Qdrant Vector Store

The Qdrant Vector Store node enables integration with a Qdrant vector database for storing and retrieving vector embeddings based on semantic similarity. It supports both Qdrant cloud servers and self-hosted deployments. 

Inputs

Documents – Embeddings to be stored in the collection
Questions – Embeddings of queries to retrieve similar documents

Outputs

Documents – Echo or transformation of the stored input documents
Answers – The retrieved, most relevant matches based on the input question
Questions – Echo or transformation of the input questions

Configuration

GUI
  • Qdrant Host Type
    • Options
      • Qdrant cloud server
      • Your own Qdrant server
  • Host Address
    • Cloud – your-instance-name.<region>.qdrant.io
    • Local – localhost
  • Port
    • Cloud Default – 443
    • Local Default – 6333
  • API Key (Cloud) – Required if using Qdrant Cloud
  • Retrieval Score – This score determines the threshold for returning similar items
    • Options – Related, Relevant, Exact, etc.
  • Collection Name – Name of the vector collection
    • Example – APARAVI

Qdrant Vector Store supports three deployment modes:

Embedded Mode

Uses an embedded Qdrant server within Aparavi. Minimal configuration required.

json

{
"profile": "embedded",
"collection": "APARAVI"
}

Local Mode

Connects to a Qdrant server running on your infrastructure.

json

{
"profile": "local",
"host": "localhost",
"port": 6333,
"collection": "APARAVI"
}

Cloud Mode

Connects to a Qdrant cloud instance.

json

{
"profile": "cloud",
"host": "your-instance-name.region.qdrant.io",
"port": 443,
"apikey": "your-api-key",
"collection": "APARAVI"
}

Advanced Settings

  • Similarity – Vector similarity metric
    • Default – Cosine
    • Options – Cosine, Euclid, Dot, Manhattan
  • Render Chunk Size – Maximum size for rendering
    • Default – 32MB
    • Note: Affects memory usage
  • Payload Limit – Maximum payload size
    • Default – 32MB
    • Note: Limits vector entry size

Configuration for Production Use

For production environments, we recommend using either Local or Cloud mode with these settings:

json

{
"profile": "cloud",
"host": "your-instance.us-east.qdrant.io",
"port": 443,
"apikey": "your-api-key",
"collection": "production_data",
"similarity": "Cosine",
"renderChunkSize": 67108864,
"payloadLimit": 67108864
}

Best Practices

  • Use separate collections for different data domains
  • Include version information in collection names for evolving datasets
  • Choose the appropriate similarity metric for your use case
  • Adjust chunk sizes based on your document characteristics
  • Implement proper access controls for cloud deployments

Troubleshooting

Connection Problems

  • Connection refused – Verify host and port settings
  • Authentication failure – Check API key validity
  • Timeout errors – Check network connectivity

Query Performance

  • Slow queries – Try a different similarity metric
  • Memory errors – Adjust chunk and payload limits
  • Poor search results – Ensure consistent embedding dimensions

Technical Reference

For detailed technical information, refer to:

  • Qdrant Official Documentation
  • Aparavi Qdrant Connector Source Code located in ../../../aparavi-connectors/connectors/qdrant/qdrant.py
  • Qdrant Configuration Schema located in ../../../aparavi-connectors/connectors/qdrant/services.json