Fix langchain4j#1747 (langchain4j#1949)

## Issue
Fixes langchain4j#1747

## Change
Fixed this bug for Milvus and Chroma.

Did not manage to fix it for Pinecone because Pinecone can store
`Integer`/`Long` only as `Double`. We could theoretically store them as
`String`s, but then metadata filtering for `Integer`/`Long` values will
not work. For now I've left the comment in the `PineconeEmbeddingStore`
Javadoc and Pinecone documentation about the bug and possible
workaround.

## General checklist
- [X] There are no breaking changes
- [X] I have added unit and integration tests for my change
- [x] I have manually run all the unit and integration tests in the
module I have added/changed, and they are all green
- [x] I have manually run all the unit and integration tests in the
[core](https://github.com/langchain4j/langchain4j/tree/main/langchain4j-core)
and
[main](https://github.com/langchain4j/langchain4j/tree/main/langchain4j)
modules, and they are all green
- [X] I have added/updated the
[documentation](https://github.com/langchain4j/langchain4j/tree/main/docs/docs)
- [ ] I have added an example in the [examples
repo](https://github.com/langchain4j/langchain4j-examples) (only for
"big" features)
- [ ] I have added/updated [Spring Boot
starter(s)](https://github.com/langchain4j/langchain4j-spring) (if
applicable)

docs/docs/integrations/embedding-stores/pinecone.md

@@ -17,6 +17,14 @@ https://www.pinecone.io/
 </dependency>
 ```
 
+## Known Issues
+
+- https://github.com/langchain4j/langchain4j/issues/1948
+Pinecone stores all numbers as [floating-point values](https://docs.pinecone.io/guides/data/filter-with-metadata#supported-metadata-types),
+which means `Integer` and `Long` values (e.g., 1746714878034235396) stored in `Metadata`
+may be corrupted and returned as incorrect numbers!
+Possible workaround: convert integer/double values to `String` before storing them in `Metadata`.
+Please note that in this case metadata filtering might not work properly!
 
 ## APIs
 

langchain4j-chroma/src/main/java/dev/langchain4j/store/embedding/chroma/ChromaClient.java

@@ -1,6 +1,7 @@
 package dev.langchain4j.store.embedding.chroma;
 
 import static com.google.gson.FieldNamingPolicy.LOWER_CASE_WITH_UNDERSCORES;
+import static com.google.gson.ToNumberPolicy.LONG_OR_DOUBLE;
 
 import com.google.gson.Gson;
 import com.google.gson.GsonBuilder;
@@ -32,7 +33,10 @@ private ChromaClient(Builder builder) {
             httpClientBuilder.addInterceptor(new ChromaResponseLoggingInterceptor());
         }
 
-        Gson gson = new GsonBuilder().setFieldNamingPolicy(LOWER_CASE_WITH_UNDERSCORES).create();
+        Gson gson = new GsonBuilder()
+                .setFieldNamingPolicy(LOWER_CASE_WITH_UNDERSCORES)
+                .setObjectToNumberStrategy(LONG_OR_DOUBLE)
+                .create();
 
         Retrofit retrofit = new Retrofit.Builder()
             .baseUrl(Utils.ensureTrailingForwardSlash(builder.baseUrl))

langchain4j-core/src/test/java/dev/langchain4j/store/embedding/EmbeddingStoreIT.java

@@ -63,6 +63,9 @@ void should_add_embedding_with_segment_with_metadata() {
         assertThat(match.embedded().metadata().getLong("long_minus_1")).isEqualTo(-1L);
         assertThat(match.embedded().metadata().getLong("long_0")).isEqualTo(0L);
         assertThat(match.embedded().metadata().getLong("long_1")).isEqualTo(1L);
+        if (testLong1746714878034235396()) {
+            assertThat(match.embedded().metadata().getLong("long_1746714878034235396")).isEqualTo(1746714878034235396L);
+        }
         assertThat(match.embedded().metadata().getLong("long_max")).isEqualTo(Long.MAX_VALUE);
 
         assertThat(match.embedded().metadata().getFloat("float_min")).isEqualTo(-Float.MAX_VALUE);
@@ -84,6 +87,10 @@ void should_add_embedding_with_segment_with_metadata() {
                 .build()).matches()).isEqualTo(relevant);
     }
 
+    protected boolean testLong1746714878034235396() {
+        return true;
+    }
+
     protected Metadata createMetadata() {
 
         Metadata metadata = new Metadata();
@@ -104,6 +111,9 @@ protected Metadata createMetadata() {
         metadata.put("long_minus_1", -1L);
         metadata.put("long_0", 0L);
         metadata.put("long_1", 1L);
+        if (testLong1746714878034235396()) {
+            metadata.put("long_1746714878034235396", 1746714878034235396L);
+        }
         metadata.put("long_max", Long.MAX_VALUE);
 
         metadata.put("float_min", -Float.MAX_VALUE);

langchain4j-milvus/src/main/java/dev/langchain4j/store/embedding/milvus/Mapper.java

@@ -1,6 +1,7 @@
 package dev.langchain4j.store.embedding.milvus;
 
 import com.google.gson.Gson;
+import com.google.gson.GsonBuilder;
 import com.google.gson.JsonObject;
 import com.google.gson.reflect.TypeToken;
 import dev.langchain4j.data.document.Metadata;
@@ -22,6 +23,7 @@
 import java.util.List;
 import java.util.Map;
 
+import static com.google.gson.ToNumberPolicy.LONG_OR_DOUBLE;
 import static dev.langchain4j.internal.Utils.isNullOrBlank;
 import static dev.langchain4j.internal.Utils.isNullOrEmpty;
 import static dev.langchain4j.store.embedding.milvus.CollectionOperationsExecutor.queryForVectors;
@@ -31,7 +33,10 @@
 
 class Mapper {
 
-    private static final Gson GSON = new Gson();
+    private static final Gson GSON = new GsonBuilder()
+            .setObjectToNumberStrategy(LONG_OR_DOUBLE)
+            .create();
+
     private static final Type MAP_TYPE = new TypeToken<Map<String, Object>>() {
     }.getType();
 

langchain4j-pgvector/src/test/java/dev/langchain4j/store/embedding/pgvector/PgVectorEmbeddingStoreWithColumnsFilteringIT.java

@@ -16,7 +16,7 @@ static void beforeAll() {
                         Arrays.asList("key text NULL", "name text NULL", "age float NULL", "city varchar null", "country varchar null",
                                 "string_empty varchar null", "string_space varchar null", "string_abc varchar null", "uuid uuid null",
                                 "integer_min int null", "integer_minus_1 int null", "integer_0 int null", "integer_1 int null", "integer_max int null",
-                                "long_min bigint null", "long_minus_1 bigint null", "long_0 bigint null", "long_1 bigint null", "long_max bigint null",
+                                "long_min bigint null", "long_minus_1 bigint null", "long_0 bigint null", "long_1 bigint null", "long_1746714878034235396 bigint null", "long_max bigint null",
                                 "float_min float null", "float_minus_1 float null", "float_0 float null", "float_1 float null", "float_123 float null", "float_max float null",
                                 "double_minus_1 float8 null", "double_0 float8 null", "double_1 float8 null", "double_123 float8 null"
                         ))

langchain4j-pinecone/src/main/java/dev/langchain4j/store/embedding/pinecone/PineconeEmbeddingStore.java

@@ -5,7 +5,12 @@
 import dev.langchain4j.data.document.Metadata;
 import dev.langchain4j.data.embedding.Embedding;
 import dev.langchain4j.data.segment.TextSegment;
-import dev.langchain4j.store.embedding.*;
+import dev.langchain4j.store.embedding.CosineSimilarity;
+import dev.langchain4j.store.embedding.EmbeddingMatch;
+import dev.langchain4j.store.embedding.EmbeddingSearchRequest;
+import dev.langchain4j.store.embedding.EmbeddingSearchResult;
+import dev.langchain4j.store.embedding.EmbeddingStore;
+import dev.langchain4j.store.embedding.RelevanceScore;
 import io.pinecone.clients.Index;
 import io.pinecone.clients.Pinecone;
 import io.pinecone.unsigned_indices_model.QueryResponseWithUnsignedIndices;
@@ -14,22 +19,34 @@
 import org.openapitools.client.model.IndexList;
 import org.openapitools.client.model.IndexModel;
 
-import java.util.*;
+import java.util.ArrayList;
+import java.util.Collection;
+import java.util.Collections;
+import java.util.List;
+import java.util.Map;
+import java.util.Objects;
 
 import static dev.langchain4j.internal.Utils.isNullOrEmpty;
 import static dev.langchain4j.internal.Utils.randomUUID;
 import static dev.langchain4j.internal.ValidationUtils.ensureNotEmpty;
 import static dev.langchain4j.store.embedding.pinecone.PineconeHelper.metadataToStruct;
 import static dev.langchain4j.store.embedding.pinecone.PineconeHelper.structToMetadata;
 import static io.pinecone.commons.IndexInterface.buildUpsertVectorWithUnsignedIndices;
-import static java.util.Collections.emptyList;
 import static java.util.Collections.singletonList;
 import static java.util.Comparator.comparingDouble;
 import static java.util.stream.Collectors.toList;
 
 /**
  * Represents a <a href="https://www.pinecone.io/">Pinecone</a> index as an embedding store.
- * <p>Current implementation assumes the index uses the cosine distance metric.</p>
+ * <p>
+ * Current implementation assumes the index uses the cosine distance metric.
+ * <p>
+ * <b>WARNING! There is a known <a href="https://github.com/langchain4j/langchain4j/issues/1948">bug</a>:
+ * Pinecone stores all numbers as floating-point values,
+ * which means {@link Integer} and {@link Long} values (e.g., 1746714878034235396) stored in {@link Metadata}
+ * may be corrupted and returned as incorrect numbers!
+ * Possible workaround: convert integer/double values to {@link String} before storing them in {@link Metadata}.
+ * Please note that in this case metadata filtering might not work properly!</b>
  */
 public class PineconeEmbeddingStore implements EmbeddingStore<TextSegment> {
 

langchain4j-pinecone/src/test/java/dev/langchain4j/store/embedding/pinecone/PineconeEmbeddingStoreIT.java

@@ -80,4 +80,9 @@ protected static Stream<Arguments> should_filter_by_metadata() {
                         }
                 );
     }
+
+    @Override
+    protected boolean testLong1746714878034235396() {
+        return false; // TODO remove after https://github.com/langchain4j/langchain4j/issues/1948 is fixed
+    }
 }

langchain4j-weaviate/src/test/java/dev/langchain4j/store/embedding/weaviate/LocalWeaviateEmbeddingStoreIT.java

@@ -37,6 +37,7 @@ class LocalWeaviateEmbeddingStoreIT extends EmbeddingStoreIT {
             "long_minus_1",
             "long_0",
             "long_1",
+            "long_1746714878034235396",
             "long_max",
             "float_min",
             "float_minus_1",