Qdrant

The Qdrant plugin registers retrievers and indexers backed by a Qdrant server (REST API) for Retrieval-Augmented Generation.

Installation

<dependency>
    <groupId>com.google.genkit</groupId>
    <artifactId>genkit-plugin-qdrant</artifactId>
    <version>1.0.0-SNAPSHOT</version>
</dependency>

Requirements

• A running Qdrant server (docker run -p 6333:6333 qdrant/qdrant)
• Java 21+
• An embedder (e.g. from the Google GenAI plugin)

Usage

QdrantPlugin registers a retriever and indexer named qdrant/<collectionName> for each configured collection.

import com.google.genkit.plugins.qdrant.QdrantCollectionConfig;
import com.google.genkit.plugins.qdrant.QdrantPlugin;

Genkit genkit = Genkit.builder()
    .plugin(GoogleGenAIPlugin.create(apiKey))
    .plugin(
        QdrantPlugin.builder()
            .url("http://localhost:6333")   // default
            .apiKey(System.getenv("QDRANT_API_KEY")) // optional; required for Qdrant Cloud
            .addCollection(
                QdrantCollectionConfig.builder()
                    .collectionName("films")
                    .embedderName("googleai/gemini-embedding-001")
                    .dimension(768)
                    .distance(QdrantCollectionConfig.Distance.COSINE)
                    .createCollectionIfNotExists(true) // default
                    .build())
            .build())
    .build();

// Index and retrieve
genkit.index("qdrant/films", documents);
List<Document> results = genkit.retrieve("qdrant/films", "a Christopher Nolan sci-fi film");

Configuration

Tune per-collection settings with QdrantCollectionConfig:

• collectionName — the Qdrant collection name (required)
• embedderName — the embedder used to vectorize documents and queries (required)
• dimension — the embedding dimension used when creating the collection (default 768). The store also probes the embedder on first use and creates the collection with the model’s actual output dimension, so this is only a fallback.
• distance — COSINE (default), EUCLIDEAN, or DOT_PRODUCT
• textPayloadKey — the payload key that stores the document text (default text)
• createCollectionIfNotExists — auto-create the collection (default true)
• addAdditionalMetadata(key, value) — metadata merged into every indexed point’s payload

Document ids that are not integers or UUIDs are hashed into a stable UUID (Qdrant point ids must be an unsigned integer or a UUID).

Sample

See the qdrant sample.