRPC (prefer_grpc speeds up bulk loads), and provides matching synchronous QdrantClient and asynchronous AsyncQdrantClient classes, plus upload_points and upload_collection helpers that batch big inserts. The standout feature is local mode: passing :memory: or a folder path gives the same API over an in-process store, so unit tests, CI jobs and notebooks run without a server, and moving to a real cluster later only changes the constructor arguments. An optional fastembed extra lets you pass raw text as models.Document objects and have embeddings computed on your machine with ONNX Runtime, on CPU or, via the separate fastembed-gpu extra, on GPU; Qdrant Cloud can embed server-side instead, but only on paid plans. Versions track the server closely and the package is Apache-2.0, free for commercial use. It suits Python RAG pipelines, LangChain or LlamaIndex vector stores, and prototypes that may later move to a managed cluster.