Qdrant (Vector Database)

Qdrant is a high-performance vector database for semantic search and RAG pipelines — store embeddings and retrieve the most similar documents in milliseconds.

Aithroyz deploys Qdrant at internal IP 10.0.0.36. Flowise's Qdrant node is pre-configured to connect to the internal address when both are in the same plan — no URL or credential setup required.

Access

Dashboard: https://qdrant.<env-name>.ops.aithroyz.com

REST API: https://qdrant.<env-name>.ops.aithroyz.com(same host, path-routed)

Internal (within sandbox): http://10.0.0.36:6333

Auth: No API key required from within the sandbox. External access is protected by Google SSO at the gateway.

Creating a collection

A collection is Qdrant's equivalent of a table. Define the vector size to match your embedding model (1536 for OpenAI text-embedding-3-small, 384 for nomic-embed-text):

# Create a collection with OpenAI-compatible 1536-dim embeddings
curl -X PUT https://qdrant.<env-name>.ops.aithroyz.com/collections/my_docs \
  -H "Content-Type: application/json" \
  -d '{
    "vectors": {
      "size": 1536,
      "distance": "Cosine"
    }
  }'

# List all collections
curl https://qdrant.<env-name>.ops.aithroyz.com/collections

Inserting and searching

# Insert a point (document chunk) with a vector and payload
curl -X PUT https://qdrant.<env-name>.ops.aithroyz.com/collections/my_docs/points \
  -H "Content-Type: application/json" \
  -d '{
    "points": [
      {
        "id": 1,
        "vector": [0.1, 0.2, 0.3, ...],
        "payload": {
          "source": "report-2024.pdf",
          "page": 3,
          "text": "The threat actor used spearphishing lures..."
        }
      }
    ]
  }'

# Semantic search — returns top 5 most similar points
curl -X POST https://qdrant.<env-name>.ops.aithroyz.com/collections/my_docs/points/search \
  -H "Content-Type: application/json" \
  -d '{
    "vector": [0.1, 0.2, 0.3, ...],
    "limit": 5,
    "with_payload": true
  }'

✓

In practice you won't generate vectors manually — use Flowise (Qdrant node + Embeddings node) or the LangChain Qdrant integration to handle chunking and embedding automatically.

Using with Flowise

In Flowise, add a Qdrant node to your chatflow and set these values:

Qdrant node settings:
  Qdrant URL:      http://10.0.0.36:6333   # internal — no TLS overhead
  Collection Name: my_docs
  Embeddings:      (connect an Embeddings node using the LLM Gateway credential)

Use the internal IP instead of the public subdomain. This keeps all embedding traffic inside the VPC — faster and free of egress costs.

Tips

Payload for citations

Store source filename and page number in the payload (e.g. {source, page, text}). Use these fields in your prompt template to cite documents in answers.

Qdrant dashboard

The web UI at the public subdomain lets you browse collections, inspect individual points, and run search queries — no code required.

Payload filtering

Combine vector search with metadata filters (e.g. only search documents from a specific date range or source). Use the filter field in the search body.

Named vectors

A single collection can store multiple named vector spaces (e.g. title_vector and body_vector) — useful for hybrid search strategies.

Flowise (LangChain Visual Builder)Read article →Open WebUI (AI Chat)Read article →LLM GatewayRead article →