Add filter-based deletion to Weaviate vector store (#2150)

Implement filter-based deletion for WeaviateVectorStore by leveraging the
existing ID-based deletion mechanism.

Key changes:

- Add doDelete(Filter.Expression) implementation that delegates to the existing
  ID-based deletion method
- Use similarity search with filter to find matching documents first
- Delete matched documents by their IDs
- Add integration tests for filter deletion scenarios:
  * Deleting by simple equality filter
  * Deleting by string-based filter expression

This maintains consistency with other vector store implementations while
working around Weaviate's limitations in direct filtered deletion.

Author: sobychacko

vector-stores/spring-ai-weaviate-store/src/main/java/org/springframework/ai/vectorstore/weaviate/WeaviateVectorStore.java

@@ -43,20 +43,20 @@
 import io.weaviate.client.v1.graphql.query.builder.GetBuilder.GetBuilderBuilder;
 import io.weaviate.client.v1.graphql.query.fields.Field;
 import io.weaviate.client.v1.graphql.query.fields.Fields;
+import org.apache.commons.logging.LogFactory;
 
 import org.springframework.ai.document.Document;
 import org.springframework.ai.document.DocumentMetadata;
-import org.springframework.ai.embedding.BatchingStrategy;
 import org.springframework.ai.embedding.EmbeddingModel;
 import org.springframework.ai.embedding.EmbeddingOptionsBuilder;
-import org.springframework.ai.embedding.TokenCountBatchingStrategy;
 import org.springframework.ai.model.EmbeddingUtils;
 import org.springframework.ai.observation.conventions.VectorStoreProvider;
 import org.springframework.ai.vectorstore.AbstractVectorStoreBuilder;
 import org.springframework.ai.vectorstore.SearchRequest;
 import org.springframework.ai.vectorstore.filter.Filter;
 import org.springframework.ai.vectorstore.observation.AbstractObservationVectorStore;
 import org.springframework.ai.vectorstore.observation.VectorStoreObservationContext;
+import org.springframework.core.log.LogAccessor;
 import org.springframework.util.Assert;
 import org.springframework.util.CollectionUtils;
 import org.springframework.util.StringUtils;
@@ -93,6 +93,8 @@
  */
 public class WeaviateVectorStore extends AbstractObservationVectorStore {
 
+	private static final LogAccessor logger = new LogAccessor(LogFactory.getLog(WeaviateVectorStore.class));
+
 	private static final String METADATA_FIELD_PREFIX = "meta_";
 
 	private static final String CONTENT_FIELD_NAME = "content";
@@ -296,6 +298,43 @@ public Optional<Boolean> doDelete(List<String> documentIds) {
 		return Optional.of(!result.hasErrors());
 	}
 
+	@Override
+	protected void doDelete(Filter.Expression filterExpression) {
+		Assert.notNull(filterExpression, "Filter expression must not be null");
+
+		try {
+			// Use similarity search with empty query to find documents matching the
+			// filter
+			SearchRequest searchRequest = SearchRequest.builder()
+				.query("") // empty query since we only want filter matches
+				.filterExpression(filterExpression)
+				.topK(10000) // large enough to get all matches
+				.similarityThresholdAll()
+				.build();
+
+			List<Document> matchingDocs = similaritySearch(searchRequest);
+
+			if (!matchingDocs.isEmpty()) {
+				List<String> idsToDelete = matchingDocs.stream().map(Document::getId).collect(Collectors.toList());
+
+				Optional<Boolean> result = delete(idsToDelete);
+
+				if (result.isPresent() && !result.get()) {
+					throw new IllegalStateException("Failed to delete some documents");
+				}
+
+				logger.debug(() -> "Deleted " + idsToDelete.size() + " documents matching filter expression");
+			}
+			else {
+				logger.debug(() -> "No documents found matching filter expression");
+			}
+		}
+		catch (Exception e) {
+			logger.error(e, () -> "Failed to delete documents by filter");
+			throw new IllegalStateException("Failed to delete documents by filter", e);
+		}
+	}
+
 	@Override
 	public List<Document> doSimilaritySearch(SearchRequest request) {
 

vector-stores/spring-ai-weaviate-store/src/test/java/org/springframework/ai/vectorstore/weaviate/WeaviateVectorStoreIT.java

@@ -1,5 +1,5 @@
 /*
- * Copyright 2023-2024 the original author or authors.
+ * Copyright 2023-2025 the original author or authors.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -37,6 +37,7 @@
 import org.springframework.ai.transformers.TransformersEmbeddingModel;
 import org.springframework.ai.vectorstore.SearchRequest;
 import org.springframework.ai.vectorstore.VectorStore;
+import org.springframework.ai.vectorstore.filter.Filter;
 import org.springframework.boot.SpringBootConfiguration;
 import org.springframework.boot.autoconfigure.EnableAutoConfiguration;
 import org.springframework.boot.test.context.runner.ApplicationContextRunner;
@@ -254,6 +255,62 @@ public void searchWithThreshold() {
 		});
 	}
 
+	@Test
+	void deleteByFilter() {
+		this.contextRunner.run(context -> {
+			VectorStore vectorStore = context.getBean(VectorStore.class);
+
+			var bgDocument = new Document("The World is Big and Salvation Lurks Around the Corner",
+					Map.of("country", "BG", "year", 2020));
+			var nlDocument = new Document("The World is Big and Salvation Lurks Around the Corner",
+					Map.of("country", "NL"));
+			var bgDocument2 = new Document("The World is Big and Salvation Lurks Around the Corner",
+					Map.of("country", "BG", "year", 2023));
+
+			vectorStore.add(List.of(bgDocument, nlDocument, bgDocument2));
+
+			Filter.Expression filterExpression = new Filter.Expression(Filter.ExpressionType.EQ,
+					new Filter.Key("country"), new Filter.Value("BG"));
+
+			vectorStore.delete(filterExpression);
+
+			List<Document> results = vectorStore
+				.similaritySearch(SearchRequest.builder().query("The World").topK(5).similarityThresholdAll().build());
+
+			assertThat(results).hasSize(1);
+			assertThat(results.get(0).getMetadata()).containsEntry("country", "NL");
+
+			vectorStore.delete(List.of(nlDocument.getId()));
+
+		});
+	}
+
+	@Test
+	void deleteWithStringFilterExpression() {
+		this.contextRunner.run(context -> {
+			VectorStore vectorStore = context.getBean(VectorStore.class);
+
+			var bgDocument = new Document("The World is Big and Salvation Lurks Around the Corner",
+					Map.of("country", "BG", "year", 2020));
+			var nlDocument = new Document("The World is Big and Salvation Lurks Around the Corner",
+					Map.of("country", "NL"));
+			var bgDocument2 = new Document("The World is Big and Salvation Lurks Around the Corner",
+					Map.of("country", "BG", "year", 2023));
+
+			vectorStore.add(List.of(bgDocument, nlDocument, bgDocument2));
+
+			vectorStore.delete("country == 'BG'");
+
+			List<Document> results = vectorStore
+				.similaritySearch(SearchRequest.builder().query("The World").topK(5).similarityThresholdAll().build());
+
+			assertThat(results).hasSize(1);
+			assertThat(results.get(0).getMetadata()).containsEntry("country", "NL");
+
+			vectorStore.delete(List.of(nlDocument.getId()));
+		});
+	}
+
 	@SpringBootConfiguration
 	@EnableAutoConfiguration
 	public static class TestApplication {