|
|
package com.supervision.springaidemo.controller;
|
|
|
|
|
|
import cn.hutool.core.io.FileUtil;
|
|
|
import com.supervision.police.domain.ModelRecordType;
|
|
|
import com.supervision.police.domain.NoteRecordSplit;
|
|
|
import com.supervision.springaidemo.dto.QARecordNodeDTO;
|
|
|
import com.supervision.police.mapper.ModelRecordTypeMapper;
|
|
|
import com.supervision.police.mapper.NoteRecordSplitMapper;
|
|
|
import com.supervision.police.service.ModelRecordTypeService;
|
|
|
import com.supervision.springaidemo.service.ModelMetricService;
|
|
|
import com.supervision.police.service.RecordService;
|
|
|
import com.supervision.police.service.NoteCheckRecordService;
|
|
|
import com.supervision.springaidemo.util.RecordRegexUtil;
|
|
|
import com.supervision.springaidemo.util.WordReadUtil;
|
|
|
import lombok.extern.slf4j.Slf4j;
|
|
|
import org.apache.ibatis.annotations.Param;
|
|
|
import org.json.JSONObject;
|
|
|
import org.springframework.ai.chat.ChatResponse;
|
|
|
import org.springframework.ai.chat.Generation;
|
|
|
import org.springframework.ai.chat.messages.Message;
|
|
|
import org.springframework.ai.chat.messages.SystemMessage;
|
|
|
import org.springframework.ai.chat.prompt.Prompt;
|
|
|
import org.springframework.ai.ollama.OllamaChatClient;
|
|
|
import org.springframework.beans.factory.annotation.Autowired;
|
|
|
import org.springframework.util.StopWatch;
|
|
|
import org.springframework.web.bind.annotation.GetMapping;
|
|
|
import org.springframework.web.bind.annotation.RequestMapping;
|
|
|
import org.springframework.web.bind.annotation.RestController;
|
|
|
|
|
|
import java.io.File;
|
|
|
import java.time.LocalDateTime;
|
|
|
import java.util.*;
|
|
|
import java.util.stream.Collectors;
|
|
|
|
|
|
/**
|
|
|
* 目前这是效果最好的方案,暂时先用这个
|
|
|
*/
|
|
|
@RestController
|
|
|
@Slf4j
|
|
|
@RequestMapping("exampleChat")
|
|
|
public class ExampleChatController {
|
|
|
|
|
|
private final OllamaChatClient chatClient;
|
|
|
|
|
|
@Autowired
|
|
|
private ModelMetricService modelMetricService;
|
|
|
|
|
|
@Autowired
|
|
|
private NoteCheckRecordService noteCheckRecordService;
|
|
|
@Autowired
|
|
|
private ModelRecordTypeMapper modelRecordTypeMapper;
|
|
|
@Autowired
|
|
|
private NoteRecordSplitMapper noteRecordSplitMapper;
|
|
|
|
|
|
@Autowired
|
|
|
public ExampleChatController(OllamaChatClient chatClient) {
|
|
|
this.chatClient = chatClient;
|
|
|
}
|
|
|
|
|
|
@Autowired
|
|
|
private RecordService recordService;
|
|
|
|
|
|
@Autowired
|
|
|
private ModelRecordTypeService modelRecordTypeService;
|
|
|
|
|
|
private static final String exampleTemplate = """
|
|
|
我们现在需要以step by step的方式进行笔录的指标分析工作,得到最终的结果并返回。
|
|
|
|
|
|
step1:分析笔录的内容;
|
|
|
以下是笔录的内容:
|
|
|
---
|
|
|
{context}
|
|
|
---
|
|
|
|
|
|
step2:现在给你指标以及指标的释义或例子可协助你参考:
|
|
|
指标释义或例子及判断标准:
|
|
|
指标名称:{metricName}
|
|
|
我现在给你为true的例子可供参考:{example};
|
|
|
如果和例子这种情况相反或偏差较大则为false;
|
|
|
如果笔录中,没有任何笔录内容涉及到该项指标,无法进行判断,则为empty。
|
|
|
|
|
|
step3:现在需要你根据上面提供的所有信息,尽可能实事求是完成判断:
|
|
|
1.判断结论:true/false/empty
|
|
|
2.得到结论的笔录原话:从笔录的对话中,得到该结论的原文(一定是摘抄的原文且为中文)。如果结论为true,则必须要有原文佐证!
|
|
|
3.得到结论的原因:分析得出该结论的原因,需明确说明为什么得到该结论,需要参考例子的格式且为中文回复。如果结论为true/false,则必须有原因!
|
|
|
|
|
|
step4:必须以json格式回复, JSON的value内容我给你的提示,在实际输出的时候不需要带上:
|
|
|
---
|
|
|
{"result":"结论", "originalContext":"笔录对应原话","reason":"原因"}
|
|
|
---
|
|
|
好了,现在可以回复了!
|
|
|
""";
|
|
|
|
|
|
@GetMapping("exampleChat")
|
|
|
public void exampleChat() {
|
|
|
File file = FileUtil.file("E:\\jc\\宁夏\\Fw_裴金禄\\裴金禄第一次.docx");
|
|
|
String context = WordReadUtil.readWord(file.getPath());
|
|
|
List<QARecordNodeDTO> qaList = RecordRegexUtil.recordRegex(context, "裴金禄");
|
|
|
for (QARecordNodeDTO qa : qaList) {
|
|
|
NoteRecordSplit noteRecord = new NoteRecordSplit();
|
|
|
noteRecord.setNoteName(file.getName());
|
|
|
noteRecord.setPersonName("裴金禄");
|
|
|
noteRecord.setQuestion(qa.getQuestion());
|
|
|
noteRecord.setAnswer(qa.getAnswer());
|
|
|
noteRecord.setCreateTime(LocalDateTime.now());
|
|
|
|
|
|
// if (qa.getAnswer().length() > 500) {
|
|
|
// continue;
|
|
|
// }
|
|
|
//根据提示词分析类型 并保存
|
|
|
// String test = "你现在是一个笔录分析人员，请用四个字描述一下下述内容属于哪种类型的对话？";
|
|
|
String test = "你是一个善于归纳分析问讯笔录的大模型，请用四个字描述下述内容属于哪种类型的对话？对话内容为：";
|
|
|
String tip = "你的回答必须严格控制为四个字。";
|
|
|
String example = "示例1：办案警官问:我们是中卫市公安局沙坡头区分局经侦大队的民警(出示工作证件)，现依法对你进行讯问，你应当如实回答我们的提问，对与案件无关的问题，你有拒绝回答的权利。你听明白了吗? 张某答:听明白了。回复：{\"type\":\"权利义务\"}。" +
|
|
|
"示例2：办案警官问：你的个人情况？张某答：我叫裴金禄，男，1984年01月10日出生，藏族，大学文化，户籍所在地陕西省西安市雁塔区丈八四路6号7号楼1单元 2304室，现住陕西省西安市雁塔区丈八四路6号7号楼1单元 2304室，无业，居民身份证号码622326198401100090，联系电话15829319393。回复：{\"type\":\"个人情况\"}。" +
|
|
|
"示例3：办案警官问：你的家庭情况？张某答：我父亲叫：裴国智，今年：67岁，退休，电话：13519353723; 母亲：李金芬，今年：61岁，无业；我弟弟裴金生，今年： 34岁，在中石油运输公司上班。回复：{\"type\":\"家庭情况\"}。";
|
|
|
List<Message> messages = new ArrayList<>(List.of(new SystemMessage(test + qa.getQuestion() + qa.getAnswer() + tip + example)));
|
|
|
Prompt prompt = new Prompt(messages);
|
|
|
StopWatch stopWatch = new StopWatch();
|
|
|
stopWatch.start();
|
|
|
// log.info("开始分析:");
|
|
|
ChatResponse call = chatClient.call(prompt);
|
|
|
stopWatch.stop();
|
|
|
// log.info("耗时:{}", stopWatch.getTotalTimeSeconds());
|
|
|
Generation result = call.getResult();
|
|
|
String content = result.getOutput().getContent();
|
|
|
// log.info("问：{}, 答：{}", qa.getQuestion(), qa.getAnswer());
|
|
|
// log.info("分析的结果是:{}", content);
|
|
|
JSONObject jsonObject = new JSONObject(content);
|
|
|
String type = jsonObject.getString("type").trim();
|
|
|
System.out.println("问："+qa.getQuestion()+"答："+qa.getAnswer());
|
|
|
System.out.println("分析的结果是:"+ type);
|
|
|
|
|
|
//保存笔录
|
|
|
noteRecord.setRecordType(type);
|
|
|
recordService.save(noteRecord);
|
|
|
|
|
|
ModelRecordType exist = modelRecordTypeService.queryByName(type);
|
|
|
if (exist == null) {
|
|
|
ModelRecordType modelRecordType = new ModelRecordType();
|
|
|
modelRecordType.setRecordType(type);
|
|
|
modelRecordTypeService.save(modelRecordType);
|
|
|
}
|
|
|
}
|
|
|
|
|
|
// for (QARecordNodeDTO qaRecordNodeDTO : qaList) {
|
|
|
// // 只查入罪指标
|
|
|
// List<ModelMetric> list = modelMetricService.lambdaQuery().likeRight(ModelMetric::getMetricCode, "RZ").list();
|
|
|
// for (ModelMetric modelMetric : list) {
|
|
|
// String systemPrompt = """
|
|
|
// 你是一个善于分析办案笔录的模型,能够根据办案笔录的回答内容,结合给定的例子,实事求是的判断给定指标是否满足。注意,仅根据笔录进行分析,不要做笔录之外的推断。笔录内容可能比较长,可能分多次提交给你。Think step by step
|
|
|
// """;
|
|
|
// List<Message> messages = new ArrayList<>(List.of(new SystemMessage(systemPrompt)));
|
|
|
// Map<String, Object> param = new HashMap<>();
|
|
|
// param.put("context", qaRecordNodeDTO.toString());
|
|
|
// param.put("metricName", modelMetric.getMetricName());
|
|
|
// param.put("example", StrUtil.format(modelMetric.getExample(), MapUtil.of("action", "裴金禄")));
|
|
|
// String format = StrUtil.format(exampleTemplate, param);
|
|
|
// List<Message> userMessageList = new ArrayList<>();
|
|
|
// if (format.length() > 8000) {
|
|
|
// log.info("分段提交");
|
|
|
// for (String s : StrUtil.split(format, 6000)) {
|
|
|
// userMessageList.add(new UserMessage(s));
|
|
|
// userMessageList.add(new AssistantMessage("继续"));
|
|
|
// }
|
|
|
// userMessageList.remove(userMessageList.size() - 1);
|
|
|
// } else {
|
|
|
// userMessageList.add(new UserMessage(format));
|
|
|
// }
|
|
|
// messages.addAll(userMessageList);
|
|
|
//
|
|
|
// RunCheckThread runCheck = new RunCheckThread("裴金禄尝试正则来做", chatClient, noteCheckRecordService, new Prompt(messages), FileUtil.getName(file), format, systemPrompt, modelMetric, 0);
|
|
|
// RunCheckThreadPool.chatExecutor.submit(runCheck);
|
|
|
// }
|
|
|
// }
|
|
|
|
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
@GetMapping("test")
|
|
|
public List test(@Param("test") String test) {
|
|
|
List<Map> list = new ArrayList<>();
|
|
|
List<Message> messages = new ArrayList<>(List.of(new SystemMessage(test)));
|
|
|
Prompt prompt = new Prompt(messages);
|
|
|
StopWatch stopWatch = new StopWatch();
|
|
|
stopWatch.start();
|
|
|
log.info("开始分析:");
|
|
|
ChatResponse call = chatClient.call(prompt);
|
|
|
stopWatch.stop();
|
|
|
log.info("耗时:{}", stopWatch.getTotalTimeSeconds());
|
|
|
Generation result = call.getResult();
|
|
|
|
|
|
String content = result.getOutput().getContent();
|
|
|
log.info("分析的结果是:{}", content);
|
|
|
Map<String, Object> resultMap = new HashMap<>();
|
|
|
resultMap.put("result", content);
|
|
|
list.add(resultMap);
|
|
|
return list;
|
|
|
}
|
|
|
|
|
|
@GetMapping("queryRecordType")
|
|
|
public List<NoteRecordSplit> queryRecordType() {
|
|
|
List<ModelRecordType> types = modelRecordTypeMapper.selectList(null);
|
|
|
String allType = types.stream().map(ModelRecordType::getRecordType).collect(Collectors.joining("."));
|
|
|
List<NoteRecordSplit> list = noteRecordSplitMapper.selectList(null);
|
|
|
for (NoteRecordSplit record : list) {
|
|
|
record.setRecordType("");
|
|
|
|
|
|
String test = "你是一个善于总结问讯内容的大模型，请判断以下对话属于【" + allType + "】哪个分类?对话内容为：";
|
|
|
String tip = "你的回答必须在所给范围内。";
|
|
|
String example = "示例：办案警官问：你的家庭情况？答：裴金禄回答：我父亲叫：裴国智，今年：67岁，退休，电话：13519353723;母亲：李金芬，今年：61岁，无业；我弟弟裴金生，今年：34岁，在中石油运输公司上班。回复：{\"type\":\"家庭情况\"}。";
|
|
|
List<Message> messages = new ArrayList<>(List.of(new SystemMessage(test + record.getQuestion() + record.getAnswer() + tip + example)));
|
|
|
Prompt prompt = new Prompt(messages);
|
|
|
StopWatch stopWatch = new StopWatch();
|
|
|
stopWatch.start();
|
|
|
// log.info("开始分析:");
|
|
|
ChatResponse call = chatClient.call(prompt);
|
|
|
stopWatch.stop();
|
|
|
// log.info("耗时:{}", stopWatch.getTotalTimeSeconds());
|
|
|
Generation result = call.getResult();
|
|
|
String content = result.getOutput().getContent();
|
|
|
// log.info("问：{}, 答：{}", qa.getQuestion(), qa.getAnswer());
|
|
|
// log.info("分析的结果是:{}", content);
|
|
|
System.out.println("问："+record.getQuestion()+"答："+record.getAnswer());
|
|
|
System.out.println("分析的结果是:"+ content);
|
|
|
record.setRecordType(content);
|
|
|
}
|
|
|
return list;
|
|
|
}
|
|
|
|
|
|
@GetMapping("test1")
|
|
|
public void test2(@Param("id") String id) {
|
|
|
NoteRecordSplit noteRecord = recordService.getById(id);
|
|
|
String question = noteRecord.getQuestion();
|
|
|
String answer = noteRecord.getAnswer();
|
|
|
String test = "请从以下对话中提取所有关于" + noteRecord.getRecordType() + "的所有三元组";
|
|
|
// String test = "请从以下对话中提取所有三元组，对话内容为：";
|
|
|
String example = "。示例：办案警官问：你的家庭情况？答：裴金禄回答：我父亲叫：裴国智，今年：67岁，电话：13519353723;母亲：李金芬，今年：61岁；我弟弟裴金生，今年：34岁。" +
|
|
|
"回复：{\"result\":[{\"主体\":\"裴金禄\",\"关系\":\"父子\",\"客体\":\"裴国智\"},{\"主体\":\"裴金禄\",\"关系\":\"母子\",\"客体\":\"李金芬\"},{\"主体\":\"裴金禄\",\"关系\":\"兄弟\",\"客体\":\"裴金生\"},{\"主体\":\"裴国智\",\"关系\":\"年龄\",\"客体\":67},{\"主体\":\"李金芬\",\"关系\":\"年龄\",\"客体\":61},{\"主体\":\"裴金生\",\"关系\":\"年龄\",\"客体\":34}]}。";
|
|
|
String a = "你是一个问讯笔录分析师，请分析以下内容中所有规定的三元组信息并补充完整，规定的三元组信息如下：[{startNodeType:'Person',entity:'',endNodeType:'Person',property:'',value:'' },{startNodeType:'Person',entity:'',endNodeType:'Organ',property:'',value:''}]，要求返回格式{result:[三元组信息]}。";
|
|
|
List<Message> messages = new ArrayList<>(List.of(new SystemMessage(a + question+answer)));
|
|
|
// List<Message> messages = new ArrayList<>(List.of(new SystemMessage(test + question+answer+example)));
|
|
|
Prompt prompt = new Prompt(messages);
|
|
|
StopWatch stopWatch = new StopWatch();
|
|
|
stopWatch.start();
|
|
|
log.info("开始分析:");
|
|
|
ChatResponse call = chatClient.call(prompt);
|
|
|
stopWatch.stop();
|
|
|
log.info("耗时:{}", stopWatch.getTotalTimeSeconds());
|
|
|
Generation result = call.getResult();
|
|
|
|
|
|
String content = result.getOutput().getContent();
|
|
|
log.info("分析的结果是:{}", content);
|
|
|
}
|
|
|
|
|
|
}
|