diff --git a/agentscope-extensions/agentscope-extensions-rag/agentscope-extensions-rag-simple/src/main/java/io/agentscope/core/embedding/openai/OpenAITextEmbedding.java b/agentscope-extensions/agentscope-extensions-rag/agentscope-extensions-rag-simple/src/main/java/io/agentscope/core/embedding/openai/OpenAITextEmbedding.java
index 35290f27a0..f1c8339cf7 100644
--- a/agentscope-extensions/agentscope-extensions-rag/agentscope-extensions-rag-simple/src/main/java/io/agentscope/core/embedding/openai/OpenAITextEmbedding.java
+++ b/agentscope-extensions/agentscope-extensions-rag/agentscope-extensions-rag-simple/src/main/java/io/agentscope/core/embedding/openai/OpenAITextEmbedding.java
@@ -47,9 +47,12 @@ public class OpenAITextEmbedding implements EmbeddingModel {
 
     private static final Logger log = LoggerFactory.getLogger(OpenAITextEmbedding.class);
 
+    /** Default dimension when not explicitly configured, per EmbeddingModel interface contract. */
+    private static final int DEFAULT_DIMENSIONS = 1024;
+
     private final String apiKey;
     private final String modelName;
-    private final int dimensions;
+    private final Integer dimensions;
     private final ExecutionConfig defaultExecutionConfig;
 
     private final String baseUrl;
@@ -57,6 +60,9 @@ public class OpenAITextEmbedding implements EmbeddingModel {
     /**
      * Creates a new OpenAI text embedding model instance.
      *
+     * <p>This constructor is kept for binary compatibility with existing compiled code.
+     * It delegates to {@link #OpenAITextEmbedding(String, String, Integer, ExecutionConfig, String)}.
+     *
      * @param apiKey the API key for OpenAI authentication
      * @param modelName the model name (e.g., "text-embedding-3-small")
      * @param dimensions the dimension of embedding vectors
@@ -69,6 +75,32 @@ public OpenAITextEmbedding(
             int dimensions,
             ExecutionConfig defaultExecutionConfig,
             String baseUrl) {
+        this(apiKey, modelName, Integer.valueOf(dimensions), defaultExecutionConfig, baseUrl);
+    }
+
+    /**
+     * Creates a new OpenAI text embedding model instance with optional dimensions.
+     *
+     * <p>When {@code dimensions} is null, the dimensions parameter will not be sent
+     * to the API, allowing open-source models that don't support matryoshka
+     * representation (e.g., BAAI/bge-large-zh-v1.5) to work correctly.
+     *
+     * @param apiKey the API key for OpenAI authentication
+     * @param modelName the model name (e.g., "text-embedding-3-small")
+     * @param dimensions the dimension of embedding vectors (null to omit from API request)
+     * @param defaultExecutionConfig default execution configuration for timeout and retry
+     * @param baseUrl custom base URL for OpenAI API (null for default)
+     * @throws IllegalArgumentException if dimensions is non-null and not positive
+     */
+    public OpenAITextEmbedding(
+            String apiKey,
+            String modelName,
+            Integer dimensions,
+            ExecutionConfig defaultExecutionConfig,
+            String baseUrl) {
+        if (dimensions != null && dimensions <= 0) {
+            throw new IllegalArgumentException("dimensions must be positive, got: " + dimensions);
+        }
         this.apiKey = apiKey;
         this.modelName = modelName;
         this.dimensions = dimensions;
@@ -132,15 +164,19 @@ public Mono<double[]> embed(ContentBlock block) {
 
                                         OpenAIClient client = clientBuilder.build();
 
-                                        EmbeddingCreateParams createParams =
+                                        EmbeddingCreateParams.Builder paramsBuilder =
                                                 EmbeddingCreateParams.builder()
                                                         .model(modelName)
-                                                        .dimensions(dimensions)
                                                         .encodingFormat(
                                                                 EmbeddingCreateParams.EncodingFormat
                                                                         .FLOAT)
-                                                        .inputOfArrayOfStrings(List.of(text))
-                                                        .build();
+                                                        .inputOfArrayOfStrings(List.of(text));
+
+                                        if (dimensions != null) {
+                                            paramsBuilder.dimensions(dimensions);
+                                        }
+
+                                        EmbeddingCreateParams createParams = paramsBuilder.build();
 
                                         log.debug(
                                                 "OpenAI embedding call: model={},"
@@ -181,8 +217,9 @@ public Mono<double[]> embed(ContentBlock block) {
                                                 EmbeddingUtils.convertFloatListToDoubleArray(
                                                         embeddingValues);
 
-                                        // Validate dimension
-                                        if (embeddingArray.length != dimensions) {
+                                        // Validate dimension only when explicitly configured
+                                        if (dimensions != null
+                                                && embeddingArray.length != dimensions) {
                                             log.warn(
                                                     "Embedding dimension mismatch: expected={},"
                                                             + " actual={}",
@@ -223,9 +260,20 @@ public String getModelName() {
         return modelName;
     }
 
+    /**
+     * {@inheritDoc}
+     *
+     * <p>Returns the explicitly configured dimensions, or {@value #DEFAULT_DIMENSIONS} if {@code
+     * dimensions} was never configured (per the {@link EmbeddingModel#getDimensions()} contract,
+     * which documents 1024 as the default when unspecified). This fallback is a placeholder only:
+     * for open-source models where dimensions is intentionally left unset, the true embedding
+     * vector length returned by {@link #embed} may differ from this value. Callers needing the
+     * exact size should inspect the length of the array returned by {@link #embed}, not this
+     * getter.
+     */
     @Override
     public int getDimensions() {
-        return dimensions;
+        return dimensions != null ? dimensions : DEFAULT_DIMENSIONS;
     }
 
     /**
@@ -234,7 +282,7 @@ public int getDimensions() {
     public static class Builder {
         private String apiKey;
         private String modelName;
-        private int dimensions = 1536;
+        private Integer dimensions = 1536;
         private ExecutionConfig defaultExecutionConfig;
         private String baseUrl;
 
@@ -261,12 +309,21 @@ public Builder modelName(String modelName) {
         }
 
         /**
-         * Sets the dimension of embedding vectors.
+         * Sets the dimension of embedding vectors. Defaults to 1536.
+         *
+         * <p>When set to a positive value, the dimensions parameter will be included in the API
+         * request. This is only supported by OpenAI official embedding models (e.g.,
+         * text-embedding-3-small, text-embedding-3-large) that support matryoshka
+         * representation.
+         *
+         * <p>For open-source models (e.g., BAAI/bge-large-zh-v1.5) that do not support this
+         * parameter and will return a 400 error if it is sent, pass {@code null} to omit the
+         * parameter from the request entirely.
          *
-         * @param dimensions the dimension
+         * @param dimensions the dimension, or null to omit the parameter from the API request
          * @return this builder instance
          */
-        public Builder dimensions(int dimensions) {
+        public Builder dimensions(Integer dimensions) {
             this.dimensions = dimensions;
             return this;
         }
@@ -311,7 +368,7 @@ public OpenAITextEmbedding build() {
                 throw new IllegalStateException(
                         "modelName is required and cannot be null or empty");
             }
-            if (dimensions <= 0) {
+            if (dimensions != null && dimensions <= 0) {
                 throw new IllegalStateException("dimensions must be positive, got: " + dimensions);
             }
 
diff --git a/agentscope-extensions/agentscope-extensions-rag/agentscope-extensions-rag-simple/src/test/java/io/agentscope/core/embedding/openai/OpenAITextEmbeddingEmbedTest.java b/agentscope-extensions/agentscope-extensions-rag/agentscope-extensions-rag-simple/src/test/java/io/agentscope/core/embedding/openai/OpenAITextEmbeddingEmbedTest.java
index d8185974d0..6f8da795c7 100644
--- a/agentscope-extensions/agentscope-extensions-rag/agentscope-extensions-rag-simple/src/test/java/io/agentscope/core/embedding/openai/OpenAITextEmbeddingEmbedTest.java
+++ b/agentscope-extensions/agentscope-extensions-rag/agentscope-extensions-rag-simple/src/test/java/io/agentscope/core/embedding/openai/OpenAITextEmbeddingEmbedTest.java
@@ -314,4 +314,100 @@ void testEmptyEmbeddingListInResponse() {
                     .verify();
         }
     }
+
+    @Test
+    @DisplayName("Should successfully embed without dimensions parameter (open-source models)")
+    void testEmbedWithoutDimensions() {
+        // Explicitly omit dimensions - simulates open-source model usage
+        OpenAITextEmbedding modelWithoutDimensions =
+                OpenAITextEmbedding.builder()
+                        .apiKey("mock_api_key")
+                        .modelName("BAAI/bge-large-zh-v1.5")
+                        .dimensions(null)
+                        .executionConfig(ExecutionConfig.builder().maxAttempts(1).build())
+                        .build();
+
+        // Prepare mock response
+        Embedding mockEmbedding = mock(Embedding.class);
+        when(mockEmbedding.embedding()).thenReturn(Arrays.asList(0.5f, 0.6f, 0.7f, 0.8f));
+
+        CreateEmbeddingResponse mockResponse = mock(CreateEmbeddingResponse.class);
+        when(mockResponse.data()).thenReturn(Arrays.asList(mockEmbedding));
+
+        try (MockedStatic<OpenAIOkHttpClient> mockedClient =
+                Mockito.mockStatic(OpenAIOkHttpClient.class)) {
+
+            OpenAIOkHttpClient.Builder mockBuilder = mock(OpenAIOkHttpClient.Builder.class);
+            OpenAIClient mockOpenAIClient = mock(OpenAIClient.class);
+            EmbeddingService mockEmbeddings = mock(EmbeddingService.class);
+
+            when(mockBuilder.build()).thenReturn(mockOpenAIClient);
+            when(mockBuilder.apiKey(any())).thenReturn(mockBuilder);
+            when(mockBuilder.baseUrl(any(String.class))).thenReturn(mockBuilder);
+            when(mockBuilder.putHeader(any(), any())).thenReturn(mockBuilder);
+
+            mockedClient.when(OpenAIOkHttpClient::builder).thenReturn(mockBuilder);
+
+            when(mockOpenAIClient.embeddings()).thenReturn(mockEmbeddings);
+            when(mockEmbeddings.create(any(EmbeddingCreateParams.class))).thenReturn(mockResponse);
+
+            TextBlock textBlock = TextBlock.builder().text("Test open-source model").build();
+
+            StepVerifier.create(modelWithoutDimensions.embed(textBlock))
+                    .assertNext(
+                            embedding -> {
+                                assertNotNull(embedding);
+                                assertEquals(4, embedding.length);
+                                assertEquals(0.5, embedding[0], 0.001);
+                                assertEquals(0.8, embedding[3], 0.001);
+                            })
+                    .verifyComplete();
+        }
+    }
+
+    @Test
+    @DisplayName("Should not warn when returned embedding length matches configured dimensions")
+    void testEmbedWithMatchingDimensions() {
+        OpenAITextEmbedding modelWithMatchingDimensions =
+                OpenAITextEmbedding.builder()
+                        .apiKey("mock_api_key")
+                        .modelName(TEST_MODEL_NAME)
+                        .dimensions(3)
+                        .executionConfig(ExecutionConfig.builder().maxAttempts(1).build())
+                        .build();
+
+        Embedding mockEmbedding = mock(Embedding.class);
+        when(mockEmbedding.embedding()).thenReturn(Arrays.asList(0.1f, 0.2f, 0.3f));
+
+        CreateEmbeddingResponse mockResponse = mock(CreateEmbeddingResponse.class);
+        when(mockResponse.data()).thenReturn(Arrays.asList(mockEmbedding));
+
+        try (MockedStatic<OpenAIOkHttpClient> mockedClient =
+                Mockito.mockStatic(OpenAIOkHttpClient.class)) {
+
+            OpenAIOkHttpClient.Builder mockBuilder = mock(OpenAIOkHttpClient.Builder.class);
+            OpenAIClient mockOpenAIClient = mock(OpenAIClient.class);
+            EmbeddingService mockEmbeddings = mock(EmbeddingService.class);
+
+            when(mockBuilder.build()).thenReturn(mockOpenAIClient);
+            when(mockBuilder.apiKey(any())).thenReturn(mockBuilder);
+            when(mockBuilder.baseUrl(any(String.class))).thenReturn(mockBuilder);
+            when(mockBuilder.putHeader(any(), any())).thenReturn(mockBuilder);
+
+            mockedClient.when(OpenAIOkHttpClient::builder).thenReturn(mockBuilder);
+
+            when(mockOpenAIClient.embeddings()).thenReturn(mockEmbeddings);
+            when(mockEmbeddings.create(any(EmbeddingCreateParams.class))).thenReturn(mockResponse);
+
+            TextBlock textBlock = TextBlock.builder().text("Matching dimensions").build();
+
+            StepVerifier.create(modelWithMatchingDimensions.embed(textBlock))
+                    .assertNext(
+                            embedding -> {
+                                assertNotNull(embedding);
+                                assertEquals(3, embedding.length);
+                            })
+                    .verifyComplete();
+        }
+    }
 }
diff --git a/agentscope-extensions/agentscope-extensions-rag/agentscope-extensions-rag-simple/src/test/java/io/agentscope/core/embedding/openai/OpenAITextEmbeddingTest.java b/agentscope-extensions/agentscope-extensions-rag/agentscope-extensions-rag-simple/src/test/java/io/agentscope/core/embedding/openai/OpenAITextEmbeddingTest.java
index 07dfd6ae6f..4571c23e1f 100644
--- a/agentscope-extensions/agentscope-extensions-rag/agentscope-extensions-rag-simple/src/test/java/io/agentscope/core/embedding/openai/OpenAITextEmbeddingTest.java
+++ b/agentscope-extensions/agentscope-extensions-rag/agentscope-extensions-rag-simple/src/test/java/io/agentscope/core/embedding/openai/OpenAITextEmbeddingTest.java
@@ -177,6 +177,52 @@ void testMinimalBuilder() {
         assertNotNull(model);
     }
 
+    @Test
+    @DisplayName(
+            "Should build successfully with dimensions explicitly omitted for open-source models")
+    void testBuilderWithoutDimensions() {
+        OpenAITextEmbedding model =
+                OpenAITextEmbedding.builder()
+                        .apiKey(TEST_API_KEY)
+                        .modelName("BAAI/bge-large-zh-v1.5")
+                        .dimensions(null)
+                        .build();
+
+        assertNotNull(model);
+        assertEquals("BAAI/bge-large-zh-v1.5", model.getModelName());
+    }
+
+    @Test
+    @DisplayName("Should default dimensions to 1536 when not explicitly set")
+    void testBuilderDefaultsDimensionsTo1536() {
+        OpenAITextEmbedding model =
+                OpenAITextEmbedding.builder()
+                        .apiKey(TEST_API_KEY)
+                        .modelName(TEST_MODEL_NAME)
+                        .build();
+
+        assertEquals(
+                1536,
+                model.getDimensions(),
+                "Builder should default dimensions to 1536 for backward compatibility");
+    }
+
+    @Test
+    @DisplayName("Should return fallback dimensions (1024) when explicitly set to null")
+    void testGetDimensionsReturnsDefaultWhenExplicitlyNull() {
+        OpenAITextEmbedding model =
+                OpenAITextEmbedding.builder()
+                        .apiKey(TEST_API_KEY)
+                        .modelName(TEST_MODEL_NAME)
+                        .dimensions(null)
+                        .build();
+
+        assertEquals(
+                1024,
+                model.getDimensions(),
+                "getDimensions() should return default 1024 per EmbeddingModel interface");
+    }
+
     @Test
     @DisplayName("Should apply timeout configuration")
     void testTimeoutConfiguration() {
@@ -259,5 +305,41 @@ void testDimensionsValidation() {
                                 .modelName(TEST_MODEL_NAME)
                                 .dimensions(-1)
                                 .build());
+
+        // Building without calling dimensions() should NOT throw
+        OpenAITextEmbedding model =
+                OpenAITextEmbedding.builder()
+                        .apiKey(TEST_API_KEY)
+                        .modelName(TEST_MODEL_NAME)
+                        .build();
+        assertNotNull(model);
+    }
+
+    @Test
+    @DisplayName("Should throw exception when constructed directly with non-positive dimensions")
+    void testDirectConstructorRejectsNonPositiveDimensions() {
+        assertThrows(
+                IllegalArgumentException.class,
+                () ->
+                        new OpenAITextEmbedding(
+                                TEST_API_KEY, TEST_MODEL_NAME, Integer.valueOf(0), null, null));
+
+        assertThrows(
+                IllegalArgumentException.class,
+                () ->
+                        new OpenAITextEmbedding(
+                                TEST_API_KEY, TEST_MODEL_NAME, Integer.valueOf(-1), null, null));
+    }
+
+    @Test
+    @DisplayName(
+            "Should support the deprecated int-dimensions constructor for binary compatibility")
+    void testDeprecatedIntConstructor() {
+        OpenAITextEmbedding model =
+                new OpenAITextEmbedding(TEST_API_KEY, TEST_MODEL_NAME, TEST_DIMENSIONS, null, null);
+
+        assertNotNull(model);
+        assertEquals(TEST_MODEL_NAME, model.getModelName());
+        assertEquals(TEST_DIMENSIONS, model.getDimensions());
     }
 }