What is Qdrant?
Qdrant is a vector similarity search engine and vector database written in Rust. It provides a production-ready service with a convenient API for storing, searching, and managing points (vectors + payload metadata).
Why Qdrant over Chroma?
- Performance — Written in Rust; significantly faster at high query volumes.
- Filtering — Rich payload filtering combined with vector search (filter by metadata AND search by vector simultaneously).
- Quantization — Reduce memory usage via scalar or product quantization without major accuracy loss.
- Distributed — Horizontal scaling, sharding, replication built-in.
- Snapshots — Point-in-time backup of collections.
Quick Start with Docker
docker run -p 6333:6333 qdrant/qdrant
Then access the dashboard at http://localhost:6333/dashboard.
Python Client
from qdrant_client import QdrantClient
from qdrant_client.models import Distance, VectorParams
client = QdrantClient("localhost", port=6333)
client.create_collection("my_collection", vectors_config=VectorParams(size=384, distance=Distance.COSINE))
When to Choose Qdrant
Choose Qdrant when your RAG app is going to production, you have >100k vectors, or you need advanced filtering. Use Chroma for quick prototyping.