From 9a63538840e2c73324e00f24676c5d58ea898087 Mon Sep 17 00:00:00 2001
From: Luca Chang <lucalc@amazon.com>
Date: Fri, 16 May 2025 15:53:33 -0700
Subject: [PATCH] feat: add AudioContent support to MCP schema (#243)

- Add AudioContent record class implementing Annotated and Content interfaces
- Update Content sealed interface to permit AudioContent
- Add audio type handling in Content.type() method
- Include AudioContent in JsonSubTypes annotation for polymorphic deserialization
- Add comprehensive tests for AudioContent serialization and deserialization
- Update error message test to include audio in known type ids

https://github.com/modelcontextprotocol/modelcontextprotocol/blob/c87a0da6d8c2436d56a6398023c80b0562224454/schema/2025-03-26/schema.ts#L987-L1009
---
 .../modelcontextprotocol/spec/McpSchema.java  | 14 ++++++++++-
 .../spec/McpSchemaTests.java                  | 24 ++++++++++++++++++-
 2 files changed, 36 insertions(+), 2 deletions(-)

diff --git a/mcp/src/main/java/io/modelcontextprotocol/spec/McpSchema.java b/mcp/src/main/java/io/modelcontextprotocol/spec/McpSchema.java
index e21d53c..5792aad 100644
--- a/mcp/src/main/java/io/modelcontextprotocol/spec/McpSchema.java
+++ b/mcp/src/main/java/io/modelcontextprotocol/spec/McpSchema.java
@@ -1359,8 +1359,9 @@ public final class McpSchema {
 	@JsonTypeInfo(use = JsonTypeInfo.Id.NAME, include = JsonTypeInfo.As.PROPERTY, property = "type")
 	@JsonSubTypes({ @JsonSubTypes.Type(value = TextContent.class, name = "text"),
 			@JsonSubTypes.Type(value = ImageContent.class, name = "image"),
+			@JsonSubTypes.Type(value = AudioContent.class, name = "audio"),
 			@JsonSubTypes.Type(value = EmbeddedResource.class, name = "resource") })
-	public sealed interface Content permits TextContent, ImageContent, EmbeddedResource {
+	public sealed interface Content permits TextContent, ImageContent, AudioContent, EmbeddedResource {
 
 		default String type() {
 			if (this instanceof TextContent) {
@@ -1369,6 +1370,9 @@ public final class McpSchema {
 			else if (this instanceof ImageContent) {
 				return "image";
 			}
+			else if (this instanceof AudioContent) {
+				return "audio";
+			}
 			else if (this instanceof EmbeddedResource) {
 				return "resource";
 			}
@@ -1398,6 +1402,14 @@ public final class McpSchema {
 		@JsonProperty("mimeType") String mimeType) implements Content { // @formatter:on
 	}
 
+	@JsonInclude(JsonInclude.Include.NON_ABSENT)
+	@JsonIgnoreProperties(ignoreUnknown = true)
+	public record AudioContent( // @formatter:off
+		@JsonProperty("annotations") Annotations annotations,
+		@JsonProperty("data") String data,
+		@JsonProperty("mimeType") String mimeType) implements Annotated, Content { // @formatter:on
+	}
+
 	@JsonInclude(JsonInclude.Include.NON_ABSENT)
 	@JsonIgnoreProperties(ignoreUnknown = true)
 	public record EmbeddedResource( // @formatter:off
diff --git a/mcp/src/test/java/io/modelcontextprotocol/spec/McpSchemaTests.java b/mcp/src/test/java/io/modelcontextprotocol/spec/McpSchemaTests.java
index 99015d8..6d027b5 100644
--- a/mcp/src/test/java/io/modelcontextprotocol/spec/McpSchemaTests.java
+++ b/mcp/src/test/java/io/modelcontextprotocol/spec/McpSchemaTests.java
@@ -59,7 +59,7 @@ public class McpSchemaTests {
 				{"type":"WRONG","text":"XXX"}""", McpSchema.TextContent.class))
 			.isInstanceOf(InvalidTypeIdException.class)
 			.hasMessageContaining(
-					"Could not resolve type id 'WRONG' as a subtype of `io.modelcontextprotocol.spec.McpSchema$TextContent`: known type ids = [image, resource, text]");
+					"Could not resolve type id 'WRONG' as a subtype of `io.modelcontextprotocol.spec.McpSchema$TextContent`: known type ids = [audio, image, resource, text]");
 	}
 
 	@Test
@@ -84,6 +84,28 @@ public class McpSchemaTests {
 		assertThat(imageContent.mimeType()).isEqualTo("image/png");
 	}
 
+	@Test
+	void testAudioContent() throws Exception {
+		McpSchema.AudioContent audioContent = new McpSchema.AudioContent(null, "base64encodeddata", "audio/wav");
+		String value = mapper.writeValueAsString(audioContent);
+
+		assertThatJson(value).when(Option.IGNORING_ARRAY_ORDER)
+			.when(Option.IGNORING_EXTRA_ARRAY_ITEMS)
+			.isObject()
+			.isEqualTo(json("""
+					{"type":"audio","data":"base64encodeddata","mimeType":"audio/wav"}"""));
+	}
+
+	@Test
+	void testAudioContentDeserialization() throws Exception {
+		McpSchema.AudioContent audioContent = mapper.readValue("""
+				{"type":"audio","data":"base64encodeddata","mimeType":"audio/wav"}""", McpSchema.AudioContent.class);
+		assertThat(audioContent).isNotNull();
+		assertThat(audioContent.type()).isEqualTo("audio");
+		assertThat(audioContent.data()).isEqualTo("base64encodeddata");
+		assertThat(audioContent.mimeType()).isEqualTo("audio/wav");
+	}
+
 	@Test
 	void testEmbeddedResource() throws Exception {
 		McpSchema.TextResourceContents resourceContents = new McpSchema.TextResourceContents("resource://test",