qdrant
**Qdrant** is a **vector database optimized for semantic search and similarity matching** — storing embeddings at scale with sub-millisecond search latency, perfect for AI applications, recommendation engines, and semantic search.
**What Is Qdrant?**
- **Type**: Specialized vector database for embeddings.
- **Performance**: Sub-millisecond search on millions of vectors.
- **Architecture**: Optimized for HNSW (hierarchical navigable small world).
- **Deployment**: Cloud, self-hosted, hybrid.
- **Scaling**: Distributed clustering for unlimited scale.
**Why Qdrant Matters**
- **Fast**: Sub-millisecond search on massive datasets.
- **Accurate**: Built specifically for vector similarity (not retrofitted).
- **Flexible Filtering**: Combine vector search with metadata filters.
- **Production-Ready**: Used by enterprises for real-time inference.
- **Open Source**: Full control and transparency.
- **Multi-Model**: Store multiple embeddings per item.
**Key Features**
**Efficient Storage**: Compressed vectors reduce memory 30-50%.
**Filtering**: Combine semantic search with exact matches.
**Payload**: Store metadata alongside vectors.
**Replication**: High availability and disaster recovery.
**Sharding**: Distribute across multiple nodes.
**Quick Start**
```python
from qdrant_client import QdrantClient
client = QdrantClient(url="http://localhost:6333")
# Create collection
client.create_collection(
collection_name="documents",
vectors_config={"size": 768}
)
# Add vectors
client.upsert(
collection_name="documents",
points={
"id": 1,
"vector": [0.1, 0.2, ...],
"payload": {"title": "doc1"}
}
)
# Search
results = client.search(
collection_name="documents",
query_vector=[0.1, 0.2, ...],
limit=10
)
```
**Alternatives**: Pinecone, Weaviate, Milvus, Chroma.
Qdrant is the **vector database for AI applications** — sub-millisecond semantic search at scale.