Skip to main content

Google Cloud SQL for PostgreSQL

Cloud SQL is a fully managed relational database service that offers high performance, seamless integration, and impressive scalability. Cloud SQL is 100% compatible with PostgreSQL. Extend your database application to build AI-powered experiences leveraging Cloud SQL's Langchain integrations.

This module implements EmbeddingStore backed by an Cloud SQL for PostgreSQL database.

Before You Begin​

In order to use this library, you first need to go through the following steps:

  1. Select or create a Cloud Platform project.
  2. Enable billing for your project.
  3. Enable the CloudSQL API.
  4. Setup Authentication.

Maven Dependency​

<dependency>
    <groupId>dev.langchain4j</groupId>
    <artificatId>langchain4j-community-cloud-sql-pg</artificatId>
    <version>1.0.0-beta3</version>
</dependency>

PostgresEmbeddingStore Usage​

Use a vector store to store text embedded data and perform vector search. Instances of PostgresEmbeddingStore can be created by configuring provided Builder, it requires the following:

  • PostgresEngine instance
  • table name
  • schema name (optional, default: "public")
  • content column (optional, default: "content")
  • embedding column (optional, default: "embedding")
  • id column (optional, default: "langchain_id")
  • metadata column names (optional)
  • additional metadata json column (optional, default: "langchain_metadata")
  • ignored metadata column names (optional)
  • distance strategy (optional, default:DistanceStrategy.COSINE_DISTANCE)
  • query options (optional)

example usage:

import dev.langchain4j.data.document.Metadata;
import dev.langchain4j.data.embedding.Embedding;
import dev.langchain4j.data.segment.TextSegment;
import dev.langchain4j.model.embedding.onnx.allminilml6v2.AllMiniLmL6V2EmbeddingModel;
import dev.langchain4j.model.embedding.EmbeddingModel;
import dev.langchain4j.store.embedding.EmbeddingMatch;
import dev.langchain4j.store.embedding.EmbeddingSearchRequest;
import dev.langchain4j.store.embedding.EmbeddingSearchResult;
import dev.langchain4j.engine.EmbeddingStoreConfig;
import dev.langchain4j.engine.PostgresEngine;
import dev.langchain4j.engine.MetadataColumn;
import dev.langchain4j.store.embedding.cloudsql.PostgresEmbeddingStore;

import java.util.ArrayList;
import java.util.List;
import java.util.UUID;
import java.util.stream.Collectors;

PostgresEngine engine = new PostgresEngine.Builder()
    .projectId("")
    .region("")
    .instance("")
    .database("")
    .build();

PostgresEmbeddingStore store = new PostgresEmbeddingStore.Builder(engine, TABLE_NAME)
    .build();

List<String> testTexts = Arrays.asList("cat", "dog", "car", "truck");
List<Embedding> embeddings = new ArrayList<>();
List<TextSegment> textSegments = new ArrayList<>();
EmbeddingModel embeddingModel = new AllMiniLmL6V2EmbeddingModel();

for (String text : testTexts) {
    Map<String, Object> metaMap = new HashMap<>();
    metaMap.put("my_metadata", "string");
    Metadata metadata = new Metadata(metaMap);
    textSegments.add(new TextSegment(text, metadata));
    embeddings.add(MyEmbeddingModel.embed(text).content());
}
List<String> ids = store.addAll(embeddings, textSegments);
// search for "cat"
EmbeddingSearchRequest request = EmbeddingSearchRequest.builder()
        .queryEmbedding(embeddings.get(0))
        .maxResults(10)
        .minScore(0.9)
        .build();
List<EmbeddingMatch<TextSegment>> result = store.search(request).matches();
// remove "cat"
store.removeAll(singletonList(result.get(0).embeddingId()));