Rebrand to 'VoyageAI by MongoDB', refresh models, support Union contextualized inputs

PACKAGES.md

@@ -40,7 +40,7 @@ A BOM is provided that can be used to define the versions of all Semantic Kernel
 : Provides a connector that can be used to interact with the OpenAI API.
 
 `semantickernel-aiservices-voyageai`
-: Provides connectors for VoyageAI's embedding and reranking services, including text embeddings, contextualized embeddings, multimodal embeddings, and document reranking.
+: Provides connectors for VoyageAI by MongoDB's embedding and reranking services, including text embeddings (e.g. voyage-4-large, voyage-4, voyage-code-4), contextualized embeddings (voyage-context-4), multimodal embeddings (voyage-multimodal-3.5), and document reranking (rerank-2.5).
 
 ## Example Configurations
 
@@ -75,9 +75,9 @@ POM XML for a simple project that uses OpenAI.
 </project>
 ```
 
-### Example: VoyageAI Embeddings and Reranking
+### Example: VoyageAI by MongoDB Embeddings and Reranking
 
-POM XML for a project that uses VoyageAI for embeddings and reranking.
+POM XML for a project that uses VoyageAI by MongoDB for embeddings and reranking.
 
 ```xml
 

aiservices/voyageai/src/main/java/com/microsoft/semantickernel/aiservices/voyageai/contextualizedembedding/VoyageAIContextualizedEmbeddingGenerationService.java

@@ -19,9 +19,9 @@
 import java.util.stream.Collectors;
 
 /**
- * VoyageAI contextualized embedding generation service.
+ * VoyageAI by MongoDB contextualized embedding generation service.
  * Generates embeddings that capture both local chunk details and global document-level metadata.
- * Supports models like voyage-3.
+ * Supports voyage-context-4 (current) and voyage-context-3.
  */
 public final class VoyageAIContextualizedEmbeddingGenerationService implements TextEmbeddingGenerationService {
 
@@ -35,7 +35,7 @@ public final class VoyageAIContextualizedEmbeddingGenerationService implements T
      * Creates a new instance of VoyageAI contextualized embedding generation service.
      *
      * @param client    VoyageAI client
-     * @param modelId   Model ID (e.g., "voyage-3")
+     * @param modelId   Model ID (e.g., "voyage-context-4")
      * @param serviceId Optional service ID
      */
     public VoyageAIContextualizedEmbeddingGenerationService(
@@ -66,9 +66,14 @@ public String getModelId() {
     }
 
     /**
-     * Generates contextualized embeddings for document chunks.
+     * Generates contextualized embeddings for pre-chunked documents.
      *
-     * @param inputs List of lists where each inner list contains document chunks
+     * <p>Sends the inputs in the nested form of the official
+     * {@code inputs: Union[List[List[str]], List[str]]} specification, where each inner
+     * list holds one document's chunks and is embedded as a group so every chunk is
+     * encoded in the context of the others.
+     *
+     * @param inputs List of lists where each inner list contains one document's chunks
      * @return A Mono containing a list of embeddings for all chunks across all documents
      */
     public Mono<List<Embedding>> generateContextualizedEmbeddingsAsync(List<List<String>> inputs) {
@@ -84,6 +89,41 @@ public Mono<List<Embedding>> generateContextualizedEmbeddingsAsync(List<List<Str
         request.setInputs(inputs);
         request.setModel(modelId);
 
+        return sendRequest(request);
+    }
+
+    /**
+     * Generates contextualized embeddings from a flat list of full-document strings.
+     *
+     * <p>Sends the inputs in the flat form of the official
+     * {@code inputs: Union[List[List[str]], List[str]]} specification and enables
+     * server-side auto-chunking, so each document is split and every resulting chunk is
+     * embedded with document-level context. Per the API contract the flat, document-typed
+     * form requires auto-chunking, so {@code input_type} is set to {@code document} and
+     * {@code enable_auto_chunking} to {@code true}.
+     *
+     * @param documents flat list of full-document strings
+     * @return A Mono containing a list of embeddings for all chunks across all documents
+     */
+    public Mono<List<Embedding>> generateContextualizedEmbeddingsForDocumentsAsync(List<String> documents) {
+        if (documents == null || documents.isEmpty()) {
+            return Mono.just(Collections.emptyList());
+        }
+
+        LOGGER.debug("Generating contextualized embeddings for {} documents (auto-chunking) using model {}",
+            documents.size(), modelId);
+
+        VoyageAIModels.ContextualizedEmbeddingRequest request =
+            new VoyageAIModels.ContextualizedEmbeddingRequest();
+        request.setFlatInputs(documents);
+        request.setModel(modelId);
+        request.setInputType("document");
+        request.setEnableAutoChunking(true);
+
+        return sendRequest(request);
+    }
+
+    private Mono<List<Embedding>> sendRequest(VoyageAIModels.ContextualizedEmbeddingRequest request) {
         return client.sendRequestAsync(
             "contextualizedembeddings",
             request,
@@ -173,7 +213,7 @@ public Builder withClient(VoyageAIClient client) {
         /**
          * Sets the model ID.
          *
-         * @param modelId Model ID (e.g., "voyage-3")
+         * @param modelId Model ID (e.g., "voyage-context-4")
          * @return This builder
          */
         public Builder withModelId(String modelId) {

aiservices/voyageai/src/main/java/com/microsoft/semantickernel/aiservices/voyageai/core/VoyageAIClient.java

@@ -18,7 +18,7 @@
 import java.util.concurrent.TimeUnit;
 
 /**
- * HTTP client for VoyageAI API.
+ * HTTP client for the VoyageAI by MongoDB API.
  */
 public final class VoyageAIClient {
     private static final Logger LOGGER = LoggerFactory.getLogger(VoyageAIClient.class);

aiservices/voyageai/src/main/java/com/microsoft/semantickernel/aiservices/voyageai/core/VoyageAIModels.java

@@ -8,7 +8,7 @@
 import java.util.List;
 
 /**
- * VoyageAI API request and response models.
+ * VoyageAI by MongoDB API request and response models.
  */
 public class VoyageAIModels {
 
@@ -122,6 +122,7 @@ public void setUsage(EmbeddingUsage usage) {
     /**
      * Embedding data item.
      */
+    @JsonIgnoreProperties(ignoreUnknown = true)
     public static class EmbeddingDataItem {
         @JsonProperty("object")
         private String object;
@@ -275,6 +276,7 @@ public void setUsage(EmbeddingUsage usage) {
     /**
      * Rerank data item.
      */
+    @JsonIgnoreProperties(ignoreUnknown = true)
     public static class RerankDataItem {
         @JsonProperty("index")
         private int index;
@@ -303,11 +305,18 @@ public void setRelevanceScore(double relevanceScore) {
 
     /**
      * Request model for contextualized embeddings.
+     *
+     * <p>The VoyageAI API accepts {@code inputs} as either a nested list of chunks
+     * ({@code List<List<String>>}, one inner list per document) or a flat list of
+     * full-document strings ({@code List<String>}), matching the official
+     * {@code inputs: Union[List[List[str]], List[str]]} specification. The flat form
+     * is used together with server-side auto-chunking.
      */
     @JsonInclude(JsonInclude.Include.NON_NULL)
     public static class ContextualizedEmbeddingRequest {
+        // Holds either List<List<String>> (nested chunks) or List<String> (flat documents).
         @JsonProperty("inputs")
-        private List<List<String>> inputs;
+        private Object inputs;
 
         @JsonProperty("model")
         private String model;
@@ -324,16 +333,41 @@ public static class ContextualizedEmbeddingRequest {
         @JsonProperty("output_dtype")
         private String outputDtype;
 
+        @JsonProperty("enable_auto_chunking")
+        private Boolean enableAutoChunking;
+
         @SuppressFBWarnings("EI_EXPOSE_REP")
-        public List<List<String>> getInputs() {
+        public Object getInputs() {
             return inputs;
         }
 
-        @SuppressFBWarnings("EI_EXPOSE_REP2")
+        /**
+         * Sets the inputs as a nested list of chunks, one inner list per document.
+         *
+         * @param inputs nested list of document chunks
+         */
         public void setInputs(List<List<String>> inputs) {
             this.inputs = inputs;
         }
 
+        /**
+         * Sets the inputs as a flat list of full-document strings. Use together with
+         * server-side auto-chunking (see {@link #setEnableAutoChunking(Boolean)}).
+         *
+         * @param inputs flat list of full-document strings
+         */
+        public void setFlatInputs(List<String> inputs) {
+            this.inputs = inputs;
+        }
+
+        public Boolean getEnableAutoChunking() {
+            return enableAutoChunking;
+        }
+
+        public void setEnableAutoChunking(Boolean enableAutoChunking) {
+            this.enableAutoChunking = enableAutoChunking;
+        }
+
         public String getModel() {
             return model;
         }
@@ -435,6 +469,7 @@ public void setEmbeddings(List<EmbeddingItem> embeddings) {
     /**
      * Embedding item with chunk information.
      */
+    @JsonIgnoreProperties(ignoreUnknown = true)
     public static class EmbeddingItem {
         @JsonProperty("embedding")
         private float[] embedding;

aiservices/voyageai/src/main/java/com/microsoft/semantickernel/aiservices/voyageai/multimodalembedding/VoyageAIMultimodalEmbeddingGenerationService.java

@@ -19,9 +19,9 @@
 import java.util.stream.Collectors;
 
 /**
- * VoyageAI multimodal embedding generation service.
+ * VoyageAI by MongoDB multimodal embedding generation service.
  * Generates embeddings for text, images, or interleaved text and images.
- * Supports the voyage-multimodal-3 model.
+ * Supports voyage-multimodal-3.5 (current) and voyage-multimodal-3.
  * <p>
  * Constraints:
  * - Maximum 1,000 inputs per request
@@ -41,7 +41,7 @@ public final class VoyageAIMultimodalEmbeddingGenerationService implements TextE
      * Creates a new instance of VoyageAI multimodal embedding generation service.
      *
      * @param client    VoyageAI client
-     * @param modelId   Model ID (e.g., "voyage-multimodal-3")
+     * @param modelId   Model ID (e.g., "voyage-multimodal-3.5")
      * @param serviceId Optional service ID
      */
     public VoyageAIMultimodalEmbeddingGenerationService(
@@ -196,7 +196,7 @@ public Builder withClient(VoyageAIClient client) {
         /**
          * Sets the model ID.
          *
-         * @param modelId Model ID (e.g., "voyage-multimodal-3")
+         * @param modelId Model ID (e.g., "voyage-multimodal-3.5")
          * @return This builder
          */
         public Builder withModelId(String modelId) {

aiservices/voyageai/src/main/java/com/microsoft/semantickernel/aiservices/voyageai/reranking/VoyageAITextRerankingService.java

@@ -17,8 +17,9 @@
 import java.util.stream.Collectors;
 
 /**
- * VoyageAI implementation of {@link TextRerankingService}.
- * Supports models like rerank-2, rerank-2-lite.
+ * VoyageAI by MongoDB implementation of {@link TextRerankingService}.
+ * Supports current models such as rerank-2.5 and rerank-2.5-lite (and preview models
+ * rerank-3, rerank-3-lite), as well as older models like rerank-2 and rerank-2-lite.
  */
 public final class VoyageAITextRerankingService implements TextRerankingService {
 
@@ -33,7 +34,7 @@ public final class VoyageAITextRerankingService implements TextRerankingService
      * Creates a new instance of VoyageAI text reranking service.
      *
      * @param client    VoyageAI client
-     * @param modelId   Model ID (e.g., "rerank-2")
+     * @param modelId   Model ID (e.g., "rerank-2.5")
      * @param serviceId Optional service ID
      * @param topK      Optional top K results to return
      */
@@ -140,7 +141,7 @@ public Builder withClient(VoyageAIClient client) {
         /**
          * Sets the model ID.
          *
-         * @param modelId Model ID (e.g., "rerank-2")
+         * @param modelId Model ID (e.g., "rerank-2.5")
          * @return This builder
          */
         public Builder withModelId(String modelId) {

aiservices/voyageai/src/main/java/com/microsoft/semantickernel/aiservices/voyageai/textembedding/VoyageAITextEmbeddingGenerationService.java

@@ -18,8 +18,10 @@
 import java.util.stream.Collectors;
 
 /**
- * VoyageAI implementation of {@link TextEmbeddingGenerationService}.
- * Supports models like voyage-3-large, voyage-3.5, voyage-code-3, voyage-finance-2, voyage-law-2.
+ * VoyageAI by MongoDB implementation of {@link TextEmbeddingGenerationService}.
+ * Supports current models such as voyage-4-large, voyage-4, voyage-4-lite, voyage-code-4,
+ * voyage-finance-2 and voyage-law-2, as well as older models like voyage-3-large,
+ * voyage-3.5, voyage-3.5-lite and voyage-code-3.
  */
 public final class VoyageAITextEmbeddingGenerationService implements TextEmbeddingGenerationService {
 
@@ -33,7 +35,7 @@ public final class VoyageAITextEmbeddingGenerationService implements TextEmbeddi
      * Creates a new instance of VoyageAI text embedding generation service.
      *
      * @param client    VoyageAI client
-     * @param modelId   Model ID (e.g., "voyage-3-large")
+     * @param modelId   Model ID (e.g., "voyage-4-large")
      * @param serviceId Optional service ID
      */
     public VoyageAITextEmbeddingGenerationService(
@@ -143,7 +145,7 @@ public Builder withClient(VoyageAIClient client) {
         /**
          * Sets the model ID.
          *
-         * @param modelId Model ID (e.g., "voyage-3-large")
+         * @param modelId Model ID (e.g., "voyage-4-large")
          * @return This builder
          */
         public Builder withModelId(String modelId) {