提示词适配

src/main/java/com/supervision/pdfqaserver/cache/PromptCache.java

@@ -191,15 +191,76 @@ public class PromptCache {
             """;
 
     private static final String TEXT_TO_CYPHER_PROMPT = """
-            结合给你的领域元数据，分析用户输入的问题，尝试将其转换为CYPHER语句。
-            领域元数据：{domainMetadata}
-            用户输入的问题：{userQuery}
+            你是一个专业的 Neo4j Cypher 查询语句生成器，专门用于构建针对特定结构的查询语句。
+             
+             ---
+             
+             **【数据库结构说明】**
+             
+             - **关系类型（relationType）**：
+             {relationTypeList}
+             
+             - **源节点类型（sourceType）**：
+             {sourceTypeList}
+             
+             - **目标节点类型（targetType）**：
+             {targetTypeList}
+             
+             ---
+             
+             **【生成规则】**
+             
+             1. 识别用户问题中的实体及意图，映射为 `Cypher 查询语句`
+             2. 使用无条件匹配，不假设任何属性名称，不添加 `WHERE` 子句过滤。
+             3. 返回所有满足该关系的查询语句，并包含节点和关系的**所有属性**。
+             4. 仅输出 Cypher 代码块，不附加任何解释。
+             5. 如无法从结构中推断 relationType、sourceType 或 targetType，输出：
+                ```
+                无法根据数据库结构生成查询
+                ```
+             
+             ---
+             
+             **【示例】**
+             
+             1. - **用户问题：** 查询“宝塔盛华商贸集团”的基本信息？
+                - **生成的 Cypher 查询：**
+                \\{
+                  "cypherQueries": [
+                    "MATCH (c:Company)-[r:HAS_LEGAL_REP]->(t) RETURN c, r, t",
+                    "MATCH (c:Company)-[r:HAS_PHONE]->(t) RETURN c, r, t",
+                    .....
+                  ]
+                \\}
+                
+             2. - **用户问题：** 查询“宝塔盛华商贸集团”出具的电子银行承兑汇票金额是多少？
+                - **生成的 Cypher 查询：**
+             	\\{
+             	  "cypherQueries": [
+             		"MATCH (c:Company)-[r:IssueDocument]->(t:FinancialBill) RETURN c, r, t",
+             		.....
+             	  ]
+             	\\}
+             
+             【用户问题】
+             {query}
+             
+             【你的任务】
+             根据以上数据库结构和用户问题，生成正确的Cypher查询语句。
            """;
 
     private static final String GENERATE_ANSWER_PROMPT = """
-            结合给你的三元组数据和用户输入的问题，生成一个简洁的回答。
-            三元组数据：{tripleMetaData}
-            用户输入的问题：{userQuery}
+            一、任务说明
+            你是一个审计报告的整理助手，你需要对用户询问的问题根据产考数据进行回答。
+            二、用户输入：
+            {query}
+            三、参考文本：
+            {example_text}
+            四、注意事项
+            1. 数据核实：需确认是否使用参考文本中的数据。
+            2. 信息时效：可能需要确认数据的时间。
+            3. 问题范围界定：当用户提问超出审计报告分析范畴时，统一采用以下话术回复：
+            "您好！当前系统功能聚焦于审计报告相关内容分析，您的问题暂不在支持范围内。如需查询财务数据、票据详情或其他审计相关信息，请提供具体问题，我们将全力协助。"
             """;
 
     private static final String CHINESE_TO_ENGLISH_PROMPT = """

src/main/java/com/supervision/pdfqaserver/config/OllamaChatModelAspect.java

@@ -26,20 +26,13 @@ public class OllamaChatModelAspect {
     @Around("execution(* org.springframework.ai.chat.model.ChatModel.call(..))")
     public Object aroundMethodExecution(ProceedingJoinPoint joinPoint) throws Throwable {
         String signature = joinPoint.getSignature().toString();
-        if (StrUtil.equals(model,"qwen3:30b-a3b") && StrUtil.equals(signature, callStringMessage)) {
-            Object[] args = joinPoint.getArgs();
-            if (args.length > 0) {
-                String arg = (String) args[0];
-                args[0] = arg + "\n /no_think";
-            }
-        }
         // 执行原方法
         Object result = joinPoint.proceed();
-
-        if (StrUtil.equals(model,"qwen3:30b-a3b") && StrUtil.equals(signature, callStringMessage)) {
+        if (StrUtil.equals(model,"qwen3:30b-a3b") ) {
+            if(StrUtil.equals(signature, callStringMessage)){
                 result = ((String) result).replaceAll("(?is)<think\\b[^>]*>(.*?)</think>", "").trim();
             }
-
+        }
         return result;
     }
 }

src/main/java/com/supervision/pdfqaserver/dao/Neo4jRepository.java

@@ -39,9 +39,9 @@ public class Neo4jRepository {
                 while (result.hasNext()) {
                     org.neo4j.driver.Record record = result.next();
                     // 从 Record 中取出三部分
-                    Node a = record.get("startNode").asNode();
+                    Node a = record.get("c").asNode();
                     Relationship r = record.get("r").asRelationship();
-                    Node b = record.get("endNode").asNode();
+                    Node b = record.get("t").asNode();
 
                     // 转成我们的 DTO
                     NodeData start = mapNode(a);

src/main/java/com/supervision/pdfqaserver/service/impl/ChatServiceImpl.java

@@ -1,5 +1,7 @@
 package com.supervision.pdfqaserver.service.impl;
 
+import cn.hutool.json.JSONObject;
+import cn.hutool.json.JSONUtil;
 import com.supervision.pdfqaserver.cache.PromptCache;
 import com.supervision.pdfqaserver.dao.Neo4jRepository;
 import com.supervision.pdfqaserver.domain.ChineseEnglishWords;
@@ -11,14 +13,13 @@ import com.supervision.pdfqaserver.service.DomainMetadataService;
 import lombok.RequiredArgsConstructor;
 import lombok.extern.slf4j.Slf4j;
 import org.springframework.ai.chat.messages.Message;
-import org.springframework.ai.chat.model.ChatResponse;
 import org.springframework.ai.chat.prompt.Prompt;
 import org.springframework.ai.chat.prompt.SystemPromptTemplate;
 import org.springframework.ai.ollama.OllamaChatModel;
 import org.springframework.stereotype.Service;
 import reactor.core.publisher.Flux;
 
-import java.util.HashMap;
+import java.util.ArrayList;
 import java.util.List;
 import java.util.Map;
 import java.util.stream.Collectors;
@@ -30,14 +31,16 @@ import static com.supervision.pdfqaserver.cache.PromptCache.TEXT_TO_CYPHER;
 @Service
 @RequiredArgsConstructor
 public class ChatServiceImpl implements ChatService {
-    private static final String PROMPT_PARAM_DOMAIN_METADATA = "domainMetadata";
-    private static final String PROMPT_PARAM_TRIPLE_METADATA = "tripleMetaData";
-    private static final String PROMPT_PARAM_USER_QUERY = "userQuery";
+    private static final String PROMPT_PARAM_SOURCE_TYPE_LIST = "sourceTypeList";
+    private static final String PROMPT_PARAM_RELATION_TYPE_LIST = "relationTypeList";
+    private static final String PROMPT_PARAM_TARGET_TYPE_LIST = "targetTypeList";
+    private static final String PROMPT_PARAM_EXAMPLE_TEXT = "example_text";
+    private static final String PROMPT_PARAM_QUERY = "query";
+    private static final String CYPHER_QUERIES = "cypherQueries";
 
     private final Neo4jRepository neo4jRepository;
 
     private final OllamaChatModel ollamaChatModel;
-
     private final DomainMetadataService domainMetadataService;
     private final ChineseEnglishWordsService chineseEnglishWordsService;
 
@@ -46,33 +49,54 @@ public class ChatServiceImpl implements ChatService {
         //拼装领域元数据
         Map<String, String> chineseEnglishWordsMap = chineseEnglishWordsService.list().stream()
                 .collect(Collectors.toMap(ChineseEnglishWords::getChineseWord, ChineseEnglishWords::getEnglishWord));
-        List<Map<String, String>> domainMappings = domainMetadataService.list().stream().map(domainMetadata -> {
-            Map<String, String> mapping = new HashMap<>();
-            mapping.put("source", domainMetadata.getSourceType());
-            mapping.put("sourceType", chineseEnglishWordsMap.get(domainMetadata.getSourceType()));
-            mapping.put("relation", domainMetadata.getRelation());
-            mapping.put("relationType", chineseEnglishWordsMap.get(domainMetadata.getRelation()));
-            mapping.put("target", domainMetadata.getTargetType());
-            mapping.put("targetType", chineseEnglishWordsMap.get(domainMetadata.getTargetType()));
-            return mapping;
-        }).toList();
-        log.info("domainMappings: {}", domainMappings);
-        //生成CYPHER
+
+        //分别得到sourceType，relation，targetType的group by后的集合
+        List<String> sourceTypeList = domainMetadataService.list().stream().map(DomainMetadata::getSourceType).distinct().toList();
+        List<String> relationList = domainMetadataService.list().stream().map(DomainMetadata::getRelation).distinct().toList();
+        List<String> targetTypeList = domainMetadataService.list().stream().map(DomainMetadata::getTargetType).distinct().toList();
+
+        //将三个集合分别结合chineseEnglishWordsMap的key转化为value集合
+        List<String> sourceTypeListEnList = sourceTypeList.stream().map(chineseEnglishWordsMap::get).toList();
+        List<String> relationListEnList = relationList.stream().map(chineseEnglishWordsMap::get).toList();
+        List<String> targetTypeListEnList = targetTypeList.stream().map(chineseEnglishWordsMap::get).toList();
+
+        //将三个集合分别转换为英文逗号分隔的字符串
+        String sourceTypeListEn = String.join(",", sourceTypeListEnList);
+        String relationListEn = String.join(",", relationListEnList);
+        String targetTypeListEn = String.join(",", targetTypeListEnList);
+        log.info("sourceTypeListEn: {}, relationListEn: {}, targetTypeListEn: {}", sourceTypeListEn, relationListEn, targetTypeListEn);
+
+        //LLM生成CYPHER
         SystemPromptTemplate textToCypherTemplate = new SystemPromptTemplate(PromptCache.promptMap.get(TEXT_TO_CYPHER));
-        Message textToCypherMessage = textToCypherTemplate.createMessage(Map.of(PROMPT_PARAM_DOMAIN_METADATA, domainMappings, PROMPT_PARAM_USER_QUERY, userQuery));
-        ChatResponse textToCypherResponse = ollamaChatModel.call(new Prompt(textToCypherMessage));
-        String queryCypher = "MATCH (startNode:公司)-[r]->(endNode) RETURN startNode,r,endNode";
-        log.info(textToCypherResponse.getResult().getOutput().getText());
-//        String queryCypher = textToCypherResponse.getResult().getOutput().getText();
-        List<RelationObject> relationObjects = neo4jRepository.execute(queryCypher, null);
+        Message textToCypherMessage = textToCypherTemplate.createMessage(Map.of(PROMPT_PARAM_SOURCE_TYPE_LIST, sourceTypeListEn, PROMPT_PARAM_RELATION_TYPE_LIST, relationListEn, PROMPT_PARAM_TARGET_TYPE_LIST, targetTypeListEn, PROMPT_PARAM_QUERY, userQuery));
+        String cypherJsonStr = ollamaChatModel.call(textToCypherMessage.getText());
+        log.info(cypherJsonStr);
+        List<String> cypherQueries;
+        try {
+            JSONObject jsonObj = JSONUtil.parseObj(cypherJsonStr);
+            cypherQueries = jsonObj.getJSONArray(CYPHER_QUERIES)
+                    .toList(String.class);
+        } catch (Exception e) {
+            log.error("解析CYPHER JSON字符串失败: {}", e.getMessage());
+            return Flux.just("查无结果");
+        }
+        log.info("转换后的Cypher语句：{}", cypherQueries.toString());
+
+        //执行CYPHER查询并汇总结果
+        List<RelationObject> relationObjects = new ArrayList<>();
+        if (!cypherQueries.isEmpty()) {
+            for (String cypher : cypherQueries) {
+                relationObjects.addAll(neo4jRepository.execute(cypher, null));
+            }
+        }
         if (relationObjects.isEmpty()) {
-            return Flux.just("没有找到相关数据");
+            return Flux.just("查无结果");
         }
-        log.info("relationObjects: {}", relationObjects);
+        log.info("三元组数据: {}", relationObjects);
 
         //生成回答
         SystemPromptTemplate generateAnswerTemplate = new SystemPromptTemplate(PromptCache.promptMap.get(GENERATE_ANSWER));
-        Message generateAnswerMessage = generateAnswerTemplate.createMessage(Map.of(PROMPT_PARAM_TRIPLE_METADATA, relationObjects, PROMPT_PARAM_USER_QUERY, userQuery));
+        Message generateAnswerMessage = generateAnswerTemplate.createMessage(Map.of(PROMPT_PARAM_EXAMPLE_TEXT, relationObjects, PROMPT_PARAM_QUERY, userQuery));
         return ollamaChatModel.stream(new Prompt(generateAnswerMessage)).map(response -> response.getResult().getOutput().getText());
     }
 }