Qdrant

Port 6333 Vector similarity search for semantic search and RAG.

Connection

QDRANT_URL=http://db.bizfylabs.com:6333
QDRANT_API_KEY=YOUR_API_KEY

Setup (Python)

pip install qdrant-client

from qdrant_client import QdrantClient
from qdrant_client.models import Distance, VectorParams, PointStruct, Filter, FieldCondition, MatchValue

client = QdrantClient(url=os.environ["QDRANT_URL"], api_key=os.environ["QDRANT_API_KEY"])

Collection CRUD

Name collections per app: myapp_documents, myapp_products

Create Collection

# Vector size must match embedding model (OpenAI text-embedding-3-small = 1536)
client.create_collection(
    collection_name="myapp_documents",
    vectors_config=VectorParams(size=1536, distance=Distance.COSINE),
)

Read — List / Get Info

client.get_collections()
client.get_collection("myapp_documents")

Update Collection

from qdrant_client.models import OptimizersConfigDiff
client.update_collection(
    collection_name="myapp_documents",
    optimizers_config=OptimizersConfigDiff(indexing_threshold=20000),
)

Delete Collection

client.delete_collection("myapp_documents")  # destructive

Point (Vector) CRUD

Create / Upsert Points

client.upsert(
    collection_name="myapp_documents",
    points=[
        PointStruct(
            id="doc-001",
            vector=embedding,  # list[float] from your model
            payload={"title": "Getting Started", "source": "wiki", "text": "..."},
        ),
        PointStruct(id="doc-002", vector=embedding2, payload={"title": "API Ref"}),
    ],
)

Read — Search & Retrieve

# Similarity search
results = client.search(
    collection_name="myapp_documents",
    query_vector=query_embedding,
    limit=5,
    score_threshold=0.7,
)

# Get by ID
client.retrieve(collection_name="myapp_documents", ids=["doc-001"])

# Scroll all (paginated)
client.scroll(collection_name="myapp_documents", limit=100)

Read — With Metadata Filter

results = client.search(
    collection_name="myapp_documents",
    query_vector=query_embedding,
    query_filter=Filter(must=[
        FieldCondition(key="source", match=MatchValue(value="wiki"))
    ]),
    limit=10,
)

Update Payload (metadata)

client.set_payload(
    collection_name="myapp_documents",
    payload={"title": "Updated Title", "status": "published"},
    points=["doc-001"],
)

# Overwrite entire payload
client.overwrite_payload(
    collection_name="myapp_documents",
    payload={"title": "New"},
    points=["doc-001"],
)

Delete Points

client.delete(
    collection_name="myapp_documents",
    points_selector=["doc-001", "doc-002"],
)

# Delete by filter
from qdrant_client.models import FilterSelector, FieldCondition, MatchValue
client.delete(
    collection_name="myapp_documents",
    points_selector=FilterSelector(filter=Filter(must=[
        FieldCondition(key="source", match=MatchValue(value="draft"))
    ])),
)

Node.js CRUD

import { QdrantClient } from '@qdrant/js-client-rest';
const client = new QdrantClient({ url: process.env.QDRANT_URL, apiKey: process.env.QDRANT_API_KEY });

await client.createCollection('myapp_documents', {
  vectors: { size: 1536, distance: 'Cosine' },
});

await client.upsert('myapp_documents', {
  points: [{ id: 'doc-001', vector: embedding, payload: { title: 'Hello' } }],
});

const results = await client.search('myapp_documents', {
  vector: queryEmbedding, limit: 5,
});

await client.delete('myapp_documents', { points: ['doc-001'] });

RAG Pipeline

def rag_query(question: str) -> str:
    query_vector = embed(question)
    hits = client.search(collection_name="myapp_documents", query_vector=query_vector, limit=5)
    context = "\n".join(hit.payload["text"] for hit in hits)
    return llm.generate(f"Context:\n{context}\n\nQ: {question}")

Admin — Collections & API Key

# List collections
curl -s -H "api-key: $KEY" http://db.bizfylabs.com:6333/collections

# Collection info
curl -s -H "api-key: $KEY" http://db.bizfylabs.com:6333/collections/myapp_documents

# Delete collection
curl -X DELETE -H "api-key: $KEY" http://db.bizfylabs.com:6333/collections/myapp_documents

# Rotate API key: update QDRANT_API_KEY in /srv/databases/.env, then:
cd /srv/databases && docker compose restart qdrant

Send to Developer

QDRANT_URL=http://db.bizfylabs.com:6333
QDRANT_API_KEY=YOUR_KEY
# Use collection prefix: myapp_documents