How to beginner · 3 min read

How to create a Qdrant collection

Q: How to create a Qdrant collection

Use the qdrant-client Python SDK to create a Qdrant collection by initializing a QdrantClient and calling create_collection() with your collection name and vector size. This sets up the collection schema for storing and searching vectors efficiently.

Quick answer

Use the qdrant-client Python SDK to create a Qdrant collection by initializing a QdrantClient and calling create_collection() with your collection name and vector size. This sets up the collection schema for storing and searching vectors efficiently.

PREREQUISITES

Python 3.8+
pip install qdrant-client
Access to a running Qdrant instance (local or cloud)

Setup

Install the official qdrant-client Python package and ensure you have access to a Qdrant server, either locally or hosted. Set the connection parameters accordingly.

bash

pip install qdrant-client

Step by step

This example shows how to create a Qdrant collection named my_collection with 128-dimensional vectors and default distance metric Cosine.

python

from qdrant_client import QdrantClient
from qdrant_client.http.models import Distance

# Connect to local Qdrant instance
client = QdrantClient(host="localhost", port=6333)

# Create collection with vector size and distance metric
client.recreate_collection(
    collection_name="my_collection",
    vector_size=128,
    distance=Distance.COSINE
)

print("Collection 'my_collection' created successfully.")

output

Collection 'my_collection' created successfully.

Common variations

Use Distance.EUCLID or Distance.DOT for different similarity metrics.
Connect to a remote Qdrant server by specifying api_key and https parameters.
Use create_collection() instead of recreate_collection() to avoid overwriting existing collections.

python

from qdrant_client import QdrantClient
from qdrant_client.http.models import Distance

# Connect to remote Qdrant with API key
client = QdrantClient(
    url="https://your-qdrant-host",
    api_key=os.environ["QDRANT_API_KEY"]
)

# Create collection with Euclidean distance
client.create_collection(
    collection_name="euclid_collection",
    vector_size=256,
    distance=Distance.EUCLID
)

print("Collection 'euclid_collection' created.")

output

Collection 'euclid_collection' created.

Troubleshooting

If you get a connection error, verify Qdrant server is running and accessible at the specified host and port.
Use recreate_collection() carefully as it deletes existing data in the collection.
Check vector size matches your embeddings dimension to avoid errors on insert.

✅

Key Takeaways

Use the official qdrant-client Python SDK to manage collections programmatically.
Specify vector size and distance metric when creating collections to optimize similarity search.
Use recreate_collection() to reset collections but beware it deletes existing data.
Ensure your Qdrant server is running and accessible before creating collections.
Adjust connection parameters for local or remote Qdrant instances including API keys.

Verified 2026-04

Verify ↗