Support Azure OpenAI and AI Search

langchain4j-azure-ai-search-spring-boot-starter/pom.xml

@@ -0,0 +1,75 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<project xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
+ xmlns="http://maven.apache.org/POM/4.0.0"
+ xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
+ <modelVersion>4.0.0</modelVersion>
+
+ <parent>
+ <groupId>dev.langchain4j</groupId>
+ <artifactId>langchain4j-spring</artifactId>
+ <version>0.29.0</version>
+ <relativePath>../pom.xml</relativePath>
+ </parent>
+
+ <artifactId>langchain4j-azure-ai-search-spring-boot-starter</artifactId>
+ <name>LangChain4j Spring Boot starter for Azure AI Search</name>
+ <packaging>jar</packaging>
+
+ <dependencies>
+
+ <dependency>
+ <groupId>dev.langchain4j</groupId>
+ <artifactId>langchain4j-azure-ai-search</artifactId>
+ <version>${project.version}</version>
+ </dependency>
+
+ <dependency>
+ <groupId>org.springframework.boot</groupId>
+ <artifactId>spring-boot-starter</artifactId>
+ </dependency>
+
+ <dependency>
+ <groupId>org.springframework.boot</groupId>
+ <artifactId>spring-boot-autoconfigure-processor</artifactId>
+ <optional>true</optional>
+ </dependency>
+
+ <!-- should be listed before spring-boot-configuration-processor -->
+ <dependency>
+ <groupId>org.projectlombok</groupId>
+ <artifactId>lombok</artifactId>
+ <scope>provided</scope>
+ </dependency>
+
+ <!-- needed to generate automatic metadata about available config properties -->
+ <dependency>
+ <groupId>org.springframework.boot</groupId>
+ <artifactId>spring-boot-configuration-processor</artifactId>
+ <optional>true</optional>
+ </dependency>
+
+ <dependency>
+ <groupId>org.springframework.boot</groupId>
+ <artifactId>spring-boot-starter-test</artifactId>
+ <scope>test</scope>
+ </dependency>
+
+ <dependency>
+ <groupId>dev.langchain4j</groupId>
+ <artifactId>langchain4j-embeddings-all-minilm-l6-v2</artifactId>
+ <version>0.29.0</version>
+ <scope>test</scope>
+ </dependency>
+
+ </dependencies>
+
+ <licenses>
+ <license>
+ <name>Apache-2.0</name>
+ <url>https://www.apache.org/licenses/LICENSE-2.0.txt</url>
+ <distribution>repo</distribution>
+ <comments>A business-friendly OSS license</comments>
+ </license>
+ </licenses>
+
+</project>

langchain4j-azure-ai-search-spring-boot-starter/src/main/java/dev/langchain4j/azure/aisearch/spring/AutoConfig.java

@@ -0,0 +1,47 @@
+package dev.langchain4j.azure.aisearch.spring;
+
+import com.azure.search.documents.indexes.models.SearchIndex;
+import dev.langchain4j.model.embedding.EmbeddingModel;
+import dev.langchain4j.rag.content.retriever.azure.search.AzureAiSearchContentRetriever;
+import dev.langchain4j.store.embedding.azure.search.AzureAiSearchEmbeddingStore;
+import org.springframework.boot.autoconfigure.AutoConfiguration;
+import org.springframework.boot.autoconfigure.condition.ConditionalOnProperty;
+import org.springframework.boot.context.properties.EnableConfigurationProperties;
+import org.springframework.context.annotation.Bean;
+import org.springframework.lang.Nullable;
+
+import static dev.langchain4j.azure.aisearch.spring.Properties.PREFIX;
+
+@AutoConfiguration
+@EnableConfigurationProperties(Properties.class)
+public class AutoConfig {
+ @Bean
+ @ConditionalOnProperty(PREFIX + ".content-retriever.api-key")
+ public AzureAiSearchContentRetriever azureAiSearchContentRetriever(Properties properties, @Nullable EmbeddingModel embeddingModel, @Nullable SearchIndex index) {
+ Properties.NestedProperties nestedProperties = properties.getContentRetriever();
+ return AzureAiSearchContentRetriever.builder()
+ .endpoint(nestedProperties.getEndpoint())
+ .apiKey(nestedProperties.getApiKey())
+ .createOrUpdateIndex(nestedProperties.getCreateOrUpdateIndex())
+ .embeddingModel(embeddingModel)
+ .dimensions(nestedProperties.getDimensions() == null ? 0 : nestedProperties.getDimensions())
+ .index(index)
+ .maxResults(nestedProperties.getMaxResults())
+ .minScore(nestedProperties.getMinScore() == null ? 0.0 : nestedProperties.getMinScore())
+ .queryType(nestedProperties.getQueryType())
+ .build();
+ }
+
+ @Bean
+ @ConditionalOnProperty(PREFIX + ".embedding-store.api-key")
+ public AzureAiSearchEmbeddingStore azureAiSearchEmbeddingStore(Properties properties, @Nullable EmbeddingModel embeddingModel, @Nullable SearchIndex index) {
+ Properties.NestedProperties nestedProperties = properties.getEmbeddingStore();
+ return AzureAiSearchEmbeddingStore.builder()
+ .endpoint(nestedProperties.getEndpoint())
+ .apiKey(nestedProperties.getApiKey())
+ .createOrUpdateIndex(nestedProperties.getCreateOrUpdateIndex())
+ .dimensions(nestedProperties.getDimensions())
+ .index(index)
+ .build();
+ }
+}

langchain4j-azure-ai-search-spring-boot-starter/src/main/java/dev/langchain4j/azure/aisearch/spring/Properties.java

@@ -0,0 +1,34 @@
+package dev.langchain4j.azure.aisearch.spring;
+
+import dev.langchain4j.rag.content.retriever.azure.search.AzureAiSearchQueryType;
+import lombok.Getter;
+import lombok.Setter;
+import org.springframework.boot.context.properties.ConfigurationProperties;
+import org.springframework.boot.context.properties.NestedConfigurationProperty;
+
+@Getter
+@Setter
+@ConfigurationProperties(prefix = Properties.PREFIX)
+public class Properties {
+
+ static final String PREFIX = "langchain4j.azure.ai-search";
+
+ @NestedConfigurationProperty
+ NestedProperties contentRetriever;
+
+ @NestedConfigurationProperty
+ NestedProperties embeddingStore;
+
+ @Getter
+ @Setter
+ public static class NestedProperties {
+ String endpoint;
+ String apiKey;
+ Integer dimensions;
+ Boolean createOrUpdateIndex;
+ String indexName;
+ Integer maxResults = 3;
+ Double minScore;
+ AzureAiSearchQueryType queryType;
+ }
+}

langchain4j-azure-ai-search-spring-boot-starter/src/main/resources/META-INF/spring.factories

@@ -0,0 +1,2 @@
+org.springframework.boot.autoconfigure.EnableAutoConfiguration=\
+dev.langchain4j.azure.aisearch.spring.AutoConfig

langchain4j-azure-ai-search-spring-boot-starter/src/main/resources/META-INF/spring/org.springframework.boot.autoconfigure.AutoConfiguration.imports

@@ -0,0 +1 @@
+dev.langchain4j.azure.aisearch.spring.AutoConfig

langchain4j-azure-ai-search-spring-boot-starter/src/test/java/dev/langchain4j/azure/aisearch/spring/AutoConfigIT.java

@@ -0,0 +1,215 @@
+package dev.langchain4j.azure.aisearch.spring;
+
+import com.azure.core.credential.AzureKeyCredential;
+import com.azure.search.documents.indexes.SearchIndexClient;
+import com.azure.search.documents.indexes.SearchIndexClientBuilder;
+import com.azure.search.documents.indexes.models.SearchIndex;
+import dev.langchain4j.data.embedding.Embedding;
+import dev.langchain4j.data.segment.TextSegment;
+import dev.langchain4j.model.embedding.AllMiniLmL6V2EmbeddingModel;
+import dev.langchain4j.model.embedding.EmbeddingModel;
+import dev.langchain4j.rag.content.Content;
+import dev.langchain4j.rag.content.retriever.ContentRetriever;
+import dev.langchain4j.rag.content.retriever.azure.search.AzureAiSearchContentRetriever;
+import dev.langchain4j.rag.content.retriever.azure.search.AzureAiSearchQueryType;
+import dev.langchain4j.rag.query.Query;
+import dev.langchain4j.store.embedding.EmbeddingMatch;
+import dev.langchain4j.store.embedding.EmbeddingStore;
+import dev.langchain4j.store.embedding.azure.search.AzureAiSearchEmbeddingStore;
+import org.junit.jupiter.api.Test;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+import org.springframework.boot.autoconfigure.AutoConfigurations;
+import org.springframework.boot.test.context.runner.ApplicationContextRunner;
+
+import java.util.List;
+
+import static dev.langchain4j.store.embedding.azure.search.AbstractAzureAiSearchEmbeddingStore.INDEX_NAME;
+import static java.util.Arrays.asList;
+import static org.assertj.core.api.Assertions.assertThat;
+
+class AutoConfigIT {
+
+ private static final String AZURE_SEARCH_KEY = System.getenv("AZURE_SEARCH_KEY");
+ private static final String AZURE_SEARCH_ENDPOINT = System.getenv("AZURE_SEARCH_ENDPOINT");
+ ApplicationContextRunner contextRunner = new ApplicationContextRunner()
+ .withConfiguration(AutoConfigurations.of(AutoConfig.class));
+
+ private static final Logger log = LoggerFactory.getLogger(AutoConfigIT.class);
+
+ private final EmbeddingModel embeddingModel = new AllMiniLmL6V2EmbeddingModel();
+ private final int dimensions = embeddingModel.embed("test").content().vector().length;
+
+ private final SearchIndexClient searchIndexClient = new SearchIndexClientBuilder()
+ .endpoint(System.getenv("AZURE_SEARCH_ENDPOINT"))
+ .credential(new AzureKeyCredential(System.getenv("AZURE_SEARCH_KEY")))
+ .buildClient();
+
+ private SearchIndex index = new SearchIndex(INDEX_NAME);
+
+ @Test
+ void should_provide_ai_search_retriever() {
+
+ searchIndexClient.deleteIndex(INDEX_NAME);
+
+ contextRunner
+ .withPropertyValues(
+ Properties.PREFIX + ".content-retriever.api-key=" + AZURE_SEARCH_KEY,
+ Properties.PREFIX + ".content-retriever.endpoint=" + AZURE_SEARCH_ENDPOINT,
+ Properties.PREFIX + ".content-retriever.dimensions=" + dimensions,
+ Properties.PREFIX + ".content-retriever.create-or-update-index=" + "true",
+ Properties.PREFIX + ".content-retriever.query-type=" + "VECTOR"
+ ).withBean(EmbeddingModel.class, () -> embeddingModel)
+ .run(context -> {
+ ContentRetriever contentRetriever = context.getBean(ContentRetriever.class);
+ assertThat(contentRetriever).isInstanceOf(AzureAiSearchContentRetriever.class);
+ AzureAiSearchContentRetriever azureAiSearchContentRetriever = (AzureAiSearchContentRetriever) contentRetriever;
+
+ String content1 = "This book is about politics";
+ String content2 = "Cats sleeps a lot.";
+ String content3 = "Sandwiches taste good.";
+ String content4 = "The house is open";
+ List<String> contents = asList(content1, content2, content3, content4);
+
+ for (String content : contents) {
+ TextSegment textSegment = TextSegment.from(content);
+ Embedding embedding = embeddingModel.embed(content).content();
+ azureAiSearchContentRetriever.add(embedding, textSegment);
+ }
+
+ awaitUntilPersisted();
+ });
+
+ String content = "house";
+ Query query = Query.from(content);
+
+ contextRunner
+ .withPropertyValues(
+ Properties.PREFIX + ".content-retriever.api-key=" + AZURE_SEARCH_KEY,
+ Properties.PREFIX + ".content-retriever.endpoint=" + AZURE_SEARCH_ENDPOINT,
+ Properties.PREFIX + ".content-retriever.create-or-update-index=" + "false",
+ Properties.PREFIX + ".content-retriever.max-results=" + "3",
+ Properties.PREFIX + ".content-retriever.min-score=" + "0.6",
+ Properties.PREFIX + ".content-retriever.query-type=" + AzureAiSearchQueryType.VECTOR
+ ).withBean(SearchIndex.class, () -> index)
+ .withBean(EmbeddingModel.class, () -> embeddingModel)
+ .run(context -> {
+ ContentRetriever contentRetriever = context.getBean(ContentRetriever.class);
+ assertThat(contentRetriever).isInstanceOf(AzureAiSearchContentRetriever.class);
+ AzureAiSearchContentRetriever contentRetrieverWithVector = (AzureAiSearchContentRetriever) contentRetriever;
+ log.info("Testing Vector Search");
+ List<Content> relevant = contentRetrieverWithVector.retrieve(query);
+ assertThat(relevant).hasSizeGreaterThan(0);
+ assertThat(relevant.get(0).textSegment().text()).isEqualTo("The house is open");
+ log.info("#1 relevant item: {}", relevant.get(0).textSegment().text());
+ });
+
+ contextRunner
+ .withPropertyValues(
+ Properties.PREFIX + ".content-retriever.api-key=" + AZURE_SEARCH_KEY,
+ Properties.PREFIX + ".content-retriever.endpoint=" + AZURE_SEARCH_ENDPOINT,
+ Properties.PREFIX + ".content-retriever.create-or-update-index=" + "false",
+ Properties.PREFIX + ".content-retriever.query-type=" + AzureAiSearchQueryType.FULL_TEXT
+ )
+ .run(context -> {
+ ContentRetriever contentRetriever = context.getBean(ContentRetriever.class);
+ assertThat(contentRetriever).isInstanceOf(AzureAiSearchContentRetriever.class);
+ AzureAiSearchContentRetriever contentRetrieverWithFullText = (AzureAiSearchContentRetriever) contentRetriever;
+ log.info("Testing Full Text Search");
+ // This uses the same storage as the vector search, so we don't need to add the content again
+ List<Content> relevant2 = contentRetrieverWithFullText.retrieve(query);
+ assertThat(relevant2).hasSizeGreaterThan(0);
+ assertThat(relevant2.get(0).textSegment().text()).isEqualTo("The house is open");
+ log.info("#1 relevant item: {}", relevant2.get(0).textSegment().text());
+ });
+
+ contextRunner
+ .withPropertyValues(
+ Properties.PREFIX + ".content-retriever.api-key=" + AZURE_SEARCH_KEY,
+ Properties.PREFIX + ".content-retriever.endpoint=" + AZURE_SEARCH_ENDPOINT,
+ Properties.PREFIX + ".content-retriever.create-or-update-index=" + "false",
+ Properties.PREFIX + ".content-retriever.query-type=" + AzureAiSearchQueryType.HYBRID
+ ).withBean(SearchIndex.class, () -> index)
+ .withBean(EmbeddingModel.class, () -> embeddingModel)
+ .run(context -> {
+ ContentRetriever contentRetriever = context.getBean(ContentRetriever.class);
+ assertThat(contentRetriever).isInstanceOf(AzureAiSearchContentRetriever.class);
+ AzureAiSearchContentRetriever contentRetrieverWithHybrid = (AzureAiSearchContentRetriever) contentRetriever;
+ log.info("Testing Hybrid Search");
+ List<Content> relevant3 = contentRetrieverWithHybrid.retrieve(query);
+ assertThat(relevant3).hasSizeGreaterThan(0);
+ assertThat(relevant3.get(0).textSegment().text()).isEqualTo("The house is open");
+ log.info("#1 relevant item: {}", relevant3.get(0).textSegment().text());
+ });
+
+ contextRunner
+ .withPropertyValues(
+ Properties.PREFIX + ".content-retriever.api-key=" + AZURE_SEARCH_KEY,
+ Properties.PREFIX + ".content-retriever.endpoint=" + AZURE_SEARCH_ENDPOINT,
+ Properties.PREFIX + ".content-retriever.create-or-update-index=" + "false",
+ Properties.PREFIX + ".content-retriever.max-results=" + "3",
+ Properties.PREFIX + ".content-retriever.min-score=" + "0.4",
+ Properties.PREFIX + ".content-retriever.query-type=" + AzureAiSearchQueryType.HYBRID_WITH_RERANKING
+ ).withBean(SearchIndex.class, () -> index)
+ .withBean(EmbeddingModel.class, () -> embeddingModel)
+ .run(context -> {
+ ContentRetriever contentRetriever = context.getBean(ContentRetriever.class);
+ assertThat(contentRetriever).isInstanceOf(AzureAiSearchContentRetriever.class);
+ AzureAiSearchContentRetriever contentRetrieverWithHybridAndReranking = (AzureAiSearchContentRetriever) contentRetriever;
+ log.info("Testing Hybrid Search with Reranking");
+ List<Content> relevant4 = contentRetrieverWithHybridAndReranking.retrieve(query);
+ assertThat(relevant4).hasSizeGreaterThan(0);
+ assertThat(relevant4.get(0).textSegment().text()).isEqualTo("The house is open");
+ log.info("#1 relevant item: {}", relevant4.get(0).textSegment().text());
+ });
+ }
+
+ protected void awaitUntilPersisted() {
+ try {
+ Thread.sleep(1_000);
+ } catch (InterruptedException e) {
+ throw new RuntimeException(e);
+ }
+ }
+
+
+ @Test
+ void should_provide_ai_search_embedding_store() {
+
+ searchIndexClient.deleteIndex(INDEX_NAME);
+
+ contextRunner
+ .withPropertyValues(
+ Properties.PREFIX + ".embedding-store.api-key=" + AZURE_SEARCH_KEY,
+ Properties.PREFIX + ".embedding-store.endpoint=" + AZURE_SEARCH_ENDPOINT,
+ Properties.PREFIX + ".embedding-store.dimensions=" + 384,
+ Properties.PREFIX + ".embedding-store.create-or-update-index=" + "true"
+ ).withBean(EmbeddingModel.class, () -> embeddingModel)
+ .run(context -> {
+ EmbeddingStore embeddingStore = context.getBean(EmbeddingStore.class);
+ assertThat(embeddingStore).isInstanceOf(AzureAiSearchEmbeddingStore.class);
+ assertThat(context.getBean(AzureAiSearchEmbeddingStore.class)).isSameAs(embeddingStore);
+
+
+ String content1 = "banana";
+ String content2 = "computer";
+ String content3 = "apple";
+ String content4 = "pizza";
+ String content5 = "strawberry";
+ String content6 = "chess";
+ List<String> contents = asList(content1, content2, content3, content4, content5, content6);
+
+ for (String content : contents) {
+ TextSegment textSegment = TextSegment.from(content);
+ Embedding embedding = embeddingModel.embed(content).content();
+ embeddingStore.add(embedding, textSegment);
+ }
+ Embedding relevantEmbedding = embeddingModel.embed("fruit").content();
+ List<EmbeddingMatch<TextSegment>> relevant = embeddingStore.findRelevant(relevantEmbedding, 3);
+ assertThat(relevant).hasSize(3);
+ assertThat(relevant.get(0).embedding()).isNotNull();
+ assertThat(relevant.get(0).embedded().text()).isIn(content1, content3, content5);
+ });
+ }
+
+}