Qdrant Integration with Graphbit¶

Overview¶

This guide explains how to connect Qdrant, an open-source vector database, to Graphbit. With this integration, you can store, index, and search high-dimensional embeddings generated by LLMs for semantic search, retrieval-augmented generation, and more.

Prerequisites¶

Qdrant running locally or remotely (see Qdrant documentation).
OpenAI API Key: For embedding generation (or another supported embedding provider).
Graphbit installed and configured (see installation guide).
Python environment with qdrant-client, graphbit, and optionally python-dotenv installed.
.env file in your project root with the following variables:
```
OPENAI_API_KEY=your_openai_api_key_here
```

Step 1: Connect to Qdrant and Create Collection¶

Set up the Qdrant client and ensure the collection is created:

import os
import uuid
from qdrant_client import QdrantClient
from qdrant_client.models import VectorParams, Distance, PointStruct

COLLECTION = "graphbit-vector"
DIMENSION = 1536

client = QdrantClient(host="localhost", port=6333)

if not client.collection_exists(COLLECTION):
    client.create_collection(
        collection_name=COLLECTION,
        vectors_config=VectorParams(size=DIMENSION, distance=Distance.COSINE),
    )

Step 2: Generate and Upsert Embeddings using Graphbit¶

Use Graphbit's embedding client to generate embeddings and upsert them into Qdrant:

from graphbit import EmbeddingClient, EmbeddingConfig

OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
EMBEDDING_MODEL = "text-embedding-3-small"
embedding_client = EmbeddingClient(
    EmbeddingConfig.openai(model=EMBEDDING_MODEL, api_key=OPENAI_API_KEY)
)

texts = [
    "GraphBit is a framework for LLM workflows and agent orchestration.",
    "Qdrant is an open-source vector database for similarity search.",
    "OpenAI offers tools for LLMs and embeddings."
]
embeds = embedding_client.embed_many(texts)

points = [
    PointStruct(id=str(uuid.uuid4()), vector=vec, payload={"text": txt})
    for vec, txt in zip(embeds, texts)
]
client.upsert(collection_name=COLLECTION, points=points, wait=True)

Step 3: Perform Similarity Search¶

Embed your query and search for similar vectors in Qdrant:

query = "What is GraphBit?"
query_vec = embedding_client.embed(query)
response = client.query_points(
    collection_name=COLLECTION,
    query=query_vec,
    limit=2,
    with_payload=True
)

Full Example¶

import os
import uuid
from qdrant_client import QdrantClient
from qdrant_client.models import VectorParams, Distance, PointStruct
from graphbit import EmbeddingClient, EmbeddingConfig

COLLECTION = "graphbit-vector"
DIMENSION = 1536

client = QdrantClient(host="localhost", port=6333)
if not client.collection_exists(COLLECTION):
    client.create_collection(
        collection_name=COLLECTION,
        vectors_config=VectorParams(size=DIMENSION, distance=Distance.COSINE),
    )

OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
EMBEDDING_MODEL = "text-embedding-3-small"
embedding_client = EmbeddingClient(
    EmbeddingConfig.openai(model=EMBEDDING_MODEL, api_key=OPENAI_API_KEY)
)

texts = [
    "GraphBit is a framework for LLM workflows and agent orchestration.",
    "Qdrant is an open-source vector database for similarity search.",
    "OpenAI offers tools for LLMs and embeddings."
]
embeds = embedding_client.embed_many(texts)

points = [
    PointStruct(id=str(uuid.uuid4()), vector=vec, payload={"text": txt})
    for vec, txt in zip(embeds, texts)
]
client.upsert(collection_name=COLLECTION, points=points, wait=True)

# Retrieve one point to double-check (optional)
check_id = points[0].id
retrieved = client.retrieve(
    collection_name=COLLECTION,
    ids=[check_id],
    with_payload=True
)
assert retrieved, "Upsert failed!"

query = "What is GraphBit?"
query_vec = embedding_client.embed(query)
response = client.query_points(
    collection_name=COLLECTION,
    query=query_vec,
    limit=2,
    with_payload=True
)

# Process response into a more readable format
response_points = response.points
result = []
for point in response_points:
    result.append({
        "id": point.id,
        "text": point.payload.get("text", ""),
        "score": point.score
    })
print(result)

This integration enables you to leverage Graphbit's embedding capabilities with Qdrant's vector database for scalable, open-source semantic search and retrieval workflows.