Getting Started

This guide walks you through building a simple document indexing and search application with Quarkus Qdrant.

Installation

Add the extension to your project:

<dependency>
    <groupId>io.quarkiverse.qdrant</groupId>
    <artifactId>quarkus-qdrant</artifactId>
    <version>999-SNAPSHOT</version>
</dependency>

The Document POJO

First, create a Document class:

package org.acme.qdrant;

public class Document {

    public String id;
    public String text;
    public String source;
}

Nothing fancy. Just a plain Java object with an ID, some text, and a source field.

The Document Service

Now create a DocumentService that uses QdrantClient to index, search, and delete documents:

package org.acme.qdrant;

import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.UUID;

import jakarta.enterprise.context.ApplicationScoped;
import jakarta.inject.Inject;

import io.quarkiverse.qdrant.runtime.QdrantClient;
import io.quarkiverse.qdrant.runtime.model.PointStruct;
import io.quarkiverse.qdrant.runtime.model.ScoredPoint;

@ApplicationScoped
public class DocumentService {

    private static final String COLLECTION = "documents";

    @Inject
    QdrantClient qdrant; (1)

    public String index(Document document) {
        if (document.id == null) {
            document.id = UUID.randomUUID().toString();
        }

        Map<String, Object> payload = new HashMap<>();
        payload.put("text", document.text);
        payload.put("source", document.source);

        List<Float> vector = computeVector(document.text); (2)

        PointStruct point = new PointStruct(document.id, vector, payload);
        qdrant.upsert(COLLECTION).point(point).execute(); (3)

        return document.id;
    }

    public List<Document> search(float[] queryVector) {
        List<ScoredPoint> results = qdrant.search(COLLECTION) (4)
                .vector(queryVector)
                .limit(10)
                .withPayload(true)
                .withVector(false)
                .execute();

        List<Document> documents = new ArrayList<>();
        if (results != null) {
            for (ScoredPoint point : results) {
                Document doc = new Document();
                doc.id = point.getId();
                if (point.getPayload() != null) {
                    doc.text = (String) point.getPayload().get("text");
                    doc.source = (String) point.getPayload().get("source");
                }
                documents.add(doc);
            }
        }
        return documents;
    }

    public void delete(String id) {
        qdrant.delete(COLLECTION).byIds(List.of(id)).execute(); (5)
    }

    List<Float> computeVector(String text) {
        // Use your embedding model here
    }
}

1	Inject `QdrantClient`.
2	Compute a vector from the text using your embedding model.
3	Upsert a point into the collection.
4	Search with a vector. Returns a list of `ScoredPoint` with id, score, and payload.
5	Delete points by their IDs.