Qdrant

​

Supported Versions

• [email protected]

​

Deployment Modes

​

Non-HA Mode

• Single node setup ideal for development and non-production projects
• Simple deployment and management

​

HA Cluster

• Automatically configured with 3 nodes
• Recommended for production environments
• Built-in data replication across nodes
• By default (automaticClusterReplication=true), automatically creates replicas of all shards across all three nodes
  • Can be disabled by setting automaticClusterReplication to false
• Automatic cluster recovery and node replacement in case of failures

​

Network Architecture & Access

​

API Keys

• apiKey: Full access API key for administrative operations (creating collections, indexing)
• readOnlyApiKey: Restricted API key for search operations

​

HTTP API

• Port: 6333
• Connection String: Available as connectionString environment variable or construct using http://${hostname}:${port}

​

gRPC API

• Port: 6334
• gRPC Connection String: Available as grpcConnectionString environment variable or construct using tcp://${hostname}:${grpcPort}

​

Connection Examples

​

Python Client

from qdrant_client import QdrantClient
import os

client = QdrantClient(
    url=os.environ['connectionString'],
    api_key=os.environ['apiKey']
)

# Create a collection
client.create_collection(
    collection_name="my_collection",
    vectors_config={
        "size": 384,
        "distance": "Cosine"
    }
)

# Insert vectors
client.upsert(
    collection_name="my_collection",
    points=[
        {
            "id": 1,
            "vector": [0.1] * 384,
            "payload": {"name": "example"}
        }
    ]
)

​

JavaScript Client

import { QdrantClient } from '@qdrant/js-client-rest';

const client = new QdrantClient({
  url: process.env.connectionString,
  apiKey: process.env.apiKey
});

// Create a collection
await client.createCollection('my_collection', {
  vectors: {
    size: 384,
    distance: 'Cosine'
  }
});

// Search
const searchResult = await client.search('my_collection', {
  vector: [0.1, 0.2, 0.3, ...],
  limit: 5
});

​

Data Backup

• Format: .snapshot (compressed)
• Tooling: Native snapshotting
• Source: Taken from the primary node (leader in HA clusters)

​

Restoring Backups

1. Download the backup file (.snapshot) from the Zerops UI
2. Prepare your target environment (clean existing collections or use a new instance)
3. Restore using the Qdrant API. Use the snapshot restore endpoint to import the snapshot file. Follow the official Qdrant documentation for detailed snapshot restore procedures.

​

Use Cases

• Semantic Search - Find similar documents based on meaning, not just keywords
• Recommendation Systems - Suggest products, content, or services based on similarity
• Image Search - Search images by visual similarity
• Anomaly Detection - Identify outliers in high-dimensional data
• RAG Applications - Retrieval-Augmented Generation for LLM applications
• Chatbots - Context-aware conversational AI

​

Best Practices

​

Collection Design

• Choose appropriate vector dimensions based on your embedding model
• Select the right distance metric (Cosine, Euclidean, Dot product)
• Use payload indexing for faster filtering
• Plan collection partitioning for large datasets

​

Performance

• Use batch operations for bulk insertions
• Implement appropriate filtering to reduce search space
• Monitor collection size and shard distribution
• Use gRPC for high-throughput scenarios

​

Security

• Use readOnlyApiKey for search-only operations
• Keep apiKey secure in backend services
• Store credentials in environment variables
• Implement proper access controls in your application

​

High Availability

• Use HA mode for production workloads
• Monitor cluster health and replication status
• Implement retry logic for transient failures
• Test failover scenarios before production deployment

​

Support

• Join our Discord community
• Contact support via email