feat: enhance BeanOutputConverter with customizable text cleaning capabilities

Signed-off-by: liugddx <[email protected]>

Author: liugddx

spring-ai-model/src/main/java/org/springframework/ai/converter/CompositeResponseTextCleaner.java

@@ -106,4 +106,3 @@ public CompositeResponseTextCleaner build() {
 	}
 
 }
-

spring-ai-model/src/main/java/org/springframework/ai/converter/MarkdownCodeBlockCleaner.java

@@ -71,4 +71,3 @@ public String clean(String text) {
 	}
 
 }
-

spring-ai-model/src/main/java/org/springframework/ai/converter/ResponseTextCleaner.java

@@ -35,4 +35,3 @@ public interface ResponseTextCleaner {
 	String clean(String text);
 
 }
-

spring-ai-model/src/main/java/org/springframework/ai/converter/ThinkingTagCleaner.java

@@ -169,4 +169,3 @@ public ThinkingTagCleaner build() {
 	}
 
 }
-

spring-ai-model/src/main/java/org/springframework/ai/converter/WhitespaceCleaner.java

@@ -30,4 +30,3 @@ public String clean(String text) {
 	}
 
 }
-

spring-ai-model/src/test/java/org/springframework/ai/converter/BeanOutputConverterTest.java

@@ -273,97 +273,97 @@ void convertWithThinkingTagsAndMarkdownCodeBlock() {
 			assertThat(testClass.getSomeString()).isEqualTo("some value");
 		}
 
-	@Test
-	void convertWithMultipleThinkingTags() {
-		var converter = new BeanOutputConverter<>(TestClass.class);
-		String textWithThinkingTags = "<thinking>First thought</thinking><thinking>Second thought</thinking>{ \"someString\": \"some value\" }";
-		var testClass = converter.convert(textWithThinkingTags);
-		assertThat(testClass.getSomeString()).isEqualTo("some value");
-	}
+		@Test
+		void convertWithMultipleThinkingTags() {
+			var converter = new BeanOutputConverter<>(TestClass.class);
+			String textWithThinkingTags = "<thinking>First thought</thinking><thinking>Second thought</thinking>{ \"someString\": \"some value\" }";
+			var testClass = converter.convert(textWithThinkingTags);
+			assertThat(testClass.getSomeString()).isEqualTo("some value");
+		}
 
-	@Test
-	void convertWithQwenThinkTags() {
-		// Test Qwen model format: <think>...</think>
-		var converter = new BeanOutputConverter<>(TestClass.class);
-		String textWithThinkTags = "<think>Let me analyze this...</think>{ \"someString\": \"qwen test\" }";
-		var testClass = converter.convert(textWithThinkTags);
-		assertThat(testClass.getSomeString()).isEqualTo("qwen test");
-	}
+		@Test
+		void convertWithQwenThinkTags() {
+			// Test Qwen model format: <think>...</think>
+			var converter = new BeanOutputConverter<>(TestClass.class);
+			String textWithThinkTags = "<think>Let me analyze this...</think>{ \"someString\": \"qwen test\" }";
+			var testClass = converter.convert(textWithThinkTags);
+			assertThat(testClass.getSomeString()).isEqualTo("qwen test");
+		}
 
-	@Test
-	void convertWithQwenThinkTagsMultiline() {
-		var converter = new BeanOutputConverter<>(TestClass.class);
-		String textWithThinkTags = """
-				<think>
-				Analyzing the request step by step
-				First, I need to understand the schema
-				Then generate the JSON
-				</think>
-				{ "someString": "qwen multiline" }
-				""";
-		var testClass = converter.convert(textWithThinkTags);
-		assertThat(testClass.getSomeString()).isEqualTo("qwen multiline");
-	}
+		@Test
+		void convertWithQwenThinkTagsMultiline() {
+			var converter = new BeanOutputConverter<>(TestClass.class);
+			String textWithThinkTags = """
+					<think>
+					Analyzing the request step by step
+					First, I need to understand the schema
+					Then generate the JSON
+					</think>
+					{ "someString": "qwen multiline" }
+					""";
+			var testClass = converter.convert(textWithThinkTags);
+			assertThat(testClass.getSomeString()).isEqualTo("qwen multiline");
+		}
 
-	@Test
-	void convertWithMixedThinkingAndThinkTags() {
-		// Test mixed format from different models
-		var converter = new BeanOutputConverter<>(TestClass.class);
-		String textWithMixedTags = "<thinking>Nova reasoning</thinking><think>Qwen analysis</think>{ \"someString\": \"mixed test\" }";
-		var testClass = converter.convert(textWithMixedTags);
-		assertThat(testClass.getSomeString()).isEqualTo("mixed test");
-	}
+		@Test
+		void convertWithMixedThinkingAndThinkTags() {
+			// Test mixed format from different models
+			var converter = new BeanOutputConverter<>(TestClass.class);
+			String textWithMixedTags = "<thinking>Nova reasoning</thinking><think>Qwen analysis</think>{ \"someString\": \"mixed test\" }";
+			var testClass = converter.convert(textWithMixedTags);
+			assertThat(testClass.getSomeString()).isEqualTo("mixed test");
+		}
 
-	@Test
-	void convertWithReasoningTags() {
-		// Test alternative reasoning tags
-		var converter = new BeanOutputConverter<>(TestClass.class);
-		String textWithReasoningTags = "<reasoning>Internal reasoning process</reasoning>{ \"someString\": \"reasoning test\" }";
-		var testClass = converter.convert(textWithReasoningTags);
-		assertThat(testClass.getSomeString()).isEqualTo("reasoning test");
-	}
+		@Test
+		void convertWithReasoningTags() {
+			// Test alternative reasoning tags
+			var converter = new BeanOutputConverter<>(TestClass.class);
+			String textWithReasoningTags = "<reasoning>Internal reasoning process</reasoning>{ \"someString\": \"reasoning test\" }";
+			var testClass = converter.convert(textWithReasoningTags);
+			assertThat(testClass.getSomeString()).isEqualTo("reasoning test");
+		}
 
-	@Test
-	void convertWithMarkdownThinkingBlock() {
-		// Test markdown-style thinking block
-		var converter = new BeanOutputConverter<>(TestClass.class);
-		String textWithMarkdownThinking = """
-				```thinking
-				This is a markdown-style thinking block
-				Used by some models
-				```
-				{ "someString": "markdown thinking" }
-				""";
-		var testClass = converter.convert(textWithMarkdownThinking);
-		assertThat(testClass.getSomeString()).isEqualTo("markdown thinking");
-	}
+		@Test
+		void convertWithMarkdownThinkingBlock() {
+			// Test markdown-style thinking block
+			var converter = new BeanOutputConverter<>(TestClass.class);
+			String textWithMarkdownThinking = """
+					```thinking
+					This is a markdown-style thinking block
+					Used by some models
+					```
+					{ "someString": "markdown thinking" }
+					""";
+			var testClass = converter.convert(textWithMarkdownThinking);
+			assertThat(testClass.getSomeString()).isEqualTo("markdown thinking");
+		}
 
-	@Test
-	void convertWithCaseInsensitiveTags() {
-		// Test case insensitive tag matching
-		var converter = new BeanOutputConverter<>(TestClass.class);
-		String textWithUpperCaseTags = "<THINKING>UPPERCASE THINKING</THINKING>{ \"someString\": \"case test\" }";
-		var testClass = converter.convert(textWithUpperCaseTags);
-		assertThat(testClass.getSomeString()).isEqualTo("case test");
-	}
+		@Test
+		void convertWithCaseInsensitiveTags() {
+			// Test case insensitive tag matching
+			var converter = new BeanOutputConverter<>(TestClass.class);
+			String textWithUpperCaseTags = "<THINKING>UPPERCASE THINKING</THINKING>{ \"someString\": \"case test\" }";
+			var testClass = converter.convert(textWithUpperCaseTags);
+			assertThat(testClass.getSomeString()).isEqualTo("case test");
+		}
 
-	@Test
-	void convertWithComplexNestedStructure() {
-		// Test complex scenario with multiple formats combined
-		var converter = new BeanOutputConverter<>(TestClass.class);
-		String complexText = """
-				<thinking>Nova model reasoning</thinking>
-				<think>Qwen model analysis</think>
-
-				```json
-				{ "someString": "complex test" }
-				```
-				""";
-		var testClass = converter.convert(complexText);
-		assertThat(testClass.getSomeString()).isEqualTo("complex test");
-	}
+		@Test
+		void convertWithComplexNestedStructure() {
+			// Test complex scenario with multiple formats combined
+			var converter = new BeanOutputConverter<>(TestClass.class);
+			String complexText = """
+					<thinking>Nova model reasoning</thinking>
+					<think>Qwen model analysis</think>
 
-}
+					```json
+					{ "someString": "complex test" }
+					```
+					""";
+			var testClass = converter.convert(complexText);
+			assertThat(testClass.getSomeString()).isEqualTo("complex test");
+		}
+
+	}
 
 	// @checkstyle:off RegexpSinglelineJavaCheck
 	@Nested

spring-ai-model/src/test/java/org/springframework/ai/converter/CompositeResponseTextCleanerTest.java

@@ -52,11 +52,11 @@ void shouldWorkWithMultipleCleaners() {
 				new MarkdownCodeBlockCleaner());
 
 		String input = """
-					<thinking>Reasoning</thinking>
-					```json
-					{"key": "value"}
-					```
-					""";
+				<thinking>Reasoning</thinking>
+				```json
+				{"key": "value"}
+				```
+				""";
 		String result = cleaner.clean(input);
 		assertThat(result).isEqualTo("{\"key\": \"value\"}");
 	}
@@ -103,4 +103,3 @@ void shouldHandleEmptyCleanersList() {
 	}
 
 }
-

spring-ai-model/src/test/java/org/springframework/ai/converter/ThinkingTagCleanerTest.java

@@ -139,17 +139,14 @@ void shouldSupportBuilderWithAdditionalPatterns() {
 
 	@Test
 	void shouldThrowExceptionWhenPatternsAreNull() {
-		assertThatThrownBy(() -> new ThinkingTagCleaner((String[]) null))
-			.isInstanceOf(IllegalArgumentException.class)
+		assertThatThrownBy(() -> new ThinkingTagCleaner((String[]) null)).isInstanceOf(IllegalArgumentException.class)
 			.hasMessageContaining("patternStrings cannot be null");
 	}
 
 	@Test
 	void shouldThrowExceptionWhenPatternsAreEmpty() {
-		assertThatThrownBy(() -> new ThinkingTagCleaner(new String[0]))
-			.isInstanceOf(IllegalArgumentException.class)
+		assertThatThrownBy(() -> new ThinkingTagCleaner(new String[0])).isInstanceOf(IllegalArgumentException.class)
 			.hasMessageContaining("patternStrings cannot be empty");
 	}
 
 }
-