feat: update cpde

Signed-off-by: yuluo-yx <yuluo08290126@gmail.com>

Author: yuluo-yx

spring-ai-alibaba-integration-example/backend/src/main/java/com/alibaba/cloud/ai/application/service/SAAWebSearchService.java

@@ -13,6 +13,7 @@
 import org.springframework.ai.chat.client.advisor.RetrievalAugmentationAdvisor;
 import org.springframework.ai.chat.client.advisor.SimpleLoggerAdvisor;
 import org.springframework.ai.chat.prompt.PromptTemplate;
+import org.springframework.ai.rag.postretrieval.ranking.DocumentRanker;
 import org.springframework.ai.rag.preretrieval.query.expansion.QueryExpander;
 import org.springframework.ai.rag.preretrieval.query.transformation.QueryTransformer;
 import org.springframework.beans.factory.annotation.Qualifier;
@@ -44,9 +45,10 @@ public SAAWebSearchService(
 			ChatClient.Builder chatClientBuilder,
 			QueryTransformer queryTransformer,
 			QueryExpander queryExpander,
-			@Qualifier("queryArgumentPromptTemplate") PromptTemplate queryArgumentPromptTemplate,
 			IQSSearchEngine searchEngine,
-			DataClean dataCleaner
+			DataClean dataCleaner,
+			DocumentRanker documentRanker,
+			@Qualifier("queryArgumentPromptTemplate") PromptTemplate queryArgumentPromptTemplate
 	) {
 
 		this.queryTransformer = queryTransformer;
@@ -60,7 +62,7 @@ public SAAWebSearchService(
 		this.chatClient = chatClientBuilder
 				.defaultOptions(
 						DashScopeChatOptions.builder()
-								.withModel("deepseek-r1")
+								.withModel("qwen-plus")
 								.withIncrementalOutput(false)
 								.build())
 				.build();
@@ -72,6 +74,8 @@ public SAAWebSearchService(
 				.searchEngine(searchEngine)
 				.dataCleaner(dataCleaner)
 				.maxResults(2)
+				.enableRanker(true)
+				.documentRanker(documentRanker)
 				.build();
 	}
 
@@ -98,8 +102,9 @@ private RetrievalAugmentationAdvisor createRetrievalAugmentationAdvisor() {
 						new CustomContextQueryAugmenter(
 								queryArgumentPromptTemplate,
 								null,
-								true))
-				.queryExpander(queryExpander).documentJoiner(new ConcatenationDocumentJoiner())
+								true)
+				).queryExpander(queryExpander)
+				.documentJoiner(new ConcatenationDocumentJoiner())
 				.build();
 	}
 

spring-ai-alibaba-integration-example/backend/src/main/java/com/alibaba/cloud/ai/application/websearch/config/WeSearchConfiguration.java

@@ -45,7 +45,7 @@ public QueryTransformer queryTransformer(
 		return RewriteQueryTransformer.builder()
 				.chatClientBuilder(chatClient.mutate())
 				.promptTemplate(transformerPromptTemplate)
-				.targetSearchSystem("Web search")
+				.targetSearchSystem("联网搜索")
 				.build();
 	}
 

spring-ai-alibaba-integration-example/backend/src/main/java/com/alibaba/cloud/ai/application/websearch/rag/WebSearchRetriever.java

@@ -5,11 +5,13 @@
 import com.alibaba.cloud.ai.application.websearch.core.IQSSearchEngine;
 import com.alibaba.cloud.ai.application.websearch.data.DataClean;
 import com.alibaba.cloud.ai.application.websearch.entity.GenericSearchResult;
+import org.jetbrains.annotations.NotNull;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
 import org.springframework.ai.document.Document;
 import org.springframework.ai.rag.Query;
+import org.springframework.ai.rag.postretrieval.ranking.DocumentRanker;
 import org.springframework.ai.rag.retrieval.search.DocumentRetriever;
 import org.springframework.lang.Nullable;
 
@@ -28,13 +30,20 @@ public class WebSearchRetriever implements DocumentRetriever {
 
 	private final DataClean dataCleaner;
 
+	private final DocumentRanker documentRanker;
+
+	private final boolean enableRanker;
+
 	private WebSearchRetriever(Builder builder) {
 
 		this.searchEngine = builder.searchEngine;
 		this.maxResults = builder.maxResults;
 		this.dataCleaner = builder.dataCleaner;
+		this.documentRanker = builder.documentRanker;
+		this.enableRanker = builder.enableRanker;
 	}
 
+	@NotNull
 	@Override
 	public List<Document> retrieve(
 			@Nullable Query query
@@ -48,7 +57,33 @@ public List<Document> retrieve(
 		logger.debug("cleaner data: {}", cleanerData);
 
 		// 返回结果
-		return dataCleaner.limitResults(cleanerData, maxResults);
+		List<Document> documents = dataCleaner.limitResults(cleanerData, maxResults);
+
+		logger.debug("WebSearchRetriever#retrieve() document size: {}, raw documents: {}",
+				documents.size(),
+				documents.stream().map(Document::getId).toArray()
+		);
+
+		return enableRanker ? ranking(query, documents) : documents;
+	}
+
+	private List<Document> ranking(Query query, List<Document> documents) {
+
+		if (documents.size() == 1) {
+			// 只有一个时，不需要 rank
+			return documents;
+		}
+
+		try {
+
+			List<Document> rankedDocuments = documentRanker.rank(query, documents);
+			logger.debug("WebSearchRetriever#ranking() Ranked documents: {}", rankedDocuments.stream().map(Document::getId).toArray());
+			return rankedDocuments;
+		} catch (Exception e) {
+			// 降级返回原始结果
+			logger.error("ranking error", e);
+			return documents;
+		}
 	}
 
 	public static WebSearchRetriever.Builder builder() {
@@ -64,6 +99,11 @@ public static final class Builder {
 
 		private DataClean dataCleaner;
 
+		private DocumentRanker documentRanker;
+
+		// 默认开启 ranking
+		private Boolean enableRanker = true;
+
 		public WebSearchRetriever.Builder searchEngine(IQSSearchEngine searchEngine) {
 
 			this.searchEngine = searchEngine;
@@ -82,6 +122,16 @@ public WebSearchRetriever.Builder maxResults(int maxResults) {
 			return this;
 		}
 
+		public WebSearchRetriever.Builder documentRanker(DocumentRanker documentRanker) {
+			this.documentRanker = documentRanker;
+			return this;
+		}
+
+		public WebSearchRetriever.Builder enableRanker(Boolean enableRanker) {
+			this.enableRanker = enableRanker;
+			return this;
+		}
+
 		public WebSearchRetriever build() {
 
 			return new WebSearchRetriever(this);

spring-ai-alibaba-integration-example/backend/src/main/java/com/alibaba/cloud/ai/application/websearch/rag/postretrieval/DashScopeDocumentRanker.java

@@ -71,7 +71,7 @@ public List<Document> rank(
 							.filter(doc ->
 							{
 								// debug rerank output.
-								logger.debug("doc id: {}, outputDocs id: {}", doc.getId(), outputDocs.getId());
+								logger.debug("DashScopeDocumentRanker#rank() doc id: {}, outputDocs id: {}", doc.getId(), outputDocs.getId());
 								return Objects.equals(doc.getId(), outputDocs.getId());
 							})
 							.findFirst();

spring-ai-alibaba-integration-example/backend/src/main/java/com/alibaba/cloud/ai/application/websearch/rag/preretrieval/query/expansion/MultiQueryExpander.java

@@ -88,6 +88,8 @@ public List<Query> expand(@Nullable Query query) {
 				.call()
 				.content();
 
+		logger.debug("MultiQueryExpander#expand() Response from chat client: {}", resp);
+
 		if (Objects.isNull(resp)) {
 
 			logger.warn("No response from chat client for query: {}. is return.", query.text());

spring-ai-alibaba-integration-example/backend/src/main/java/com/alibaba/cloud/ai/application/websearch/rag/prompt/CustomContextQueryAugmenter.java

@@ -69,6 +69,10 @@ public CustomContextQueryAugmenter(
 		this.emptyPromptTemplate = emptyPromptTemplate != null ? emptyPromptTemplate : DEFAULT_EMPTY_PROMPT_TEMPLATE;
 		this.allowEmptyContext = allowEmptyContext != null ? allowEmptyContext : DEFAULT_ALLOW_EMPTY_PROMPT;
 
+		logger.debug("CustomContextQueryAugmenter promptTemplate: {}", promptTemplate.getTemplate());
+		logger.debug("CustomContextQueryAugmenter emptyPromptTemplate: {}", emptyPromptTemplate);
+		logger.debug("CustomContextQueryAugmenter allowEmptyContext: {}", allowEmptyContext);
+
 		PromptAssert.templateHasRequiredPlaceholders(this.promptTemplate, "query", "context");
 	}
 

spring-ai-alibaba-integration-example/backend/src/main/java/com/alibaba/cloud/ai/application/websearch/rag/prompt/PromptTemplateConfig.java

@@ -17,15 +17,18 @@ public PromptTemplate transformerPromptTemplate() {
 
 		return new PromptTemplate(
 				"""
-				Given a user query, rewrite it to provide better results when querying a {target}.
-				Remove any irrelevant information, and ensure the query is concise and specific.
+				给定一个用户查询，重写用户问题，使得在 {target} 查询时提供更好的结果。
 				
-				The output must be consistent with the user query's language.
+				你应该遵循以下规则：
 				
-				Original query:
+				1. 删除任何不相关的信息，并确保查询简洁具体；
+				2. 输出必须与用户查询的语言一致；
+				3. 保证在大模型的角度更好理解与做出回答。
+				
+				原始查询：
 				{query}
 				
-				Rewritten query:
+				重写之后的查询:
 				"""
 		);
 	}
@@ -36,38 +39,39 @@ public PromptTemplate queryArgumentPromptTemplate() {
 
 		return new PromptTemplate(
 				"""
-				You will be given a set of related document contexts to the question.
-				Each document starts with a reference number like [[x]], where x is a number that may be repeated.
-				Document without reference will be marked as [[null]].
-				Please use the context and cite the context at the end of each sentence if applicable.
-				Context information is below.
+				你将获得一组与问题相关的文档上下文。
+				每个文档都以参考编号开头，如 [[x]]，其中 x 是可以重复的数字。
+				没有引用的文档将被标记为 [[null]]。
+				请使用上下文，并在每个句子的末尾引用上下文（如果适用）。
+				上下文信息如下:
 				
 				---------------------
 				{context}
 				---------------------
 				
-				Given the context information and no prior knowledge, generate a structured response to the user's question.
+				在给定上下文信息且没有先验知识的情况下，生成对用户问题的结构化响应。
 				
-				Follow these rules:
+				在你回答用户问题时，请遵循以下规则：
 				
-				1. If the answer is not in the context, just say that you don't know.
-				2. Do not give any information that is not related to the question, and do not repeat.
-				3. Avoid statements like "Based on the context..." or "The provided information...".
-				4. Your answer must be correct, accurate and written by an expert using an unbiased and professional tone.
-				5. The appropriate structure is automatically determined based on the characteristics of the content, but be sure to include subheadings to enhance readability.
-				6. When generating a response, provide a clear conclusion or main idea first, and do not need to carry a title.
-				7. Make sure each section has clear subheadings so the reader can better understand and follow the content.
-				8. If the information is complex or has multiple sections, make sure each section has an appropriate title to create a hierarchical structure.
-				9. Please cite the sentence or section with the reference numbers in the end, in the format [[x]].
-				10. If a sentence or sections comes from multiple contexts, please list all applicable citations, like [[x]][[y]].
-				11. You don`t have to cite if the reference is marked as [[null]].
-				12. Other than code and specific names and citations, your answer must be written in the same language as the question.
-
-				Query: {query}
+				1. 如果答案不在上下文中，就说你不知道；
+				2. 不要提供任何与问题无关的信息，也不要输出任何的重复内容；
+				3. 避免使用 “基于上下文...” 或 “The provided information...” 的说法；
+				4. 你的答案必须正确、准确，并使用专家般公正和专业的语气撰写；
+				5. 回答中适当的文本结构是根据内容的特点来确定的，请在输出中包含副标题以提高可读性；
+				6. 生成回复时，先提供明确的结论或中心思想，不需要带有标题；
+				7. 确保每个部分都有清晰的副标题，以便用户可以更好地理解和参考你的输出内容；
+				8. 如果信息复杂或包含多个部分，请确保每个部分都有适当的标题以创建分层结构；
+				9. 请以 [x] 格式引用末尾带有参考编号的句子或部分；
+				10. 如果一个句子或章节来自多个上下文，请列出所有适用的引用，例如 [x][y]；
+				11. 你的输出答案必须保持美观且严谨的 markdown 格式。
+				12. 因为你的输出保持 markdown 格式，请在引用上下文时，以超链接的形式带上参考文档中的链接，方便用户点击查看；
+				13. 如果参考文献被标记为 [null]，则不必引用；
+				14. 除了代码。具体名称和引文外，你的答案必须用与问题相同的语言编写。
 				
-				Answer:
+				用户问题: {query}
+				
+				你的回答:
 				"""
 		);
 	}
-
 }