feat: 接入 DeepSeek，新增 AI 分类建议、自然语言查询与智能记账
- 后端新增 AI 模块：DeepSeek 客户端、用户级配置表 t_ai_config、用量记录表 t_ai_usage - 批量设置支持 AI 推荐分类与标签，并可一键应用 - 交易列表支持自然语言查询，解析为筛选条件确认后应用 - 支持自然语言记账，解析后预填新增表单 - 新增 AI 设置页：启用开关、API Key、接口地址、模型、连通性测试与用量统计 - 交易列表重排：AI 查询区、已激活筛选标签、备注与查询历史缓存 - 账单导入默认分类下拉补充分类图标 - 顶部用户入口改为下拉菜单（AI 设置 / 退出登录）
This commit is contained in:
@@ -0,0 +1,140 @@
|
||||
package cn.rosercode.finance.client;
|
||||
|
||||
import cn.rosercode.finance.common.BusinessException;
|
||||
import com.fasterxml.jackson.core.JsonProcessingException;
|
||||
import com.fasterxml.jackson.databind.JsonNode;
|
||||
import com.fasterxml.jackson.databind.ObjectMapper;
|
||||
import org.springframework.stereotype.Component;
|
||||
|
||||
import java.io.IOException;
|
||||
import java.net.URI;
|
||||
import java.net.http.HttpClient;
|
||||
import java.net.http.HttpRequest;
|
||||
import java.net.http.HttpResponse;
|
||||
import java.nio.charset.StandardCharsets;
|
||||
import java.time.Duration;
|
||||
import java.util.LinkedHashMap;
|
||||
import java.util.List;
|
||||
import java.util.Map;
|
||||
|
||||
/**
|
||||
* DeepSeek 客户端，封装 Chat Completions 接口的 HTTP 调用与响应解析。
|
||||
*
|
||||
* @date 2026-09-17 17:52:17
|
||||
*/
|
||||
@Component
|
||||
public class DeepSeekClient {
|
||||
|
||||
/** 单次请求的读取超时时间 */
|
||||
private static final Duration REQUEST_TIMEOUT = Duration.ofSeconds(60);
|
||||
|
||||
/** 连接超时时间 */
|
||||
private static final Duration CONNECT_TIMEOUT = Duration.ofSeconds(10);
|
||||
|
||||
/** 错误信息中保留的响应体最大长度 */
|
||||
private static final int ERROR_BODY_LIMIT = 300;
|
||||
|
||||
private final ObjectMapper objectMapper;
|
||||
private final HttpClient httpClient;
|
||||
|
||||
/**
|
||||
* 构造方法，初始化 JSON 映射器与 HTTP 客户端。
|
||||
*
|
||||
* @param objectMapper Jackson 对象映射器
|
||||
*/
|
||||
public DeepSeekClient(ObjectMapper objectMapper) {
|
||||
this.objectMapper = objectMapper;
|
||||
this.httpClient = HttpClient.newBuilder()
|
||||
.connectTimeout(CONNECT_TIMEOUT)
|
||||
.build();
|
||||
}
|
||||
|
||||
/**
|
||||
* 调用 Chat Completions 接口，返回模型输出的文本内容。
|
||||
*
|
||||
* @param baseUrl API 基础地址
|
||||
* @param apiKey API Key
|
||||
* @param model 模型名称
|
||||
* @param systemPrompt 系统提示词
|
||||
* @param userPrompt 用户提示词
|
||||
* @return 模型输出内容与 token 用量
|
||||
*/
|
||||
public DeepSeekResponse chat(String baseUrl, String apiKey, String model, String systemPrompt, String userPrompt) {
|
||||
// 兼容以斜杠结尾的 baseUrl
|
||||
String normalized = baseUrl.endsWith("/") ? baseUrl.substring(0, baseUrl.length() - 1) : baseUrl;
|
||||
String url = normalized + "/chat/completions";
|
||||
|
||||
// 构造请求体，强制返回 JSON 对象
|
||||
Map<String, Object> body = new LinkedHashMap<>();
|
||||
body.put("model", model);
|
||||
body.put("temperature", 0.2);
|
||||
body.put("stream", false);
|
||||
body.put("response_format", Map.of("type", "json_object"));
|
||||
body.put("messages", List.of(
|
||||
Map.of("role", "system", "content", systemPrompt),
|
||||
Map.of("role", "user", "content", userPrompt)
|
||||
));
|
||||
|
||||
String requestBody;
|
||||
try {
|
||||
requestBody = objectMapper.writeValueAsString(body);
|
||||
} catch (JsonProcessingException e) {
|
||||
throw new BusinessException("构建 AI 请求失败: " + e.getMessage());
|
||||
}
|
||||
|
||||
HttpRequest request = HttpRequest.newBuilder(URI.create(url))
|
||||
.timeout(REQUEST_TIMEOUT)
|
||||
.header("Authorization", "Bearer " + apiKey)
|
||||
.header("Content-Type", "application/json")
|
||||
.POST(HttpRequest.BodyPublishers.ofString(requestBody, StandardCharsets.UTF_8))
|
||||
.build();
|
||||
|
||||
HttpResponse<String> response;
|
||||
try {
|
||||
response = httpClient.send(request, HttpResponse.BodyHandlers.ofString(StandardCharsets.UTF_8));
|
||||
} catch (IOException e) {
|
||||
throw new BusinessException("调用 DeepSeek 失败: " + e.getMessage());
|
||||
} catch (InterruptedException e) {
|
||||
// 恢复中断标记，避免吞掉中断信号
|
||||
Thread.currentThread().interrupt();
|
||||
throw new BusinessException("调用 DeepSeek 被中断");
|
||||
}
|
||||
|
||||
if (response.statusCode() != 200) {
|
||||
throw new BusinessException("DeepSeek 返回异常(" + response.statusCode() + "): " + truncate(response.body()));
|
||||
}
|
||||
|
||||
// 解析 choices[0].message.content 与 usage 用量
|
||||
try {
|
||||
JsonNode root = objectMapper.readTree(response.body());
|
||||
JsonNode content = root.path("choices").path(0).path("message").path("content");
|
||||
if (content.isMissingNode() || content.isNull() || content.asText().isBlank()) {
|
||||
throw new BusinessException("DeepSeek 返回内容为空");
|
||||
}
|
||||
JsonNode usage = root.path("usage");
|
||||
DeepSeekResponse result = new DeepSeekResponse();
|
||||
result.setContent(content.asText());
|
||||
result.setPromptTokens(usage.path("prompt_tokens").asInt(0));
|
||||
result.setCompletionTokens(usage.path("completion_tokens").asInt(0));
|
||||
result.setTotalTokens(usage.path("total_tokens").asInt(0));
|
||||
// DeepSeek 用 prompt_cache_hit_tokens 表示缓存命中的输入 token
|
||||
result.setCachedTokens(usage.path("prompt_cache_hit_tokens").asInt(0));
|
||||
return result;
|
||||
} catch (JsonProcessingException e) {
|
||||
throw new BusinessException("解析 DeepSeek 响应失败: " + e.getMessage());
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 截断过长的响应体，避免错误信息过长。
|
||||
*
|
||||
* @param body 响应体
|
||||
* @return 截断后的字符串
|
||||
*/
|
||||
private String truncate(String body) {
|
||||
if (body == null) {
|
||||
return "";
|
||||
}
|
||||
return body.length() <= ERROR_BODY_LIMIT ? body : body.substring(0, ERROR_BODY_LIMIT) + "...";
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,27 @@
|
||||
package cn.rosercode.finance.client;
|
||||
|
||||
import lombok.Data;
|
||||
|
||||
/**
|
||||
* DeepSeek 调用结果，包含模型输出内容与 token 用量。
|
||||
*
|
||||
* @date 2026-09-17 18:02:08
|
||||
*/
|
||||
@Data
|
||||
public class DeepSeekResponse {
|
||||
|
||||
/** 模型输出的文本内容 */
|
||||
private String content;
|
||||
|
||||
/** 输入 token 数 */
|
||||
private int promptTokens;
|
||||
|
||||
/** 输出 token 数 */
|
||||
private int completionTokens;
|
||||
|
||||
/** 总 token 数 */
|
||||
private int totalTokens;
|
||||
|
||||
/** 缓存命中的输入 token 数 */
|
||||
private int cachedTokens;
|
||||
}
|
||||
@@ -0,0 +1,177 @@
|
||||
package cn.rosercode.finance.controller;
|
||||
|
||||
import cn.rosercode.finance.common.Result;
|
||||
import cn.rosercode.finance.dto.AiApplyRequest;
|
||||
import cn.rosercode.finance.dto.AiConfigRequest;
|
||||
import cn.rosercode.finance.dto.AiQueryRequest;
|
||||
import cn.rosercode.finance.dto.AiRecordRequest;
|
||||
import cn.rosercode.finance.dto.AiSuggestRequest;
|
||||
import cn.rosercode.finance.service.AiConfigService;
|
||||
import cn.rosercode.finance.service.AiQueryService;
|
||||
import cn.rosercode.finance.service.AiRecordService;
|
||||
import cn.rosercode.finance.service.AiSuggestionService;
|
||||
import cn.rosercode.finance.service.AiUsageService;
|
||||
import cn.rosercode.finance.vo.AiConfigVO;
|
||||
import cn.rosercode.finance.vo.AiQueryResultVO;
|
||||
import cn.rosercode.finance.vo.AiRecordResultVO;
|
||||
import cn.rosercode.finance.vo.AiSuggestionVO;
|
||||
import cn.rosercode.finance.vo.AiUsageResultVO;
|
||||
import io.swagger.v3.oas.annotations.Operation;
|
||||
import io.swagger.v3.oas.annotations.tags.Tag;
|
||||
import jakarta.servlet.http.HttpServletRequest;
|
||||
import jakarta.validation.Valid;
|
||||
import org.springframework.web.bind.annotation.GetMapping;
|
||||
import org.springframework.web.bind.annotation.PostMapping;
|
||||
import org.springframework.web.bind.annotation.PutMapping;
|
||||
import org.springframework.web.bind.annotation.RequestBody;
|
||||
import org.springframework.web.bind.annotation.RequestMapping;
|
||||
import org.springframework.web.bind.annotation.RequestParam;
|
||||
import org.springframework.web.bind.annotation.RestController;
|
||||
|
||||
import java.util.List;
|
||||
|
||||
/**
|
||||
* AI 控制器
|
||||
*
|
||||
* <p>提供 AI 配置的读写、连通性测试以及交易记录分类与标签智能建议接口</p>
|
||||
*
|
||||
* @date 2026-09-17 17:52:40
|
||||
*/
|
||||
@RestController
|
||||
@RequestMapping("/api/ai")
|
||||
@Tag(name = "AI 智能建议", description = "DeepSeek 配置与交易分类标签建议相关接口")
|
||||
public class AiController {
|
||||
|
||||
private final AiConfigService aiConfigService;
|
||||
private final AiSuggestionService aiSuggestionService;
|
||||
private final AiUsageService aiUsageService;
|
||||
private final AiQueryService aiQueryService;
|
||||
private final AiRecordService aiRecordService;
|
||||
|
||||
public AiController(AiConfigService aiConfigService, AiSuggestionService aiSuggestionService,
|
||||
AiUsageService aiUsageService, AiQueryService aiQueryService,
|
||||
AiRecordService aiRecordService) {
|
||||
this.aiConfigService = aiConfigService;
|
||||
this.aiSuggestionService = aiSuggestionService;
|
||||
this.aiUsageService = aiUsageService;
|
||||
this.aiQueryService = aiQueryService;
|
||||
this.aiRecordService = aiRecordService;
|
||||
}
|
||||
|
||||
/**
|
||||
* 获取当前用户的 AI 配置
|
||||
*
|
||||
* @param request HTTP请求，从中获取用户ID
|
||||
* @return AI 配置（API Key 脱敏）
|
||||
*/
|
||||
@GetMapping("/config")
|
||||
@Operation(summary = "获取 AI 配置", description = "获取当前用户的 AI 智能建议配置，API Key 已脱敏")
|
||||
public Result<AiConfigVO> getConfig(HttpServletRequest request) {
|
||||
Long userId = (Long) request.getAttribute("userId");
|
||||
return Result.success(aiConfigService.getConfig(userId));
|
||||
}
|
||||
|
||||
/**
|
||||
* 保存当前用户的 AI 配置
|
||||
*
|
||||
* @param request HTTP请求，从中获取用户ID
|
||||
* @param configRequest 配置保存请求
|
||||
* @return 操作结果
|
||||
*/
|
||||
@PutMapping("/config")
|
||||
@Operation(summary = "保存 AI 配置", description = "更新当前用户的 AI 配置，API Key 留空表示不修改")
|
||||
public Result<Void> saveConfig(HttpServletRequest request, @RequestBody AiConfigRequest configRequest) {
|
||||
Long userId = (Long) request.getAttribute("userId");
|
||||
aiConfigService.saveConfig(userId, configRequest);
|
||||
return Result.success();
|
||||
}
|
||||
|
||||
/**
|
||||
* 测试 DeepSeek 连通性
|
||||
*
|
||||
* @param request HTTP请求，从中获取用户ID
|
||||
* @return 操作结果
|
||||
*/
|
||||
@PostMapping("/config/test")
|
||||
@Operation(summary = "测试连接", description = "使用当前配置向 DeepSeek 发送一条测试请求")
|
||||
public Result<Void> testConfig(HttpServletRequest request) {
|
||||
Long userId = (Long) request.getAttribute("userId");
|
||||
aiConfigService.testConnection(userId);
|
||||
return Result.success();
|
||||
}
|
||||
|
||||
/**
|
||||
* 生成分类与标签建议
|
||||
*
|
||||
* @param request HTTP请求，从中获取用户ID
|
||||
* @param suggestRequest 建议请求，携带记录ID列表
|
||||
* @return 逐条建议列表
|
||||
*/
|
||||
@PostMapping("/suggest")
|
||||
@Operation(summary = "生成 AI 建议", description = "根据交易记录内容生成分类与标签建议")
|
||||
public Result<List<AiSuggestionVO>> suggest(HttpServletRequest request,
|
||||
@RequestBody @Valid AiSuggestRequest suggestRequest) {
|
||||
Long userId = (Long) request.getAttribute("userId");
|
||||
return Result.success(aiSuggestionService.suggest(userId, suggestRequest.getIds()));
|
||||
}
|
||||
|
||||
/**
|
||||
* 一键应用 AI 建议
|
||||
*
|
||||
* @param request HTTP请求，从中获取用户ID
|
||||
* @param applyRequest 应用请求，携带逐条建议
|
||||
* @return 实际发生变更的记录数
|
||||
*/
|
||||
@PostMapping("/apply")
|
||||
@Operation(summary = "应用 AI 建议", description = "覆盖分类并追加标签，一键写入建议")
|
||||
public Result<Integer> apply(HttpServletRequest request,
|
||||
@RequestBody @Valid AiApplyRequest applyRequest) {
|
||||
Long userId = (Long) request.getAttribute("userId");
|
||||
return Result.success(aiSuggestionService.apply(userId, applyRequest));
|
||||
}
|
||||
|
||||
/**
|
||||
* 获取 AI 使用情况
|
||||
*
|
||||
* @param request HTTP请求，从中获取用户ID
|
||||
* @param limit 最近明细条数，默认 20
|
||||
* @return 累计汇总与最近调用明细
|
||||
*/
|
||||
@GetMapping("/usage")
|
||||
@Operation(summary = "获取 AI 使用情况", description = "获取当前用户 AI 调用的累计汇总与最近调用明细")
|
||||
public Result<AiUsageResultVO> usage(HttpServletRequest request,
|
||||
@RequestParam(defaultValue = "20") int limit) {
|
||||
Long userId = (Long) request.getAttribute("userId");
|
||||
return Result.success(aiUsageService.getUsage(userId, limit));
|
||||
}
|
||||
|
||||
/**
|
||||
* 解析自然语言查询为筛选条件
|
||||
*
|
||||
* @param request HTTP请求，从中获取用户ID
|
||||
* @param queryRequest 查询请求，携带自然语言语句
|
||||
* @return 结构化筛选条件
|
||||
*/
|
||||
@PostMapping("/query")
|
||||
@Operation(summary = "自然语言查询解析", description = "将自然语言查询解析为交易筛选条件，供用户确认后应用")
|
||||
public Result<AiQueryResultVO> query(HttpServletRequest request,
|
||||
@RequestBody @Valid AiQueryRequest queryRequest) {
|
||||
Long userId = (Long) request.getAttribute("userId");
|
||||
return Result.success(aiQueryService.parse(userId, queryRequest.getQuery()));
|
||||
}
|
||||
|
||||
/**
|
||||
* 解析自然语言记账内容为交易记录
|
||||
*
|
||||
* @param request HTTP请求，从中获取用户ID
|
||||
* @param recordRequest 记账请求，携带自然语言语句
|
||||
* @return 结构化交易记录
|
||||
*/
|
||||
@PostMapping("/record")
|
||||
@Operation(summary = "自然语言记账解析", description = "将自然语言记账内容解析为交易记录，供用户确认后保存")
|
||||
public Result<AiRecordResultVO> record(HttpServletRequest request,
|
||||
@RequestBody @Valid AiRecordRequest recordRequest) {
|
||||
Long userId = (Long) request.getAttribute("userId");
|
||||
return Result.success(aiRecordService.parse(userId, recordRequest.getText()));
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,39 @@
|
||||
package cn.rosercode.finance.dto;
|
||||
|
||||
import jakarta.validation.Valid;
|
||||
import jakarta.validation.constraints.NotEmpty;
|
||||
import jakarta.validation.constraints.NotNull;
|
||||
import lombok.Data;
|
||||
|
||||
import java.util.List;
|
||||
|
||||
/**
|
||||
* AI 建议应用请求DTO，携带逐条记录的建议分类与标签，用于一键写入。
|
||||
*
|
||||
* @date 2026-09-17 17:52:00
|
||||
*/
|
||||
@Data
|
||||
public class AiApplyRequest {
|
||||
|
||||
/** 待应用的建议列表 */
|
||||
@NotEmpty(message = "请至少提供一条建议")
|
||||
@Valid
|
||||
private List<Item> suggestions;
|
||||
|
||||
/**
|
||||
* 单条建议项，描述某条交易记录的目标分类与待追加标签。
|
||||
*/
|
||||
@Data
|
||||
public static class Item {
|
||||
|
||||
/** 交易记录ID */
|
||||
@NotNull(message = "记录ID不能为空")
|
||||
private Long transactionId;
|
||||
|
||||
/** 建议分类ID，为 null 表示不修改分类 */
|
||||
private Long categoryId;
|
||||
|
||||
/** 建议追加的标签ID列表，可为空表示不追加标签 */
|
||||
private List<Long> tagIds;
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,24 @@
|
||||
package cn.rosercode.finance.dto;
|
||||
|
||||
import lombok.Data;
|
||||
|
||||
/**
|
||||
* AI 配置保存请求DTO，用于更新用户的 AI 智能建议配置。
|
||||
*
|
||||
* @date 2026-09-17 17:51:58
|
||||
*/
|
||||
@Data
|
||||
public class AiConfigRequest {
|
||||
|
||||
/** 是否启用 AI 智能建议 */
|
||||
private Boolean enabled;
|
||||
|
||||
/** DeepSeek API Key，为空或 null 时保留原值（不修改） */
|
||||
private String apiKey;
|
||||
|
||||
/** API 基础地址，为空时使用默认值 */
|
||||
private String baseUrl;
|
||||
|
||||
/** 模型名称，为空时使用默认值 */
|
||||
private String model;
|
||||
}
|
||||
@@ -0,0 +1,17 @@
|
||||
package cn.rosercode.finance.dto;
|
||||
|
||||
import jakarta.validation.constraints.NotBlank;
|
||||
import lombok.Data;
|
||||
|
||||
/**
|
||||
* AI 自然语言查询请求DTO，携带用户输入的查询语句。
|
||||
*
|
||||
* @date 2026-09-17 18:13:06
|
||||
*/
|
||||
@Data
|
||||
public class AiQueryRequest {
|
||||
|
||||
/** 用户输入的自然语言查询语句 */
|
||||
@NotBlank(message = "请输入查询内容")
|
||||
private String query;
|
||||
}
|
||||
@@ -0,0 +1,17 @@
|
||||
package cn.rosercode.finance.dto;
|
||||
|
||||
import jakarta.validation.constraints.NotBlank;
|
||||
import lombok.Data;
|
||||
|
||||
/**
|
||||
* AI 自然语言记账请求DTO，携带用户输入的记账语句。
|
||||
*
|
||||
* @date 2026-09-17 18:23:10
|
||||
*/
|
||||
@Data
|
||||
public class AiRecordRequest {
|
||||
|
||||
/** 用户输入的自然语言记账内容 */
|
||||
@NotBlank(message = "请输入记账内容")
|
||||
private String text;
|
||||
}
|
||||
@@ -0,0 +1,19 @@
|
||||
package cn.rosercode.finance.dto;
|
||||
|
||||
import jakarta.validation.constraints.NotEmpty;
|
||||
import lombok.Data;
|
||||
|
||||
import java.util.List;
|
||||
|
||||
/**
|
||||
* AI 建议请求DTO，携带需要分析分类与标签的交易记录ID列表。
|
||||
*
|
||||
* @date 2026-09-17 17:51:59
|
||||
*/
|
||||
@Data
|
||||
public class AiSuggestRequest {
|
||||
|
||||
/** 待分析建议的交易记录ID列表 */
|
||||
@NotEmpty(message = "请至少选择一条记录")
|
||||
private List<Long> ids;
|
||||
}
|
||||
@@ -0,0 +1,43 @@
|
||||
package cn.rosercode.finance.entity;
|
||||
|
||||
import com.baomidou.mybatisplus.annotation.IdType;
|
||||
import com.baomidou.mybatisplus.annotation.TableId;
|
||||
import com.baomidou.mybatisplus.annotation.TableName;
|
||||
import lombok.Data;
|
||||
|
||||
import java.time.LocalDateTime;
|
||||
|
||||
/**
|
||||
* AI 配置实体类，对应数据库表 t_ai_config，用于记录用户的 AI 智能建议配置。
|
||||
*
|
||||
* @date 2026-09-17 17:51:57
|
||||
*/
|
||||
@Data
|
||||
@TableName("t_ai_config")
|
||||
public class AiConfig {
|
||||
|
||||
/** 配置ID，主键自增 */
|
||||
@TableId(type = IdType.AUTO)
|
||||
private Long id;
|
||||
|
||||
/** 所属用户ID */
|
||||
private Long userId;
|
||||
|
||||
/** 是否启用 AI 智能建议 */
|
||||
private Boolean enabled;
|
||||
|
||||
/** DeepSeek API Key（明文存储） */
|
||||
private String apiKey;
|
||||
|
||||
/** API 基础地址 */
|
||||
private String baseUrl;
|
||||
|
||||
/** 使用的模型名称 */
|
||||
private String model;
|
||||
|
||||
/** 创建时间 */
|
||||
private LocalDateTime createdAt;
|
||||
|
||||
/** 更新时间 */
|
||||
private LocalDateTime updatedAt;
|
||||
}
|
||||
@@ -0,0 +1,58 @@
|
||||
package cn.rosercode.finance.entity;
|
||||
|
||||
import com.baomidou.mybatisplus.annotation.IdType;
|
||||
import com.baomidou.mybatisplus.annotation.TableId;
|
||||
import com.baomidou.mybatisplus.annotation.TableName;
|
||||
import lombok.Data;
|
||||
|
||||
import java.time.LocalDateTime;
|
||||
|
||||
/**
|
||||
* AI 使用记录实体类，对应数据库表 t_ai_usage，记录每次调用 DeepSeek 的用量与结果。
|
||||
*
|
||||
* @date 2026-09-17 18:02:05
|
||||
*/
|
||||
@Data
|
||||
@TableName("t_ai_usage")
|
||||
public class AiUsage {
|
||||
|
||||
/** 记录ID，主键自增 */
|
||||
@TableId(type = IdType.AUTO)
|
||||
private Long id;
|
||||
|
||||
/** 所属用户ID */
|
||||
private Long userId;
|
||||
|
||||
/** 调用类型：SUGGEST-生成建议，TEST-测试连接 */
|
||||
private String type;
|
||||
|
||||
/** 模型名称 */
|
||||
private String model;
|
||||
|
||||
/** 本次分析的记录数 */
|
||||
private Integer transactionCount;
|
||||
|
||||
/** 输入 token 数 */
|
||||
private Integer promptTokens;
|
||||
|
||||
/** 输出 token 数 */
|
||||
private Integer completionTokens;
|
||||
|
||||
/** 总 token 数 */
|
||||
private Integer totalTokens;
|
||||
|
||||
/** 缓存命中的输入 token 数 */
|
||||
private Integer cachedTokens;
|
||||
|
||||
/** 是否成功 */
|
||||
private Boolean success;
|
||||
|
||||
/** 失败原因 */
|
||||
private String errorMessage;
|
||||
|
||||
/** 调用耗时（毫秒） */
|
||||
private Long durationMs;
|
||||
|
||||
/** 创建时间 */
|
||||
private LocalDateTime createdAt;
|
||||
}
|
||||
@@ -0,0 +1,21 @@
|
||||
package cn.rosercode.finance.enums;
|
||||
|
||||
/**
|
||||
* AI 调用类型枚举，标识使用记录的来源。
|
||||
*
|
||||
* @date 2026-09-17 18:02:04
|
||||
*/
|
||||
public enum AiCallType {
|
||||
|
||||
/** 生成分类与标签建议 */
|
||||
SUGGEST,
|
||||
|
||||
/** 测试连接 */
|
||||
TEST,
|
||||
|
||||
/** 自然语言查询解析 */
|
||||
QUERY,
|
||||
|
||||
/** 自然语言记账解析 */
|
||||
RECORD
|
||||
}
|
||||
@@ -0,0 +1,14 @@
|
||||
package cn.rosercode.finance.mapper;
|
||||
|
||||
import com.baomidou.mybatisplus.core.mapper.BaseMapper;
|
||||
import cn.rosercode.finance.entity.AiConfig;
|
||||
import org.apache.ibatis.annotations.Mapper;
|
||||
|
||||
/**
|
||||
* AI 配置Mapper接口，继承MyBatis-Plus的BaseMapper，提供AI配置表的基础CRUD操作。
|
||||
*
|
||||
* @date 2026-09-17 17:51:58
|
||||
*/
|
||||
@Mapper
|
||||
public interface AiConfigMapper extends BaseMapper<AiConfig> {
|
||||
}
|
||||
@@ -0,0 +1,33 @@
|
||||
package cn.rosercode.finance.mapper;
|
||||
|
||||
import com.baomidou.mybatisplus.core.mapper.BaseMapper;
|
||||
import cn.rosercode.finance.entity.AiUsage;
|
||||
import cn.rosercode.finance.vo.AiUsageSummaryVO;
|
||||
import org.apache.ibatis.annotations.Mapper;
|
||||
import org.apache.ibatis.annotations.Param;
|
||||
import org.apache.ibatis.annotations.Select;
|
||||
|
||||
/**
|
||||
* AI 使用记录Mapper接口，继承MyBatis-Plus的BaseMapper，并提供汇总统计查询。
|
||||
*
|
||||
* @date 2026-09-17 18:02:08
|
||||
*/
|
||||
@Mapper
|
||||
public interface AiUsageMapper extends BaseMapper<AiUsage> {
|
||||
|
||||
/**
|
||||
* 统计指定用户的累计使用情况（调用次数与 token 用量）。
|
||||
*
|
||||
* @param userId 用户ID
|
||||
* @return 汇总结果，无记录时各项为 0
|
||||
*/
|
||||
@Select("SELECT COUNT(*) AS totalCalls, "
|
||||
+ "COALESCE(SUM(CASE WHEN success = 1 THEN 1 ELSE 0 END), 0) AS successCalls, "
|
||||
+ "COALESCE(SUM(CASE WHEN success = 0 THEN 1 ELSE 0 END), 0) AS failedCalls, "
|
||||
+ "COALESCE(SUM(prompt_tokens), 0) AS promptTokens, "
|
||||
+ "COALESCE(SUM(completion_tokens), 0) AS completionTokens, "
|
||||
+ "COALESCE(SUM(total_tokens), 0) AS totalTokens, "
|
||||
+ "COALESCE(SUM(cached_tokens), 0) AS cachedTokens "
|
||||
+ "FROM t_ai_usage WHERE user_id = #{userId}")
|
||||
AiUsageSummaryVO selectSummary(@Param("userId") Long userId);
|
||||
}
|
||||
@@ -0,0 +1,44 @@
|
||||
package cn.rosercode.finance.service;
|
||||
|
||||
import cn.rosercode.finance.dto.AiConfigRequest;
|
||||
import cn.rosercode.finance.entity.AiConfig;
|
||||
import cn.rosercode.finance.vo.AiConfigVO;
|
||||
|
||||
/**
|
||||
* AI 配置服务接口，提供用户级 AI 配置的读取、保存与连通性测试。
|
||||
*
|
||||
* @date 2026-09-17 17:52:14
|
||||
*/
|
||||
public interface AiConfigService {
|
||||
|
||||
/**
|
||||
* 获取当前用户的 AI 配置（API Key 脱敏），未配置时返回默认值。
|
||||
*
|
||||
* @param userId 用户ID
|
||||
* @return 脱敏后的配置视图对象
|
||||
*/
|
||||
AiConfigVO getConfig(Long userId);
|
||||
|
||||
/**
|
||||
* 保存当前用户的 AI 配置。
|
||||
*
|
||||
* @param userId 用户ID
|
||||
* @param request 配置保存请求
|
||||
*/
|
||||
void saveConfig(Long userId, AiConfigRequest request);
|
||||
|
||||
/**
|
||||
* 获取原始配置实体，供内部调用（含明文 API Key）。
|
||||
*
|
||||
* @param userId 用户ID
|
||||
* @return 配置实体，不存在时返回 null
|
||||
*/
|
||||
AiConfig getRawConfig(Long userId);
|
||||
|
||||
/**
|
||||
* 测试当前用户配置的 DeepSeek 连通性。
|
||||
*
|
||||
* @param userId 用户ID
|
||||
*/
|
||||
void testConnection(Long userId);
|
||||
}
|
||||
@@ -0,0 +1,20 @@
|
||||
package cn.rosercode.finance.service;
|
||||
|
||||
import cn.rosercode.finance.vo.AiQueryResultVO;
|
||||
|
||||
/**
|
||||
* AI 自然语言查询服务接口，将自然语言解析为交易筛选条件。
|
||||
*
|
||||
* @date 2026-09-17 18:13:18
|
||||
*/
|
||||
public interface AiQueryService {
|
||||
|
||||
/**
|
||||
* 解析自然语言查询。
|
||||
*
|
||||
* @param userId 用户ID
|
||||
* @param query 查询语句
|
||||
* @return 结构化筛选条件
|
||||
*/
|
||||
AiQueryResultVO parse(Long userId, String query);
|
||||
}
|
||||
@@ -0,0 +1,20 @@
|
||||
package cn.rosercode.finance.service;
|
||||
|
||||
import cn.rosercode.finance.vo.AiRecordResultVO;
|
||||
|
||||
/**
|
||||
* AI 自然语言记账服务接口，将自然语言解析为一条交易记录。
|
||||
*
|
||||
* @date 2026-09-17 18:23:17
|
||||
*/
|
||||
public interface AiRecordService {
|
||||
|
||||
/**
|
||||
* 解析自然语言记账内容。
|
||||
*
|
||||
* @param userId 用户ID
|
||||
* @param text 记账语句
|
||||
* @return 结构化交易记录
|
||||
*/
|
||||
AiRecordResultVO parse(Long userId, String text);
|
||||
}
|
||||
@@ -0,0 +1,32 @@
|
||||
package cn.rosercode.finance.service;
|
||||
|
||||
import cn.rosercode.finance.dto.AiApplyRequest;
|
||||
import cn.rosercode.finance.vo.AiSuggestionVO;
|
||||
|
||||
import java.util.List;
|
||||
|
||||
/**
|
||||
* AI 建议服务接口，提供交易记录的分类与标签智能建议及应用能力。
|
||||
*
|
||||
* @date 2026-09-17 17:52:23
|
||||
*/
|
||||
public interface AiSuggestionService {
|
||||
|
||||
/**
|
||||
* 根据交易记录内容调用 DeepSeek 生成分类与标签建议。
|
||||
*
|
||||
* @param userId 用户ID
|
||||
* @param ids 待分析的建议记录ID列表
|
||||
* @return 逐条建议列表
|
||||
*/
|
||||
List<AiSuggestionVO> suggest(Long userId, List<Long> ids);
|
||||
|
||||
/**
|
||||
* 一键应用建议：覆盖分类、追加标签。
|
||||
*
|
||||
* @param userId 用户ID
|
||||
* @param request 建议应用请求
|
||||
* @return 实际发生变更的记录数
|
||||
*/
|
||||
int apply(Long userId, AiApplyRequest request);
|
||||
}
|
||||
@@ -0,0 +1,48 @@
|
||||
package cn.rosercode.finance.service;
|
||||
|
||||
import cn.rosercode.finance.client.DeepSeekResponse;
|
||||
import cn.rosercode.finance.enums.AiCallType;
|
||||
import cn.rosercode.finance.vo.AiUsageResultVO;
|
||||
|
||||
/**
|
||||
* AI 使用记录服务接口，提供调用用量的记录与查询能力。
|
||||
*
|
||||
* @date 2026-09-17 18:02:15
|
||||
*/
|
||||
public interface AiUsageService {
|
||||
|
||||
/**
|
||||
* 记录一次成功的 AI 调用。
|
||||
*
|
||||
* @param userId 用户ID
|
||||
* @param type 调用类型
|
||||
* @param model 模型名称
|
||||
* @param transactionCount 本次分析的记录数
|
||||
* @param response 模型返回结果（含 token 用量）
|
||||
* @param durationMs 调用耗时（毫秒）
|
||||
*/
|
||||
void recordSuccess(Long userId, AiCallType type, String model, int transactionCount,
|
||||
DeepSeekResponse response, long durationMs);
|
||||
|
||||
/**
|
||||
* 记录一次失败的 AI 调用。
|
||||
*
|
||||
* @param userId 用户ID
|
||||
* @param type 调用类型
|
||||
* @param model 模型名称
|
||||
* @param transactionCount 本次分析的记录数
|
||||
* @param durationMs 调用耗时（毫秒）
|
||||
* @param errorMessage 失败原因
|
||||
*/
|
||||
void recordFailure(Long userId, AiCallType type, String model, int transactionCount,
|
||||
long durationMs, String errorMessage);
|
||||
|
||||
/**
|
||||
* 查询指定用户的使用情况（累计汇总 + 最近明细）。
|
||||
*
|
||||
* @param userId 用户ID
|
||||
* @param recentLimit 最近明细条数
|
||||
* @return 使用情况视图对象
|
||||
*/
|
||||
AiUsageResultVO getUsage(Long userId, int recentLimit);
|
||||
}
|
||||
@@ -0,0 +1,158 @@
|
||||
package cn.rosercode.finance.service.impl;
|
||||
|
||||
import com.baomidou.mybatisplus.core.conditions.query.LambdaQueryWrapper;
|
||||
import cn.rosercode.finance.client.DeepSeekClient;
|
||||
import cn.rosercode.finance.client.DeepSeekResponse;
|
||||
import cn.rosercode.finance.common.BusinessException;
|
||||
import cn.rosercode.finance.dto.AiConfigRequest;
|
||||
import cn.rosercode.finance.entity.AiConfig;
|
||||
import cn.rosercode.finance.enums.AiCallType;
|
||||
import cn.rosercode.finance.mapper.AiConfigMapper;
|
||||
import cn.rosercode.finance.service.AiConfigService;
|
||||
import cn.rosercode.finance.service.AiUsageService;
|
||||
import cn.rosercode.finance.vo.AiConfigVO;
|
||||
import org.springframework.stereotype.Service;
|
||||
import org.springframework.transaction.annotation.Transactional;
|
||||
|
||||
/**
|
||||
* AI 配置服务实现类，处理用户级 AI 配置的读写与连通性测试。
|
||||
*
|
||||
* @date 2026-09-17 17:52:22
|
||||
*/
|
||||
@Service
|
||||
public class AiConfigServiceImpl implements AiConfigService {
|
||||
|
||||
/** 默认 API 基础地址 */
|
||||
private static final String DEFAULT_BASE_URL = "https://api.deepseek.com";
|
||||
|
||||
/** 默认模型名称 */
|
||||
private static final String DEFAULT_MODEL = "deepseek-chat";
|
||||
|
||||
private final AiConfigMapper aiConfigMapper;
|
||||
private final DeepSeekClient deepSeekClient;
|
||||
private final AiUsageService aiUsageService;
|
||||
|
||||
public AiConfigServiceImpl(AiConfigMapper aiConfigMapper, DeepSeekClient deepSeekClient,
|
||||
AiUsageService aiUsageService) {
|
||||
this.aiConfigMapper = aiConfigMapper;
|
||||
this.deepSeekClient = deepSeekClient;
|
||||
this.aiUsageService = aiUsageService;
|
||||
}
|
||||
|
||||
/**
|
||||
* {@inheritDoc}
|
||||
*
|
||||
* 未配置时返回关闭状态与默认地址、模型。
|
||||
*/
|
||||
@Override
|
||||
public AiConfigVO getConfig(Long userId) {
|
||||
AiConfig config = getRawConfig(userId);
|
||||
AiConfigVO vo = new AiConfigVO();
|
||||
if (config == null) {
|
||||
vo.setEnabled(false);
|
||||
vo.setHasApiKey(false);
|
||||
vo.setBaseUrl(DEFAULT_BASE_URL);
|
||||
vo.setModel(DEFAULT_MODEL);
|
||||
return vo;
|
||||
}
|
||||
vo.setEnabled(Boolean.TRUE.equals(config.getEnabled()));
|
||||
vo.setHasApiKey(config.getApiKey() != null && !config.getApiKey().isBlank());
|
||||
vo.setApiKey(maskApiKey(config.getApiKey()));
|
||||
vo.setBaseUrl(config.getBaseUrl());
|
||||
vo.setModel(config.getModel());
|
||||
return vo;
|
||||
}
|
||||
|
||||
/**
|
||||
* {@inheritDoc}
|
||||
*
|
||||
* API Key 为空时保留原值，地址与模型为空时回退默认值。
|
||||
*/
|
||||
@Override
|
||||
@Transactional
|
||||
public void saveConfig(Long userId, AiConfigRequest request) {
|
||||
AiConfig config = getRawConfig(userId);
|
||||
if (config == null) {
|
||||
config = new AiConfig();
|
||||
config.setUserId(userId);
|
||||
}
|
||||
config.setEnabled(request.getEnabled() != null ? request.getEnabled() : Boolean.FALSE);
|
||||
// API Key 为空表示不修改，保留原值
|
||||
if (request.getApiKey() != null && !request.getApiKey().isBlank()) {
|
||||
config.setApiKey(request.getApiKey().trim());
|
||||
}
|
||||
config.setBaseUrl(isBlank(request.getBaseUrl()) ? DEFAULT_BASE_URL : request.getBaseUrl().trim());
|
||||
config.setModel(isBlank(request.getModel()) ? DEFAULT_MODEL : request.getModel().trim());
|
||||
|
||||
if (config.getId() == null) {
|
||||
aiConfigMapper.insert(config);
|
||||
} else {
|
||||
aiConfigMapper.updateById(config);
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* {@inheritDoc}
|
||||
*/
|
||||
@Override
|
||||
public AiConfig getRawConfig(Long userId) {
|
||||
return aiConfigMapper.selectOne(new LambdaQueryWrapper<AiConfig>()
|
||||
.eq(AiConfig::getUserId, userId));
|
||||
}
|
||||
|
||||
/**
|
||||
* {@inheritDoc}
|
||||
*
|
||||
* 发送一条极简请求验证 Key 与地址是否可用。
|
||||
*/
|
||||
@Override
|
||||
public void testConnection(Long userId) {
|
||||
AiConfig config = getRawConfig(userId);
|
||||
if (config == null || isBlank(config.getApiKey())) {
|
||||
throw new BusinessException("请先配置 DeepSeek API Key");
|
||||
}
|
||||
long start = System.currentTimeMillis();
|
||||
DeepSeekResponse response;
|
||||
try {
|
||||
response = deepSeekClient.chat(config.getBaseUrl(), config.getApiKey(), config.getModel(),
|
||||
"你是一个测试助手，请只输出 JSON。",
|
||||
"请返回 {\"ok\":true}");
|
||||
} catch (RuntimeException e) {
|
||||
// 失败也记录一次使用，便于排查问题
|
||||
aiUsageService.recordFailure(userId, AiCallType.TEST, config.getModel(), 0,
|
||||
System.currentTimeMillis() - start, e.getMessage());
|
||||
throw e;
|
||||
}
|
||||
aiUsageService.recordSuccess(userId, AiCallType.TEST, config.getModel(), 0,
|
||||
response, System.currentTimeMillis() - start);
|
||||
if (isBlank(response.getContent())) {
|
||||
throw new BusinessException("连接成功但未返回内容");
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 对 API Key 进行脱敏，仅保留前缀与后缀。
|
||||
*
|
||||
* @param apiKey 原始 API Key
|
||||
* @return 脱敏后的字符串，未配置时返回 null
|
||||
*/
|
||||
private String maskApiKey(String apiKey) {
|
||||
if (isBlank(apiKey)) {
|
||||
return null;
|
||||
}
|
||||
if (apiKey.length() <= 7) {
|
||||
return "****";
|
||||
}
|
||||
return apiKey.substring(0, 3) + "****" + apiKey.substring(apiKey.length() - 4);
|
||||
}
|
||||
|
||||
/**
|
||||
* 判断字符串是否为空白。
|
||||
*
|
||||
* @param value 待判断字符串
|
||||
* @return 为 null 或全空白时返回 true
|
||||
*/
|
||||
private boolean isBlank(String value) {
|
||||
return value == null || value.isBlank();
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,436 @@
|
||||
package cn.rosercode.finance.service.impl;
|
||||
|
||||
import com.baomidou.mybatisplus.core.conditions.query.LambdaQueryWrapper;
|
||||
import cn.rosercode.finance.client.DeepSeekClient;
|
||||
import cn.rosercode.finance.client.DeepSeekResponse;
|
||||
import cn.rosercode.finance.common.BusinessException;
|
||||
import cn.rosercode.finance.entity.AiConfig;
|
||||
import cn.rosercode.finance.entity.Category;
|
||||
import cn.rosercode.finance.entity.Tag;
|
||||
import cn.rosercode.finance.enums.AiCallType;
|
||||
import cn.rosercode.finance.mapper.CategoryMapper;
|
||||
import cn.rosercode.finance.mapper.TagMapper;
|
||||
import cn.rosercode.finance.service.AiConfigService;
|
||||
import cn.rosercode.finance.service.AiQueryService;
|
||||
import cn.rosercode.finance.service.AiUsageService;
|
||||
import cn.rosercode.finance.vo.AiQueryResultVO;
|
||||
import com.fasterxml.jackson.core.JsonProcessingException;
|
||||
import com.fasterxml.jackson.databind.JsonNode;
|
||||
import com.fasterxml.jackson.databind.ObjectMapper;
|
||||
import org.springframework.stereotype.Service;
|
||||
|
||||
import java.math.BigDecimal;
|
||||
import java.time.DayOfWeek;
|
||||
import java.time.LocalDate;
|
||||
import java.time.format.DateTimeFormatter;
|
||||
import java.util.ArrayList;
|
||||
import java.util.List;
|
||||
import java.util.stream.Collectors;
|
||||
|
||||
/**
|
||||
* AI 自然语言查询服务实现类，负责提示词组装、日期推算与名称映射。
|
||||
*
|
||||
* @date 2026-09-17 18:13:28
|
||||
*/
|
||||
@Service
|
||||
public class AiQueryServiceImpl implements AiQueryService {
|
||||
|
||||
/** 日期格式 */
|
||||
private static final DateTimeFormatter DATE_FMT = DateTimeFormatter.ofPattern("yyyy-MM-dd");
|
||||
|
||||
/** 日期时间格式，与前端筛选参数保持一致 */
|
||||
private static final DateTimeFormatter DATETIME_FMT = DateTimeFormatter.ofPattern("yyyy-MM-dd'T'HH:mm:ss");
|
||||
|
||||
/** 合法的收支类型 */
|
||||
private static final List<String> VALID_TYPES = List.of("INCOME", "EXPENSE");
|
||||
|
||||
/** 合法的支付方式 */
|
||||
private static final List<String> VALID_PAYMENTS = List.of("WECHAT", "ALIPAY", "CASH", "BANK");
|
||||
|
||||
private final AiConfigService aiConfigService;
|
||||
private final CategoryMapper categoryMapper;
|
||||
private final TagMapper tagMapper;
|
||||
private final DeepSeekClient deepSeekClient;
|
||||
private final AiUsageService aiUsageService;
|
||||
private final ObjectMapper objectMapper;
|
||||
|
||||
public AiQueryServiceImpl(AiConfigService aiConfigService,
|
||||
CategoryMapper categoryMapper,
|
||||
TagMapper tagMapper,
|
||||
DeepSeekClient deepSeekClient,
|
||||
AiUsageService aiUsageService,
|
||||
ObjectMapper objectMapper) {
|
||||
this.aiConfigService = aiConfigService;
|
||||
this.categoryMapper = categoryMapper;
|
||||
this.tagMapper = tagMapper;
|
||||
this.deepSeekClient = deepSeekClient;
|
||||
this.aiUsageService = aiUsageService;
|
||||
this.objectMapper = objectMapper;
|
||||
}
|
||||
|
||||
/**
|
||||
* {@inheritDoc}
|
||||
*
|
||||
* 将用户分类、标签与当前日期注入提示词，解析模型输出并映射为筛选条件。
|
||||
*/
|
||||
@Override
|
||||
public AiQueryResultVO parse(Long userId, String query) {
|
||||
if (query == null || query.isBlank()) {
|
||||
throw new BusinessException("请输入查询内容");
|
||||
}
|
||||
AiConfig config = aiConfigService.getRawConfig(userId);
|
||||
if (config == null || !Boolean.TRUE.equals(config.getEnabled())) {
|
||||
throw new BusinessException("AI 功能未启用，请先在 AI 设置中开启");
|
||||
}
|
||||
if (isBlank(config.getApiKey())) {
|
||||
throw new BusinessException("请先配置 DeepSeek API Key");
|
||||
}
|
||||
|
||||
List<Category> categories = categoryMapper.selectList(new LambdaQueryWrapper<Category>()
|
||||
.eq(Category::getUserId, userId));
|
||||
List<Tag> tags = tagMapper.selectList(new LambdaQueryWrapper<Tag>()
|
||||
.eq(Tag::getUserId, userId));
|
||||
|
||||
String systemPrompt = buildSystemPrompt(categories, tags);
|
||||
long start = System.currentTimeMillis();
|
||||
DeepSeekResponse aiResponse;
|
||||
try {
|
||||
aiResponse = deepSeekClient.chat(config.getBaseUrl(), config.getApiKey(), config.getModel(),
|
||||
systemPrompt, query);
|
||||
} catch (RuntimeException e) {
|
||||
// 失败也记录一次使用，便于排查问题
|
||||
aiUsageService.recordFailure(userId, AiCallType.QUERY, config.getModel(), 0,
|
||||
System.currentTimeMillis() - start, e.getMessage());
|
||||
throw e;
|
||||
}
|
||||
aiUsageService.recordSuccess(userId, AiCallType.QUERY, config.getModel(), 0,
|
||||
aiResponse, System.currentTimeMillis() - start);
|
||||
|
||||
return parseResult(aiResponse.getContent(), categories, tags);
|
||||
}
|
||||
|
||||
/**
|
||||
* 构建系统提示词，注入可用分类、标签与当前日期。
|
||||
*
|
||||
* @param categories 用户分类列表
|
||||
* @param tags 用户标签列表
|
||||
* @return 系统提示词
|
||||
*/
|
||||
private String buildSystemPrompt(List<Category> categories, List<Tag> tags) {
|
||||
String categoryNames = categories.stream().map(Category::getName).collect(Collectors.joining("、"));
|
||||
String tagNames = tags.stream().map(Tag::getName).collect(Collectors.joining("、"));
|
||||
String today = LocalDate.now().format(DATE_FMT);
|
||||
return "你是一个记账系统的查询解析助手。请把用户的自然语言查询解析为结构化筛选条件，只输出 JSON，不要输出任何多余内容。\n"
|
||||
+ "可用的收支类型：INCOME（收入）、EXPENSE（支出）。\n"
|
||||
+ "可用的支付方式：WECHAT（微信）、ALIPAY（支付宝）、CASH（现金）、BANK（银行卡）。\n"
|
||||
+ "可用分类：" + (categoryNames.isEmpty() ? "无" : categoryNames) + "。\n"
|
||||
+ "可用标签：" + (tagNames.isEmpty() ? "无" : tagNames) + "。\n"
|
||||
+ "时间范围请使用 preset 表示，取值：TODAY、YESTERDAY、THIS_WEEK、LAST_WEEK、THIS_MONTH、LAST_MONTH、"
|
||||
+ "LAST_7_DAYS、LAST_30_DAYS、THIS_YEAR、LAST_YEAR、CUSTOM、NONE；若为 CUSTOM，请同时给出 start 与 end（格式 yyyy-MM-dd）。\n"
|
||||
+ "未提及的字段请留空（数组用 []，其余用 null）。分类和标签只能使用上面给出的名称。\n"
|
||||
+ "今天是 " + today + "。\n"
|
||||
+ "输出格式：{\"summary\":\"简要说明识别到的查询意图\",\"types\":[],\"categoryNames\":[],\"tagNames\":[],"
|
||||
+ "\"paymentMethods\":[],\"dateRange\":{\"preset\":\"NONE\",\"start\":null,\"end\":null},\"minAmount\":null,"
|
||||
+ "\"maxAmount\":null,\"remark\":null,\"includeInStats\":null,\"uncategorized\":false,\"unmatched\":[]}";
|
||||
}
|
||||
|
||||
/**
|
||||
* 解析模型返回的 JSON 并映射为筛选条件。
|
||||
*
|
||||
* @param content 模型输出内容
|
||||
* @param categories 用户分类列表
|
||||
* @param tags 用户标签列表
|
||||
* @return 结构化筛选条件
|
||||
*/
|
||||
private AiQueryResultVO parseResult(String content, List<Category> categories, List<Tag> tags) {
|
||||
JsonNode root;
|
||||
try {
|
||||
root = objectMapper.readTree(stripCodeFence(content));
|
||||
} catch (JsonProcessingException e) {
|
||||
throw new BusinessException("AI 返回内容不是有效 JSON，请重试");
|
||||
}
|
||||
|
||||
AiQueryResultVO vo = new AiQueryResultVO();
|
||||
vo.setSummary(root.path("summary").asText(""));
|
||||
vo.setTypes(filterValid(root.path("types"), VALID_TYPES));
|
||||
vo.setPaymentMethods(filterValid(root.path("paymentMethods"), VALID_PAYMENTS));
|
||||
vo.setUncategorized(root.path("uncategorized").asBoolean(false));
|
||||
|
||||
JsonNode statsNode = root.path("includeInStats");
|
||||
if (!statsNode.isMissingNode() && !statsNode.isNull()) {
|
||||
vo.setIncludeInStats(statsNode.asBoolean());
|
||||
}
|
||||
vo.setMinAmount(decimalOrNull(root.path("minAmount")));
|
||||
vo.setMaxAmount(decimalOrNull(root.path("maxAmount")));
|
||||
|
||||
String remark = root.path("remark").asText(null);
|
||||
vo.setRemark(isBlank(remark) ? null : remark.trim());
|
||||
|
||||
// 分类名称映射为ID
|
||||
List<Long> categoryIds = new ArrayList<>();
|
||||
List<String> categoryNames = new ArrayList<>();
|
||||
JsonNode categoryNode = root.path("categoryNames");
|
||||
if (categoryNode.isArray()) {
|
||||
for (JsonNode node : categoryNode) {
|
||||
Category category = matchCategory(categories, node.asText(""));
|
||||
if (category != null && !categoryIds.contains(category.getId())) {
|
||||
categoryIds.add(category.getId());
|
||||
categoryNames.add(category.getName());
|
||||
}
|
||||
}
|
||||
}
|
||||
vo.setCategoryIds(categoryIds);
|
||||
vo.setCategoryNames(categoryNames);
|
||||
|
||||
// 标签名称映射为ID
|
||||
List<Long> tagIds = new ArrayList<>();
|
||||
List<String> tagNames = new ArrayList<>();
|
||||
JsonNode tagNode = root.path("tagNames");
|
||||
if (tagNode.isArray()) {
|
||||
for (JsonNode node : tagNode) {
|
||||
Tag tag = matchTag(tags, node.asText(""));
|
||||
if (tag != null && !tagIds.contains(tag.getId())) {
|
||||
tagIds.add(tag.getId());
|
||||
tagNames.add(tag.getName());
|
||||
}
|
||||
}
|
||||
}
|
||||
vo.setTagIds(tagIds);
|
||||
vo.setTagNames(tagNames);
|
||||
|
||||
// 时间范围推算
|
||||
String[] range = resolveDateRange(root.path("dateRange"));
|
||||
if (range != null) {
|
||||
vo.setStartTime(range[0]);
|
||||
vo.setEndTime(range[1]);
|
||||
}
|
||||
|
||||
// 未识别关键词
|
||||
List<String> unmatched = new ArrayList<>();
|
||||
JsonNode unmatchedNode = root.path("unmatched");
|
||||
if (unmatchedNode.isArray()) {
|
||||
for (JsonNode node : unmatchedNode) {
|
||||
String text = node.asText("");
|
||||
if (!isBlank(text)) {
|
||||
unmatched.add(text);
|
||||
}
|
||||
}
|
||||
}
|
||||
vo.setUnmatched(unmatched);
|
||||
return vo;
|
||||
}
|
||||
|
||||
/**
|
||||
* 将模型给出的日期 preset 推算为绝对时间范围。
|
||||
*
|
||||
* @param dateRange 模型输出的 dateRange 节点
|
||||
* @return 起止时间数组，无法识别时返回 null
|
||||
*/
|
||||
private String[] resolveDateRange(JsonNode dateRange) {
|
||||
if (dateRange == null || dateRange.isMissingNode() || dateRange.isNull()) {
|
||||
return null;
|
||||
}
|
||||
String preset = dateRange.path("preset").asText("NONE");
|
||||
if (isBlank(preset) || "NONE".equalsIgnoreCase(preset)) {
|
||||
return null;
|
||||
}
|
||||
LocalDate today = LocalDate.now();
|
||||
LocalDate start;
|
||||
LocalDate end;
|
||||
switch (preset.toUpperCase()) {
|
||||
case "TODAY":
|
||||
start = today;
|
||||
end = today;
|
||||
break;
|
||||
case "YESTERDAY":
|
||||
start = today.minusDays(1);
|
||||
end = start;
|
||||
break;
|
||||
case "THIS_WEEK":
|
||||
start = today.with(DayOfWeek.MONDAY);
|
||||
end = today;
|
||||
break;
|
||||
case "LAST_WEEK":
|
||||
start = today.with(DayOfWeek.MONDAY).minusWeeks(1);
|
||||
end = start.plusDays(6);
|
||||
break;
|
||||
case "THIS_MONTH":
|
||||
start = today.withDayOfMonth(1);
|
||||
end = today;
|
||||
break;
|
||||
case "LAST_MONTH":
|
||||
start = today.withDayOfMonth(1).minusMonths(1);
|
||||
end = start.withDayOfMonth(start.lengthOfMonth());
|
||||
break;
|
||||
case "LAST_7_DAYS":
|
||||
start = today.minusDays(6);
|
||||
end = today;
|
||||
break;
|
||||
case "LAST_30_DAYS":
|
||||
start = today.minusDays(29);
|
||||
end = today;
|
||||
break;
|
||||
case "THIS_YEAR":
|
||||
start = today.withDayOfYear(1);
|
||||
end = today;
|
||||
break;
|
||||
case "LAST_YEAR":
|
||||
start = today.withDayOfYear(1).minusYears(1);
|
||||
end = start.withDayOfYear(start.lengthOfYear());
|
||||
break;
|
||||
case "CUSTOM": {
|
||||
LocalDate customStart = parseDate(dateRange.path("start").asText(null));
|
||||
LocalDate customEnd = parseDate(dateRange.path("end").asText(null));
|
||||
if (customStart == null && customEnd == null) {
|
||||
return null;
|
||||
}
|
||||
start = customStart != null ? customStart : customEnd;
|
||||
end = customEnd != null ? customEnd : customStart;
|
||||
break;
|
||||
}
|
||||
default:
|
||||
return null;
|
||||
}
|
||||
return new String[]{
|
||||
start.atStartOfDay().format(DATETIME_FMT),
|
||||
end.atTime(23, 59, 59).format(DATETIME_FMT)
|
||||
};
|
||||
}
|
||||
|
||||
/**
|
||||
* 解析日期文本，失败返回 null。
|
||||
*
|
||||
* @param text 日期文本
|
||||
* @return 日期对象
|
||||
*/
|
||||
private LocalDate parseDate(String text) {
|
||||
if (isBlank(text)) {
|
||||
return null;
|
||||
}
|
||||
try {
|
||||
return LocalDate.parse(text.trim().substring(0, 10), DATE_FMT);
|
||||
} catch (Exception e) {
|
||||
return null;
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 按名称匹配分类：优先精确匹配，其次包含匹配。
|
||||
*
|
||||
* @param categories 分类列表
|
||||
* @param name 目标名称
|
||||
* @return 匹配到的分类，未命中返回 null
|
||||
*/
|
||||
private Category matchCategory(List<Category> categories, String name) {
|
||||
if (isBlank(name)) {
|
||||
return null;
|
||||
}
|
||||
String target = name.trim();
|
||||
for (Category category : categories) {
|
||||
if (category.getName().equalsIgnoreCase(target)) {
|
||||
return category;
|
||||
}
|
||||
}
|
||||
for (Category category : categories) {
|
||||
if (category.getName().contains(target) || target.contains(category.getName())) {
|
||||
return category;
|
||||
}
|
||||
}
|
||||
return null;
|
||||
}
|
||||
|
||||
/**
|
||||
* 按名称匹配标签：优先精确匹配，其次包含匹配。
|
||||
*
|
||||
* @param tags 标签列表
|
||||
* @param name 目标名称
|
||||
* @return 匹配到的标签，未命中返回 null
|
||||
*/
|
||||
private Tag matchTag(List<Tag> tags, String name) {
|
||||
if (isBlank(name)) {
|
||||
return null;
|
||||
}
|
||||
String target = name.trim();
|
||||
for (Tag tag : tags) {
|
||||
if (tag.getName().equalsIgnoreCase(target)) {
|
||||
return tag;
|
||||
}
|
||||
}
|
||||
for (Tag tag : tags) {
|
||||
if (tag.getName().contains(target) || target.contains(tag.getName())) {
|
||||
return tag;
|
||||
}
|
||||
}
|
||||
return null;
|
||||
}
|
||||
|
||||
/**
|
||||
* 过滤出合法且去重的枚举值。
|
||||
*
|
||||
* @param node 模型输出节点
|
||||
* @param valid 合法值列表
|
||||
* @return 过滤后的列表
|
||||
*/
|
||||
private List<String> filterValid(JsonNode node, List<String> valid) {
|
||||
List<String> result = new ArrayList<>();
|
||||
if (node != null && node.isArray()) {
|
||||
for (JsonNode item : node) {
|
||||
String value = item.asText("").toUpperCase();
|
||||
if (valid.contains(value) && !result.contains(value)) {
|
||||
result.add(value);
|
||||
}
|
||||
}
|
||||
}
|
||||
return result;
|
||||
}
|
||||
|
||||
/**
|
||||
* 将节点转换为金额，非数字时返回 null。
|
||||
*
|
||||
* @param node 模型输出节点
|
||||
* @return 金额
|
||||
*/
|
||||
private BigDecimal decimalOrNull(JsonNode node) {
|
||||
if (node == null || node.isMissingNode() || node.isNull()) {
|
||||
return null;
|
||||
}
|
||||
try {
|
||||
return node.decimalValue();
|
||||
} catch (Exception e) {
|
||||
return null;
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 去除模型可能附加的 Markdown 代码块围栏。
|
||||
*
|
||||
* @param content 原始内容
|
||||
* @return 清理后的内容
|
||||
*/
|
||||
private String stripCodeFence(String content) {
|
||||
if (content == null) {
|
||||
return "";
|
||||
}
|
||||
String cleaned = content.trim();
|
||||
if (cleaned.startsWith("```")) {
|
||||
int firstLineEnd = cleaned.indexOf('\n');
|
||||
if (firstLineEnd > 0) {
|
||||
cleaned = cleaned.substring(firstLineEnd + 1);
|
||||
}
|
||||
if (cleaned.endsWith("```")) {
|
||||
cleaned = cleaned.substring(0, cleaned.length() - 3);
|
||||
}
|
||||
}
|
||||
return cleaned.trim();
|
||||
}
|
||||
|
||||
/**
|
||||
* 判断字符串是否为空白。
|
||||
*
|
||||
* @param value 待判断字符串
|
||||
* @return 为 null 或全空白时返回 true
|
||||
*/
|
||||
private boolean isBlank(String value) {
|
||||
return value == null || value.isBlank();
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,351 @@
|
||||
package cn.rosercode.finance.service.impl;
|
||||
|
||||
import com.baomidou.mybatisplus.core.conditions.query.LambdaQueryWrapper;
|
||||
import cn.rosercode.finance.client.DeepSeekClient;
|
||||
import cn.rosercode.finance.client.DeepSeekResponse;
|
||||
import cn.rosercode.finance.common.BusinessException;
|
||||
import cn.rosercode.finance.entity.AiConfig;
|
||||
import cn.rosercode.finance.entity.Category;
|
||||
import cn.rosercode.finance.entity.Tag;
|
||||
import cn.rosercode.finance.enums.AiCallType;
|
||||
import cn.rosercode.finance.mapper.CategoryMapper;
|
||||
import cn.rosercode.finance.mapper.TagMapper;
|
||||
import cn.rosercode.finance.service.AiConfigService;
|
||||
import cn.rosercode.finance.service.AiRecordService;
|
||||
import cn.rosercode.finance.service.AiUsageService;
|
||||
import cn.rosercode.finance.vo.AiRecordResultVO;
|
||||
import com.fasterxml.jackson.core.JsonProcessingException;
|
||||
import com.fasterxml.jackson.databind.JsonNode;
|
||||
import com.fasterxml.jackson.databind.ObjectMapper;
|
||||
import org.springframework.stereotype.Service;
|
||||
|
||||
import java.math.BigDecimal;
|
||||
import java.time.LocalDate;
|
||||
import java.time.LocalDateTime;
|
||||
import java.time.format.DateTimeFormatter;
|
||||
import java.util.ArrayList;
|
||||
import java.util.List;
|
||||
import java.util.stream.Collectors;
|
||||
|
||||
/**
|
||||
* AI 自然语言记账服务实现类，负责提示词组装、时间推算与名称映射。
|
||||
*
|
||||
* @date 2026-09-17 18:23:25
|
||||
*/
|
||||
@Service
|
||||
public class AiRecordServiceImpl implements AiRecordService {
|
||||
|
||||
/** 日期时间格式，与前端表单保持一致 */
|
||||
private static final DateTimeFormatter DATETIME_FMT = DateTimeFormatter.ofPattern("yyyy-MM-dd'T'HH:mm:ss");
|
||||
|
||||
/** 合法的收支类型 */
|
||||
private static final List<String> VALID_TYPES = List.of("INCOME", "EXPENSE");
|
||||
|
||||
/** 合法的支付方式 */
|
||||
private static final List<String> VALID_PAYMENTS = List.of("WECHAT", "ALIPAY", "CASH", "BANK");
|
||||
|
||||
private final AiConfigService aiConfigService;
|
||||
private final CategoryMapper categoryMapper;
|
||||
private final TagMapper tagMapper;
|
||||
private final DeepSeekClient deepSeekClient;
|
||||
private final AiUsageService aiUsageService;
|
||||
private final ObjectMapper objectMapper;
|
||||
|
||||
public AiRecordServiceImpl(AiConfigService aiConfigService,
|
||||
CategoryMapper categoryMapper,
|
||||
TagMapper tagMapper,
|
||||
DeepSeekClient deepSeekClient,
|
||||
AiUsageService aiUsageService,
|
||||
ObjectMapper objectMapper) {
|
||||
this.aiConfigService = aiConfigService;
|
||||
this.categoryMapper = categoryMapper;
|
||||
this.tagMapper = tagMapper;
|
||||
this.deepSeekClient = deepSeekClient;
|
||||
this.aiUsageService = aiUsageService;
|
||||
this.objectMapper = objectMapper;
|
||||
}
|
||||
|
||||
/**
|
||||
* {@inheritDoc}
|
||||
*
|
||||
* 将用户分类、标签与当前日期时间注入提示词，解析模型输出并映射为交易记录字段。
|
||||
*/
|
||||
@Override
|
||||
public AiRecordResultVO parse(Long userId, String text) {
|
||||
if (text == null || text.isBlank()) {
|
||||
throw new BusinessException("请输入记账内容");
|
||||
}
|
||||
AiConfig config = aiConfigService.getRawConfig(userId);
|
||||
if (config == null || !Boolean.TRUE.equals(config.getEnabled())) {
|
||||
throw new BusinessException("AI 功能未启用，请先在 AI 设置中开启");
|
||||
}
|
||||
if (isBlank(config.getApiKey())) {
|
||||
throw new BusinessException("请先配置 DeepSeek API Key");
|
||||
}
|
||||
|
||||
List<Category> categories = categoryMapper.selectList(new LambdaQueryWrapper<Category>()
|
||||
.eq(Category::getUserId, userId));
|
||||
List<Tag> tags = tagMapper.selectList(new LambdaQueryWrapper<Tag>()
|
||||
.eq(Tag::getUserId, userId));
|
||||
|
||||
String systemPrompt = buildSystemPrompt(categories, tags);
|
||||
long start = System.currentTimeMillis();
|
||||
DeepSeekResponse aiResponse;
|
||||
try {
|
||||
aiResponse = deepSeekClient.chat(config.getBaseUrl(), config.getApiKey(), config.getModel(),
|
||||
systemPrompt, text);
|
||||
} catch (RuntimeException e) {
|
||||
// 失败也记录一次使用，便于排查问题
|
||||
aiUsageService.recordFailure(userId, AiCallType.RECORD, config.getModel(), 0,
|
||||
System.currentTimeMillis() - start, e.getMessage());
|
||||
throw e;
|
||||
}
|
||||
aiUsageService.recordSuccess(userId, AiCallType.RECORD, config.getModel(), 0,
|
||||
aiResponse, System.currentTimeMillis() - start);
|
||||
|
||||
return parseResult(aiResponse.getContent(), categories, tags);
|
||||
}
|
||||
|
||||
/**
|
||||
* 构建系统提示词，注入可用分类、标签与当前日期时间。
|
||||
*
|
||||
* @param categories 用户分类列表
|
||||
* @param tags 用户标签列表
|
||||
* @return 系统提示词
|
||||
*/
|
||||
private String buildSystemPrompt(List<Category> categories, List<Tag> tags) {
|
||||
String categoryNames = categories.stream().map(Category::getName).collect(Collectors.joining("、"));
|
||||
String tagNames = tags.stream().map(Tag::getName).collect(Collectors.joining("、"));
|
||||
String now = LocalDateTime.now().format(DateTimeFormatter.ofPattern("yyyy-MM-dd HH:mm:ss"));
|
||||
return "你是一个记账助手。请把用户的自然语言记账内容解析为一条结构化交易记录，只输出 JSON，不要输出任何多余内容。\n"
|
||||
+ "可用的收支类型：INCOME（收入）、EXPENSE（支出）。\n"
|
||||
+ "可用的支付方式：WECHAT（微信）、ALIPAY（支付宝）、CASH（现金）、BANK（银行卡）。\n"
|
||||
+ "可用分类：" + (categoryNames.isEmpty() ? "无" : categoryNames) + "。\n"
|
||||
+ "可用标签：" + (tagNames.isEmpty() ? "无" : tagNames) + "。\n"
|
||||
+ "规则：1. type 默认为 EXPENSE，除非明显是收入；2. amount 为必填数字；"
|
||||
+ "3. categoryNames 从可用分类中选择最合适的一个（数组，最多一个），无法判断则为空数组；"
|
||||
+ "4. tagNames 从可用标签中选择（数组，可为空）；5. paymentMethod 从可用支付方式中选择，无法判断为 null；"
|
||||
+ "6. transactionTime 输出 yyyy-MM-ddTHH:mm:ss，未提及时间用当前时间，相对时间（如昨天、上周五）请换算；"
|
||||
+ "7. includeInStats 默认 true；8. remark 给出简洁备注。\n"
|
||||
+ "当前时间：" + now + "。\n"
|
||||
+ "输出格式：{\"summary\":\"简要说明识别到的记账内容\",\"type\":\"EXPENSE\",\"amount\":35.5,"
|
||||
+ "\"categoryNames\":[\"餐饮\"],\"tagNames\":[],\"paymentMethod\":\"ALIPAY\","
|
||||
+ "\"transactionTime\":\"" + LocalDateTime.now().format(DATETIME_FMT) + "\",\"remark\":\"午餐\","
|
||||
+ "\"includeInStats\":true,\"unmatched\":[]}";
|
||||
}
|
||||
|
||||
/**
|
||||
* 解析模型返回的 JSON 并映射为交易记录字段。
|
||||
*
|
||||
* @param content 模型输出内容
|
||||
* @param categories 用户分类列表
|
||||
* @param tags 用户标签列表
|
||||
* @return 结构化交易记录
|
||||
*/
|
||||
private AiRecordResultVO parseResult(String content, List<Category> categories, List<Tag> tags) {
|
||||
JsonNode root;
|
||||
try {
|
||||
root = objectMapper.readTree(stripCodeFence(content));
|
||||
} catch (JsonProcessingException e) {
|
||||
throw new BusinessException("AI 返回内容不是有效 JSON，请重试");
|
||||
}
|
||||
|
||||
AiRecordResultVO vo = new AiRecordResultVO();
|
||||
vo.setSummary(root.path("summary").asText(""));
|
||||
|
||||
// 类型：仅接受合法值，默认支出
|
||||
String type = root.path("type").asText("EXPENSE").toUpperCase();
|
||||
vo.setType(VALID_TYPES.contains(type) ? type : "EXPENSE");
|
||||
|
||||
// 金额
|
||||
vo.setAmount(decimalOrNull(root.path("amount")));
|
||||
if (vo.getAmount() == null || vo.getAmount().compareTo(BigDecimal.ZERO) <= 0) {
|
||||
throw new BusinessException("未能识别到有效金额，请补充金额后重试");
|
||||
}
|
||||
|
||||
// 分类：取第一个匹配项
|
||||
JsonNode categoryNode = root.path("categoryNames");
|
||||
if (categoryNode.isArray()) {
|
||||
for (JsonNode node : categoryNode) {
|
||||
Category category = matchCategory(categories, node.asText(""));
|
||||
if (category != null) {
|
||||
vo.setCategoryId(category.getId());
|
||||
vo.setCategoryName(category.getName());
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// 标签
|
||||
List<Long> tagIds = new ArrayList<>();
|
||||
List<String> tagNames = new ArrayList<>();
|
||||
JsonNode tagNode = root.path("tagNames");
|
||||
if (tagNode.isArray()) {
|
||||
for (JsonNode node : tagNode) {
|
||||
Tag tag = matchTag(tags, node.asText(""));
|
||||
if (tag != null && !tagIds.contains(tag.getId())) {
|
||||
tagIds.add(tag.getId());
|
||||
tagNames.add(tag.getName());
|
||||
}
|
||||
}
|
||||
}
|
||||
vo.setTagIds(tagIds);
|
||||
vo.setTagNames(tagNames);
|
||||
|
||||
// 支付方式
|
||||
String payment = root.path("paymentMethod").asText("").toUpperCase();
|
||||
if (VALID_PAYMENTS.contains(payment)) {
|
||||
vo.setPaymentMethod(payment);
|
||||
}
|
||||
|
||||
// 交易时间
|
||||
vo.setTransactionTime(resolveTransactionTime(root.path("transactionTime")));
|
||||
|
||||
String remark = root.path("remark").asText(null);
|
||||
vo.setRemark(isBlank(remark) ? null : remark.trim());
|
||||
|
||||
JsonNode statsNode = root.path("includeInStats");
|
||||
vo.setIncludeInStats(statsNode.isMissingNode() || statsNode.isNull() ? Boolean.TRUE : statsNode.asBoolean(true));
|
||||
|
||||
// 未识别关键词
|
||||
List<String> unmatched = new ArrayList<>();
|
||||
JsonNode unmatchedNode = root.path("unmatched");
|
||||
if (unmatchedNode.isArray()) {
|
||||
for (JsonNode node : unmatchedNode) {
|
||||
String value = node.asText("");
|
||||
if (!isBlank(value)) {
|
||||
unmatched.add(value);
|
||||
}
|
||||
}
|
||||
}
|
||||
vo.setUnmatched(unmatched);
|
||||
return vo;
|
||||
}
|
||||
|
||||
/**
|
||||
* 解析模型给出的交易时间，无法解析时回退当前时间。
|
||||
*
|
||||
* @param node 模型输出的时间节点
|
||||
* @return 格式化后的时间字符串
|
||||
*/
|
||||
private String resolveTransactionTime(JsonNode node) {
|
||||
LocalDateTime now = LocalDateTime.now();
|
||||
String text = node == null || node.isMissingNode() || node.isNull() ? null : node.asText(null);
|
||||
if (isBlank(text)) {
|
||||
return now.format(DATETIME_FMT);
|
||||
}
|
||||
String value = text.trim().replace(' ', 'T');
|
||||
try {
|
||||
// 完整日期时间
|
||||
return LocalDateTime.parse(value.substring(0, Math.min(value.length(), 19))).format(DATETIME_FMT);
|
||||
} catch (Exception e) {
|
||||
// 仅日期：补齐当前时分秒
|
||||
try {
|
||||
LocalDate date = LocalDate.parse(value.substring(0, 10));
|
||||
return date.atTime(now.getHour(), now.getMinute(), now.getSecond()).format(DATETIME_FMT);
|
||||
} catch (Exception ex) {
|
||||
return now.format(DATETIME_FMT);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 按名称匹配分类：优先精确匹配，其次包含匹配。
|
||||
*
|
||||
* @param categories 分类列表
|
||||
* @param name 目标名称
|
||||
* @return 匹配到的分类，未命中返回 null
|
||||
*/
|
||||
private Category matchCategory(List<Category> categories, String name) {
|
||||
if (isBlank(name)) {
|
||||
return null;
|
||||
}
|
||||
String target = name.trim();
|
||||
for (Category category : categories) {
|
||||
if (category.getName().equalsIgnoreCase(target)) {
|
||||
return category;
|
||||
}
|
||||
}
|
||||
for (Category category : categories) {
|
||||
if (category.getName().contains(target) || target.contains(category.getName())) {
|
||||
return category;
|
||||
}
|
||||
}
|
||||
return null;
|
||||
}
|
||||
|
||||
/**
|
||||
* 按名称匹配标签：优先精确匹配，其次包含匹配。
|
||||
*
|
||||
* @param tags 标签列表
|
||||
* @param name 目标名称
|
||||
* @return 匹配到的标签，未命中返回 null
|
||||
*/
|
||||
private Tag matchTag(List<Tag> tags, String name) {
|
||||
if (isBlank(name)) {
|
||||
return null;
|
||||
}
|
||||
String target = name.trim();
|
||||
for (Tag tag : tags) {
|
||||
if (tag.getName().equalsIgnoreCase(target)) {
|
||||
return tag;
|
||||
}
|
||||
}
|
||||
for (Tag tag : tags) {
|
||||
if (tag.getName().contains(target) || target.contains(tag.getName())) {
|
||||
return tag;
|
||||
}
|
||||
}
|
||||
return null;
|
||||
}
|
||||
|
||||
/**
|
||||
* 将节点转换为金额，非数字时返回 null。
|
||||
*
|
||||
* @param node 模型输出节点
|
||||
* @return 金额
|
||||
*/
|
||||
private BigDecimal decimalOrNull(JsonNode node) {
|
||||
if (node == null || node.isMissingNode() || node.isNull()) {
|
||||
return null;
|
||||
}
|
||||
try {
|
||||
return node.decimalValue();
|
||||
} catch (Exception e) {
|
||||
return null;
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 去除模型可能附加的 Markdown 代码块围栏。
|
||||
*
|
||||
* @param content 原始内容
|
||||
* @return 清理后的内容
|
||||
*/
|
||||
private String stripCodeFence(String content) {
|
||||
if (content == null) {
|
||||
return "";
|
||||
}
|
||||
String cleaned = content.trim();
|
||||
if (cleaned.startsWith("```")) {
|
||||
int firstLineEnd = cleaned.indexOf('\n');
|
||||
if (firstLineEnd > 0) {
|
||||
cleaned = cleaned.substring(firstLineEnd + 1);
|
||||
}
|
||||
if (cleaned.endsWith("```")) {
|
||||
cleaned = cleaned.substring(0, cleaned.length() - 3);
|
||||
}
|
||||
}
|
||||
return cleaned.trim();
|
||||
}
|
||||
|
||||
/**
|
||||
* 判断字符串是否为空白。
|
||||
*
|
||||
* @param value 待判断字符串
|
||||
* @return 为 null 或全空白时返回 true
|
||||
*/
|
||||
private boolean isBlank(String value) {
|
||||
return value == null || value.isBlank();
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,386 @@
|
||||
package cn.rosercode.finance.service.impl;
|
||||
|
||||
import com.baomidou.mybatisplus.core.conditions.query.LambdaQueryWrapper;
|
||||
import com.baomidou.mybatisplus.core.conditions.update.LambdaUpdateWrapper;
|
||||
import cn.rosercode.finance.client.DeepSeekClient;
|
||||
import cn.rosercode.finance.client.DeepSeekResponse;
|
||||
import cn.rosercode.finance.common.BusinessException;
|
||||
import cn.rosercode.finance.dto.AiApplyRequest;
|
||||
import cn.rosercode.finance.entity.AiConfig;
|
||||
import cn.rosercode.finance.entity.Category;
|
||||
import cn.rosercode.finance.entity.Tag;
|
||||
import cn.rosercode.finance.entity.Transaction;
|
||||
import cn.rosercode.finance.entity.TransactionTag;
|
||||
import cn.rosercode.finance.enums.AiCallType;
|
||||
import cn.rosercode.finance.mapper.CategoryMapper;
|
||||
import cn.rosercode.finance.mapper.TagMapper;
|
||||
import cn.rosercode.finance.mapper.TransactionMapper;
|
||||
import cn.rosercode.finance.mapper.TransactionTagMapper;
|
||||
import cn.rosercode.finance.service.AiConfigService;
|
||||
import cn.rosercode.finance.service.AiSuggestionService;
|
||||
import cn.rosercode.finance.service.AiUsageService;
|
||||
import cn.rosercode.finance.vo.AiSuggestionVO;
|
||||
import com.fasterxml.jackson.core.JsonProcessingException;
|
||||
import com.fasterxml.jackson.databind.JsonNode;
|
||||
import com.fasterxml.jackson.databind.ObjectMapper;
|
||||
import org.springframework.stereotype.Service;
|
||||
import org.springframework.transaction.annotation.Transactional;
|
||||
|
||||
import java.util.ArrayList;
|
||||
import java.util.LinkedHashMap;
|
||||
import java.util.List;
|
||||
import java.util.Map;
|
||||
import java.util.Set;
|
||||
import java.util.function.Function;
|
||||
import java.util.stream.Collectors;
|
||||
|
||||
/**
|
||||
* AI 建议服务实现类，负责组装提示词、调用 DeepSeek、校验并应用建议。
|
||||
*
|
||||
* @date 2026-09-17 17:52:35
|
||||
*/
|
||||
@Service
|
||||
public class AiSuggestionServiceImpl implements AiSuggestionService {
|
||||
|
||||
/** 单次最多分析的记录数 */
|
||||
private static final int MAX_SUGGEST_SIZE = 100;
|
||||
|
||||
/** 原始账单数据注入提示词时的最大长度 */
|
||||
private static final int RAW_DATA_LIMIT = 300;
|
||||
|
||||
/** 系统提示词，要求模型严格返回 JSON */
|
||||
private static final String SYSTEM_PROMPT = "你是一个专业的个人记账助手。请根据每条交易记录的类型、金额、备注等信息，"
|
||||
+ "从给定的分类列表和标签列表中，为该记录选择最合适的分类与标签。"
|
||||
+ "要求：1. 分类必须与记录的类型一致（INCOME 只能选 INCOME 分类，EXPENSE 只能选 EXPENSE 分类）；"
|
||||
+ "2. 只能使用给定列表中的 ID，不要编造；3. 标签可多选，也可为空数组；"
|
||||
+ "4. 只输出 JSON，不要输出任何多余内容。"
|
||||
+ "输出格式：{\"suggestions\":[{\"id\":记录ID,\"categoryId\":分类ID,\"tagIds\":[标签ID],\"reason\":\"简要中文理由\"}]}";
|
||||
|
||||
private final AiConfigService aiConfigService;
|
||||
private final TransactionMapper transactionMapper;
|
||||
private final CategoryMapper categoryMapper;
|
||||
private final TagMapper tagMapper;
|
||||
private final TransactionTagMapper transactionTagMapper;
|
||||
private final DeepSeekClient deepSeekClient;
|
||||
private final ObjectMapper objectMapper;
|
||||
private final AiUsageService aiUsageService;
|
||||
|
||||
public AiSuggestionServiceImpl(AiConfigService aiConfigService,
|
||||
TransactionMapper transactionMapper,
|
||||
CategoryMapper categoryMapper,
|
||||
TagMapper tagMapper,
|
||||
TransactionTagMapper transactionTagMapper,
|
||||
DeepSeekClient deepSeekClient,
|
||||
ObjectMapper objectMapper,
|
||||
AiUsageService aiUsageService) {
|
||||
this.aiConfigService = aiConfigService;
|
||||
this.transactionMapper = transactionMapper;
|
||||
this.categoryMapper = categoryMapper;
|
||||
this.tagMapper = tagMapper;
|
||||
this.transactionTagMapper = transactionTagMapper;
|
||||
this.deepSeekClient = deepSeekClient;
|
||||
this.objectMapper = objectMapper;
|
||||
this.aiUsageService = aiUsageService;
|
||||
}
|
||||
|
||||
/**
|
||||
* {@inheritDoc}
|
||||
*
|
||||
* 校验配置与数量后，将分类、标签与交易记录注入提示词，调用模型并解析校验结果。
|
||||
*/
|
||||
@Override
|
||||
public List<AiSuggestionVO> suggest(Long userId, List<Long> ids) {
|
||||
if (ids == null || ids.isEmpty()) {
|
||||
throw new BusinessException("请至少选择一条记录");
|
||||
}
|
||||
if (ids.size() > MAX_SUGGEST_SIZE) {
|
||||
throw new BusinessException("单次最多分析 " + MAX_SUGGEST_SIZE + " 条记录，请缩小范围后重试");
|
||||
}
|
||||
AiConfig config = aiConfigService.getRawConfig(userId);
|
||||
if (config == null || !Boolean.TRUE.equals(config.getEnabled())) {
|
||||
throw new BusinessException("AI 智能建议未启用，请先在 AI 设置中开启");
|
||||
}
|
||||
if (isBlank(config.getApiKey())) {
|
||||
throw new BusinessException("请先配置 DeepSeek API Key");
|
||||
}
|
||||
|
||||
// 仅分析属于当前用户的记录
|
||||
List<Transaction> transactions = transactionMapper.selectList(new LambdaQueryWrapper<Transaction>()
|
||||
.eq(Transaction::getUserId, userId)
|
||||
.in(Transaction::getId, ids));
|
||||
if (transactions.isEmpty()) {
|
||||
throw new BusinessException("未找到可分析的交易记录");
|
||||
}
|
||||
|
||||
List<Category> categories = categoryMapper.selectList(new LambdaQueryWrapper<Category>()
|
||||
.eq(Category::getUserId, userId));
|
||||
if (categories.isEmpty()) {
|
||||
throw new BusinessException("请先创建分类后再使用 AI 建议");
|
||||
}
|
||||
List<Tag> tags = tagMapper.selectList(new LambdaQueryWrapper<Tag>()
|
||||
.eq(Tag::getUserId, userId));
|
||||
|
||||
String userPrompt = buildUserPrompt(categories, tags, transactions);
|
||||
long start = System.currentTimeMillis();
|
||||
DeepSeekResponse aiResponse;
|
||||
try {
|
||||
aiResponse = deepSeekClient.chat(config.getBaseUrl(), config.getApiKey(), config.getModel(),
|
||||
SYSTEM_PROMPT, userPrompt);
|
||||
} catch (RuntimeException e) {
|
||||
// 失败也记录一次使用，便于排查问题
|
||||
aiUsageService.recordFailure(userId, AiCallType.SUGGEST, config.getModel(), transactions.size(),
|
||||
System.currentTimeMillis() - start, e.getMessage());
|
||||
throw e;
|
||||
}
|
||||
aiUsageService.recordSuccess(userId, AiCallType.SUGGEST, config.getModel(), transactions.size(),
|
||||
aiResponse, System.currentTimeMillis() - start);
|
||||
|
||||
List<AiSuggestionVO> result = parseSuggestions(aiResponse.getContent(), transactions, categories, tags);
|
||||
if (result.isEmpty()) {
|
||||
throw new BusinessException("AI 未返回有效建议，请重试");
|
||||
}
|
||||
return result;
|
||||
}
|
||||
|
||||
/**
|
||||
* {@inheritDoc}
|
||||
*
|
||||
* 分类覆盖写入，标签在保留原有基础上追加。
|
||||
*/
|
||||
@Override
|
||||
@Transactional
|
||||
public int apply(Long userId, AiApplyRequest request) {
|
||||
List<AiApplyRequest.Item> items = request.getSuggestions();
|
||||
List<Long> transactionIds = items.stream()
|
||||
.map(AiApplyRequest.Item::getTransactionId)
|
||||
.distinct()
|
||||
.collect(Collectors.toList());
|
||||
|
||||
// 仅处理属于当前用户的记录
|
||||
Map<Long, Transaction> transactionMap = transactionMapper.selectList(new LambdaQueryWrapper<Transaction>()
|
||||
.eq(Transaction::getUserId, userId)
|
||||
.in(Transaction::getId, transactionIds))
|
||||
.stream().collect(Collectors.toMap(Transaction::getId, Function.identity()));
|
||||
if (transactionMap.isEmpty()) {
|
||||
return 0;
|
||||
}
|
||||
|
||||
Map<Long, Category> categoryMap = categoryMapper.selectList(new LambdaQueryWrapper<Category>()
|
||||
.eq(Category::getUserId, userId))
|
||||
.stream().collect(Collectors.toMap(Category::getId, Function.identity()));
|
||||
Map<Long, Tag> tagMap = tagMapper.selectList(new LambdaQueryWrapper<Tag>()
|
||||
.eq(Tag::getUserId, userId))
|
||||
.stream().collect(Collectors.toMap(Tag::getId, Function.identity()));
|
||||
|
||||
int applied = 0;
|
||||
for (AiApplyRequest.Item item : items) {
|
||||
Transaction transaction = transactionMap.get(item.getTransactionId());
|
||||
if (transaction == null) {
|
||||
continue;
|
||||
}
|
||||
boolean changed = false;
|
||||
|
||||
// 分类：仅接受类型匹配且确属当前用户的分类
|
||||
if (item.getCategoryId() != null) {
|
||||
Category category = categoryMap.get(item.getCategoryId());
|
||||
if (category != null
|
||||
&& category.getType().equals(transaction.getType())
|
||||
&& !category.getId().equals(transaction.getCategoryId())) {
|
||||
transactionMapper.update(null, new LambdaUpdateWrapper<Transaction>()
|
||||
.eq(Transaction::getId, transaction.getId())
|
||||
.eq(Transaction::getUserId, userId)
|
||||
.set(Transaction::getCategoryId, category.getId()));
|
||||
changed = true;
|
||||
}
|
||||
}
|
||||
|
||||
// 标签：追加尚未关联的有效标签
|
||||
List<Long> validTagIds = item.getTagIds() == null ? List.of() : item.getTagIds().stream()
|
||||
.filter(tagMap::containsKey)
|
||||
.distinct()
|
||||
.collect(Collectors.toList());
|
||||
if (!validTagIds.isEmpty()) {
|
||||
Set<Long> existingTagIds = transactionTagMapper.selectList(new LambdaQueryWrapper<TransactionTag>()
|
||||
.eq(TransactionTag::getTransactionId, transaction.getId()))
|
||||
.stream().map(TransactionTag::getTagId).collect(Collectors.toSet());
|
||||
for (Long tagId : validTagIds) {
|
||||
if (existingTagIds.contains(tagId)) {
|
||||
continue;
|
||||
}
|
||||
TransactionTag relation = new TransactionTag();
|
||||
relation.setTransactionId(transaction.getId());
|
||||
relation.setTagId(tagId);
|
||||
transactionTagMapper.insert(relation);
|
||||
changed = true;
|
||||
}
|
||||
}
|
||||
|
||||
if (changed) {
|
||||
applied++;
|
||||
}
|
||||
}
|
||||
return applied;
|
||||
}
|
||||
|
||||
/**
|
||||
* 构建用户提示词，将分类、标签与交易记录序列化为 JSON。
|
||||
*
|
||||
* @param categories 用户分类列表
|
||||
* @param tags 用户标签列表
|
||||
* @param transactions 待分析记录列表
|
||||
* @return JSON 格式的用户提示词
|
||||
*/
|
||||
private String buildUserPrompt(List<Category> categories, List<Tag> tags, List<Transaction> transactions) {
|
||||
Map<String, Object> payload = new LinkedHashMap<>();
|
||||
payload.put("categories", categories.stream().map(category -> {
|
||||
Map<String, Object> node = new LinkedHashMap<>();
|
||||
node.put("id", category.getId());
|
||||
node.put("name", category.getName());
|
||||
node.put("type", category.getType());
|
||||
return node;
|
||||
}).collect(Collectors.toList()));
|
||||
payload.put("tags", tags.stream().map(tag -> {
|
||||
Map<String, Object> node = new LinkedHashMap<>();
|
||||
node.put("id", tag.getId());
|
||||
node.put("name", tag.getName());
|
||||
return node;
|
||||
}).collect(Collectors.toList()));
|
||||
payload.put("transactions", transactions.stream().map(transaction -> {
|
||||
Map<String, Object> node = new LinkedHashMap<>();
|
||||
node.put("id", transaction.getId());
|
||||
node.put("type", transaction.getType());
|
||||
node.put("amount", transaction.getAmount());
|
||||
node.put("time", transaction.getTransactionTime());
|
||||
node.put("paymentMethod", transaction.getPaymentMethod());
|
||||
node.put("remark", transaction.getRemark());
|
||||
node.put("extra", extractExtra(transaction.getRawData()));
|
||||
return node;
|
||||
}).collect(Collectors.toList()));
|
||||
|
||||
try {
|
||||
return objectMapper.writeValueAsString(payload);
|
||||
} catch (JsonProcessingException e) {
|
||||
throw new BusinessException("构建 AI 提示词失败: " + e.getMessage());
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 解析模型返回的 JSON，并校验分类与标签的合法性。
|
||||
*
|
||||
* @param content 模型输出内容
|
||||
* @param transactions 参与分析的记录列表
|
||||
* @param categories 用户分类列表
|
||||
* @param tags 用户标签列表
|
||||
* @return 有效建议列表
|
||||
*/
|
||||
private List<AiSuggestionVO> parseSuggestions(String content,
|
||||
List<Transaction> transactions,
|
||||
List<Category> categories,
|
||||
List<Tag> tags) {
|
||||
JsonNode root;
|
||||
try {
|
||||
root = objectMapper.readTree(stripCodeFence(content));
|
||||
} catch (JsonProcessingException e) {
|
||||
throw new BusinessException("AI 返回内容不是有效 JSON，请重试");
|
||||
}
|
||||
JsonNode array = root.path("suggestions");
|
||||
if (!array.isArray()) {
|
||||
throw new BusinessException("AI 返回格式不正确，请重试");
|
||||
}
|
||||
|
||||
Map<Long, Transaction> transactionMap = transactions.stream()
|
||||
.collect(Collectors.toMap(Transaction::getId, Function.identity()));
|
||||
Map<Long, Category> categoryMap = categories.stream()
|
||||
.collect(Collectors.toMap(Category::getId, Function.identity()));
|
||||
Map<Long, Tag> tagMap = tags.stream()
|
||||
.collect(Collectors.toMap(Tag::getId, Function.identity()));
|
||||
|
||||
List<AiSuggestionVO> result = new ArrayList<>();
|
||||
for (JsonNode node : array) {
|
||||
long id = node.path("id").asLong(0);
|
||||
Transaction transaction = transactionMap.get(id);
|
||||
if (transaction == null) {
|
||||
continue;
|
||||
}
|
||||
AiSuggestionVO vo = new AiSuggestionVO();
|
||||
vo.setTransactionId(id);
|
||||
|
||||
// 分类必须存在且与记录类型一致
|
||||
long categoryId = node.path("categoryId").asLong(0);
|
||||
Category category = categoryMap.get(categoryId);
|
||||
if (category != null && category.getType().equals(transaction.getType())) {
|
||||
vo.setCategoryId(category.getId());
|
||||
vo.setCategoryName(category.getName());
|
||||
}
|
||||
|
||||
// 标签过滤为属于当前用户的标签
|
||||
List<Long> tagIds = new ArrayList<>();
|
||||
List<String> tagNames = new ArrayList<>();
|
||||
JsonNode tagNode = node.path("tagIds");
|
||||
if (tagNode.isArray()) {
|
||||
for (JsonNode item : tagNode) {
|
||||
Tag tag = tagMap.get(item.asLong(0));
|
||||
if (tag != null && !tagIds.contains(tag.getId())) {
|
||||
tagIds.add(tag.getId());
|
||||
tagNames.add(tag.getName());
|
||||
}
|
||||
}
|
||||
}
|
||||
vo.setTagIds(tagIds);
|
||||
vo.setTagNames(tagNames);
|
||||
vo.setReason(node.path("reason").asText(""));
|
||||
|
||||
// 至少存在一项有效建议才返回
|
||||
if (vo.getCategoryId() != null || !tagIds.isEmpty()) {
|
||||
result.add(vo);
|
||||
}
|
||||
}
|
||||
return result;
|
||||
}
|
||||
|
||||
/**
|
||||
* 去除模型可能附加的 Markdown 代码块围栏。
|
||||
*
|
||||
* @param content 原始内容
|
||||
* @return 清理后的内容
|
||||
*/
|
||||
private String stripCodeFence(String content) {
|
||||
if (content == null) {
|
||||
return "";
|
||||
}
|
||||
String cleaned = content.trim();
|
||||
if (cleaned.startsWith("```")) {
|
||||
int firstLineEnd = cleaned.indexOf('\n');
|
||||
if (firstLineEnd > 0) {
|
||||
cleaned = cleaned.substring(firstLineEnd + 1);
|
||||
}
|
||||
if (cleaned.endsWith("```")) {
|
||||
cleaned = cleaned.substring(0, cleaned.length() - 3);
|
||||
}
|
||||
}
|
||||
return cleaned.trim();
|
||||
}
|
||||
|
||||
/**
|
||||
* 从原始账单数据中截取摘要，用于注入提示词。
|
||||
*
|
||||
* @param rawData 原始账单 JSON 字符串
|
||||
* @return 截断后的摘要，为空时返回 null
|
||||
*/
|
||||
private String extractExtra(String rawData) {
|
||||
if (isBlank(rawData)) {
|
||||
return null;
|
||||
}
|
||||
return rawData.length() <= RAW_DATA_LIMIT ? rawData : rawData.substring(0, RAW_DATA_LIMIT);
|
||||
}
|
||||
|
||||
/**
|
||||
* 判断字符串是否为空白。
|
||||
*
|
||||
* @param value 待判断字符串
|
||||
* @return 为 null 或全空白时返回 true
|
||||
*/
|
||||
private boolean isBlank(String value) {
|
||||
return value == null || value.isBlank();
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,173 @@
|
||||
package cn.rosercode.finance.service.impl;
|
||||
|
||||
import com.baomidou.mybatisplus.core.conditions.query.LambdaQueryWrapper;
|
||||
import cn.rosercode.finance.client.DeepSeekResponse;
|
||||
import cn.rosercode.finance.entity.AiUsage;
|
||||
import cn.rosercode.finance.enums.AiCallType;
|
||||
import cn.rosercode.finance.mapper.AiUsageMapper;
|
||||
import cn.rosercode.finance.service.AiUsageService;
|
||||
import cn.rosercode.finance.vo.AiUsageResultVO;
|
||||
import cn.rosercode.finance.vo.AiUsageSummaryVO;
|
||||
import cn.rosercode.finance.vo.AiUsageVO;
|
||||
import org.slf4j.Logger;
|
||||
import org.slf4j.LoggerFactory;
|
||||
import org.springframework.stereotype.Service;
|
||||
|
||||
import java.util.List;
|
||||
import java.util.stream.Collectors;
|
||||
|
||||
/**
|
||||
* AI 使用记录服务实现类，负责写入调用明细并汇总查询。
|
||||
*
|
||||
* @date 2026-09-17 18:02:23
|
||||
*/
|
||||
@Service
|
||||
public class AiUsageServiceImpl implements AiUsageService {
|
||||
|
||||
private static final Logger log = LoggerFactory.getLogger(AiUsageServiceImpl.class);
|
||||
|
||||
/** 最近明细条数上限 */
|
||||
private static final int MAX_RECENT_LIMIT = 100;
|
||||
|
||||
/** 失败原因最大长度 */
|
||||
private static final int ERROR_MESSAGE_LIMIT = 500;
|
||||
|
||||
private final AiUsageMapper aiUsageMapper;
|
||||
|
||||
public AiUsageServiceImpl(AiUsageMapper aiUsageMapper) {
|
||||
this.aiUsageMapper = aiUsageMapper;
|
||||
}
|
||||
|
||||
/**
|
||||
* {@inheritDoc}
|
||||
*
|
||||
* 记录写入失败仅打印告警，不影响主流程。
|
||||
*/
|
||||
@Override
|
||||
public void recordSuccess(Long userId, AiCallType type, String model, int transactionCount,
|
||||
DeepSeekResponse response, long durationMs) {
|
||||
AiUsage usage = new AiUsage();
|
||||
usage.setUserId(userId);
|
||||
usage.setType(type.name());
|
||||
usage.setModel(model);
|
||||
usage.setTransactionCount(transactionCount);
|
||||
usage.setPromptTokens(response != null ? response.getPromptTokens() : 0);
|
||||
usage.setCompletionTokens(response != null ? response.getCompletionTokens() : 0);
|
||||
usage.setTotalTokens(response != null ? response.getTotalTokens() : 0);
|
||||
usage.setCachedTokens(response != null ? response.getCachedTokens() : 0);
|
||||
usage.setSuccess(true);
|
||||
usage.setDurationMs(durationMs);
|
||||
saveQuietly(usage);
|
||||
}
|
||||
|
||||
/**
|
||||
* {@inheritDoc}
|
||||
*/
|
||||
@Override
|
||||
public void recordFailure(Long userId, AiCallType type, String model, int transactionCount,
|
||||
long durationMs, String errorMessage) {
|
||||
AiUsage usage = new AiUsage();
|
||||
usage.setUserId(userId);
|
||||
usage.setType(type.name());
|
||||
usage.setModel(model);
|
||||
usage.setTransactionCount(transactionCount);
|
||||
usage.setPromptTokens(0);
|
||||
usage.setCompletionTokens(0);
|
||||
usage.setTotalTokens(0);
|
||||
usage.setCachedTokens(0);
|
||||
usage.setSuccess(false);
|
||||
usage.setErrorMessage(truncate(errorMessage));
|
||||
usage.setDurationMs(durationMs);
|
||||
saveQuietly(usage);
|
||||
}
|
||||
|
||||
/**
|
||||
* {@inheritDoc}
|
||||
*/
|
||||
@Override
|
||||
public AiUsageResultVO getUsage(Long userId, int recentLimit) {
|
||||
int limit = Math.min(Math.max(recentLimit, 1), MAX_RECENT_LIMIT);
|
||||
AiUsageSummaryVO summary = aiUsageMapper.selectSummary(userId);
|
||||
if (summary == null) {
|
||||
summary = emptySummary();
|
||||
}
|
||||
|
||||
List<AiUsageVO> recent = aiUsageMapper.selectList(new LambdaQueryWrapper<AiUsage>()
|
||||
.eq(AiUsage::getUserId, userId)
|
||||
.orderByDesc(AiUsage::getCreatedAt)
|
||||
.orderByDesc(AiUsage::getId)
|
||||
.last("LIMIT " + limit))
|
||||
.stream().map(this::toVO).collect(Collectors.toList());
|
||||
|
||||
AiUsageResultVO result = new AiUsageResultVO();
|
||||
result.setSummary(summary);
|
||||
result.setRecent(recent);
|
||||
return result;
|
||||
}
|
||||
|
||||
/**
|
||||
* 安静地保存使用记录，失败时仅打印告警。
|
||||
*
|
||||
* @param usage 使用记录实体
|
||||
*/
|
||||
private void saveQuietly(AiUsage usage) {
|
||||
try {
|
||||
aiUsageMapper.insert(usage);
|
||||
} catch (Exception e) {
|
||||
log.warn("记录 AI 使用情况失败: {}", e.getMessage());
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 构造全 0 的汇总对象。
|
||||
*
|
||||
* @return 空的汇总对象
|
||||
*/
|
||||
private AiUsageSummaryVO emptySummary() {
|
||||
AiUsageSummaryVO summary = new AiUsageSummaryVO();
|
||||
summary.setTotalCalls(0L);
|
||||
summary.setSuccessCalls(0L);
|
||||
summary.setFailedCalls(0L);
|
||||
summary.setPromptTokens(0L);
|
||||
summary.setCompletionTokens(0L);
|
||||
summary.setTotalTokens(0L);
|
||||
summary.setCachedTokens(0L);
|
||||
return summary;
|
||||
}
|
||||
|
||||
/**
|
||||
* 将使用记录实体转换为视图对象。
|
||||
*
|
||||
* @param usage 使用记录实体
|
||||
* @return 使用明细视图对象
|
||||
*/
|
||||
private AiUsageVO toVO(AiUsage usage) {
|
||||
AiUsageVO vo = new AiUsageVO();
|
||||
vo.setId(usage.getId());
|
||||
vo.setType(usage.getType());
|
||||
vo.setModel(usage.getModel());
|
||||
vo.setTransactionCount(usage.getTransactionCount());
|
||||
vo.setPromptTokens(usage.getPromptTokens());
|
||||
vo.setCompletionTokens(usage.getCompletionTokens());
|
||||
vo.setTotalTokens(usage.getTotalTokens());
|
||||
vo.setCachedTokens(usage.getCachedTokens());
|
||||
vo.setSuccess(usage.getSuccess());
|
||||
vo.setErrorMessage(usage.getErrorMessage());
|
||||
vo.setDurationMs(usage.getDurationMs());
|
||||
vo.setCreatedAt(usage.getCreatedAt());
|
||||
return vo;
|
||||
}
|
||||
|
||||
/**
|
||||
* 截断过长的失败原因。
|
||||
*
|
||||
* @param value 原始文本
|
||||
* @return 截断后的文本
|
||||
*/
|
||||
private String truncate(String value) {
|
||||
if (value == null) {
|
||||
return null;
|
||||
}
|
||||
return value.length() <= ERROR_MESSAGE_LIMIT ? value : value.substring(0, ERROR_MESSAGE_LIMIT);
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,27 @@
|
||||
package cn.rosercode.finance.vo;
|
||||
|
||||
import lombok.Data;
|
||||
|
||||
/**
|
||||
* AI 配置视图对象，用于向前端返回配置信息（API Key 脱敏）。
|
||||
*
|
||||
* @date 2026-09-17 17:52:00
|
||||
*/
|
||||
@Data
|
||||
public class AiConfigVO {
|
||||
|
||||
/** 是否启用 AI 智能建议 */
|
||||
private Boolean enabled;
|
||||
|
||||
/** 脱敏后的 API Key，未配置时为 null */
|
||||
private String apiKey;
|
||||
|
||||
/** 是否已配置 API Key */
|
||||
private Boolean hasApiKey;
|
||||
|
||||
/** API 基础地址 */
|
||||
private String baseUrl;
|
||||
|
||||
/** 使用的模型名称 */
|
||||
private String model;
|
||||
}
|
||||
@@ -0,0 +1,60 @@
|
||||
package cn.rosercode.finance.vo;
|
||||
|
||||
import lombok.Data;
|
||||
|
||||
import java.math.BigDecimal;
|
||||
import java.util.List;
|
||||
|
||||
/**
|
||||
* AI 自然语言查询解析结果视图对象，字段与交易查询条件对应。
|
||||
*
|
||||
* @date 2026-09-17 18:13:07
|
||||
*/
|
||||
@Data
|
||||
public class AiQueryResultVO {
|
||||
|
||||
/** 解析摘要，用于向用户说明识别到的查询意图 */
|
||||
private String summary;
|
||||
|
||||
/** 收支类型：INCOME / EXPENSE */
|
||||
private List<String> types;
|
||||
|
||||
/** 命中的分类ID列表 */
|
||||
private List<Long> categoryIds;
|
||||
|
||||
/** 命中的分类名称列表 */
|
||||
private List<String> categoryNames;
|
||||
|
||||
/** 是否筛选未分类记录 */
|
||||
private Boolean uncategorized;
|
||||
|
||||
/** 支付方式列表 */
|
||||
private List<String> paymentMethods;
|
||||
|
||||
/** 开始时间（yyyy-MM-ddTHH:mm:ss） */
|
||||
private String startTime;
|
||||
|
||||
/** 结束时间（yyyy-MM-ddTHH:mm:ss） */
|
||||
private String endTime;
|
||||
|
||||
/** 是否计入统计 */
|
||||
private Boolean includeInStats;
|
||||
|
||||
/** 最小金额 */
|
||||
private BigDecimal minAmount;
|
||||
|
||||
/** 最大金额 */
|
||||
private BigDecimal maxAmount;
|
||||
|
||||
/** 备注关键字 */
|
||||
private String remark;
|
||||
|
||||
/** 命中的标签ID列表 */
|
||||
private List<Long> tagIds;
|
||||
|
||||
/** 命中的标签名称列表 */
|
||||
private List<String> tagNames;
|
||||
|
||||
/** 未能识别的关键词，供用户参考 */
|
||||
private List<String> unmatched;
|
||||
}
|
||||
@@ -0,0 +1,51 @@
|
||||
package cn.rosercode.finance.vo;
|
||||
|
||||
import lombok.Data;
|
||||
|
||||
import java.math.BigDecimal;
|
||||
import java.util.List;
|
||||
|
||||
/**
|
||||
* AI 自然语言记账解析结果视图对象，字段与交易记录表单对应。
|
||||
*
|
||||
* @date 2026-09-17 18:23:11
|
||||
*/
|
||||
@Data
|
||||
public class AiRecordResultVO {
|
||||
|
||||
/** 解析摘要，用于向用户说明识别到的记账内容 */
|
||||
private String summary;
|
||||
|
||||
/** 收支类型：INCOME / EXPENSE */
|
||||
private String type;
|
||||
|
||||
/** 金额 */
|
||||
private BigDecimal amount;
|
||||
|
||||
/** 命中的分类ID，未命中为 null */
|
||||
private Long categoryId;
|
||||
|
||||
/** 命中的分类名称 */
|
||||
private String categoryName;
|
||||
|
||||
/** 支付方式 */
|
||||
private String paymentMethod;
|
||||
|
||||
/** 交易时间（yyyy-MM-ddTHH:mm:ss） */
|
||||
private String transactionTime;
|
||||
|
||||
/** 备注 */
|
||||
private String remark;
|
||||
|
||||
/** 是否计入统计 */
|
||||
private Boolean includeInStats;
|
||||
|
||||
/** 命中的标签ID列表 */
|
||||
private List<Long> tagIds;
|
||||
|
||||
/** 命中的标签名称列表 */
|
||||
private List<String> tagNames;
|
||||
|
||||
/** 未能识别的关键词 */
|
||||
private List<String> unmatched;
|
||||
}
|
||||
@@ -0,0 +1,32 @@
|
||||
package cn.rosercode.finance.vo;
|
||||
|
||||
import lombok.Data;
|
||||
|
||||
import java.util.List;
|
||||
|
||||
/**
|
||||
* AI 建议视图对象，描述某条交易记录的分类与标签建议。
|
||||
*
|
||||
* @date 2026-09-17 17:52:01
|
||||
*/
|
||||
@Data
|
||||
public class AiSuggestionVO {
|
||||
|
||||
/** 交易记录ID */
|
||||
private Long transactionId;
|
||||
|
||||
/** 建议分类ID，为 null 表示无分类建议 */
|
||||
private Long categoryId;
|
||||
|
||||
/** 建议分类名称 */
|
||||
private String categoryName;
|
||||
|
||||
/** 建议标签ID列表 */
|
||||
private List<Long> tagIds;
|
||||
|
||||
/** 建议标签名称列表 */
|
||||
private List<String> tagNames;
|
||||
|
||||
/** 建议理由 */
|
||||
private String reason;
|
||||
}
|
||||
@@ -0,0 +1,20 @@
|
||||
package cn.rosercode.finance.vo;
|
||||
|
||||
import lombok.Data;
|
||||
|
||||
import java.util.List;
|
||||
|
||||
/**
|
||||
* AI 使用情况视图对象，包含累计汇总与最近调用明细。
|
||||
*
|
||||
* @date 2026-09-17 18:02:14
|
||||
*/
|
||||
@Data
|
||||
public class AiUsageResultVO {
|
||||
|
||||
/** 累计汇总 */
|
||||
private AiUsageSummaryVO summary;
|
||||
|
||||
/** 最近调用明细（按时间倒序） */
|
||||
private List<AiUsageVO> recent;
|
||||
}
|
||||
@@ -0,0 +1,33 @@
|
||||
package cn.rosercode.finance.vo;
|
||||
|
||||
import lombok.Data;
|
||||
|
||||
/**
|
||||
* AI 使用汇总视图对象，展示累计调用次数与 token 用量。
|
||||
*
|
||||
* @date 2026-09-17 18:02:13
|
||||
*/
|
||||
@Data
|
||||
public class AiUsageSummaryVO {
|
||||
|
||||
/** 累计调用次数 */
|
||||
private Long totalCalls;
|
||||
|
||||
/** 成功次数 */
|
||||
private Long successCalls;
|
||||
|
||||
/** 失败次数 */
|
||||
private Long failedCalls;
|
||||
|
||||
/** 累计输入 token 数 */
|
||||
private Long promptTokens;
|
||||
|
||||
/** 累计输出 token 数 */
|
||||
private Long completionTokens;
|
||||
|
||||
/** 累计总 token 数 */
|
||||
private Long totalTokens;
|
||||
|
||||
/** 累计缓存命中 token 数 */
|
||||
private Long cachedTokens;
|
||||
}
|
||||
@@ -0,0 +1,50 @@
|
||||
package cn.rosercode.finance.vo;
|
||||
|
||||
import lombok.Data;
|
||||
|
||||
import java.time.LocalDateTime;
|
||||
|
||||
/**
|
||||
* AI 使用明细视图对象，描述一次调用的用量与结果。
|
||||
*
|
||||
* @date 2026-09-17 18:02:13
|
||||
*/
|
||||
@Data
|
||||
public class AiUsageVO {
|
||||
|
||||
/** 记录ID */
|
||||
private Long id;
|
||||
|
||||
/** 调用类型：SUGGEST-生成建议，TEST-测试连接 */
|
||||
private String type;
|
||||
|
||||
/** 模型名称 */
|
||||
private String model;
|
||||
|
||||
/** 本次分析的记录数 */
|
||||
private Integer transactionCount;
|
||||
|
||||
/** 输入 token 数 */
|
||||
private Integer promptTokens;
|
||||
|
||||
/** 输出 token 数 */
|
||||
private Integer completionTokens;
|
||||
|
||||
/** 总 token 数 */
|
||||
private Integer totalTokens;
|
||||
|
||||
/** 缓存命中的输入 token 数 */
|
||||
private Integer cachedTokens;
|
||||
|
||||
/** 是否成功 */
|
||||
private Boolean success;
|
||||
|
||||
/** 失败原因 */
|
||||
private String errorMessage;
|
||||
|
||||
/** 调用耗时（毫秒） */
|
||||
private Long durationMs;
|
||||
|
||||
/** 创建时间 */
|
||||
private LocalDateTime createdAt;
|
||||
}
|
||||
@@ -65,3 +65,34 @@ CREATE TABLE IF NOT EXISTS t_transaction_tag (
|
||||
UNIQUE INDEX uk_transaction_tag (transaction_id, tag_id) COMMENT '交易+标签唯一索引',
|
||||
INDEX idx_tag_id (tag_id) COMMENT '标签ID索引'
|
||||
);
|
||||
|
||||
-- AI 配置表：存储每个用户的 AI 智能建议相关配置
|
||||
CREATE TABLE IF NOT EXISTS t_ai_config (
|
||||
id BIGINT AUTO_INCREMENT PRIMARY KEY COMMENT '配置ID，主键自增',
|
||||
user_id BIGINT NOT NULL COMMENT '所属用户ID',
|
||||
enabled TINYINT(1) NOT NULL DEFAULT 0 COMMENT '是否启用AI智能建议：1-启用，0-关闭',
|
||||
api_key VARCHAR(255) COMMENT 'DeepSeek API Key（明文存储）',
|
||||
base_url VARCHAR(255) NOT NULL DEFAULT 'https://api.deepseek.com' COMMENT 'API 基础地址',
|
||||
model VARCHAR(50) NOT NULL DEFAULT 'deepseek-chat' COMMENT '使用的模型名称',
|
||||
created_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP COMMENT '创建时间',
|
||||
updated_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT '更新时间',
|
||||
UNIQUE INDEX uk_user_id (user_id) COMMENT '用户ID唯一索引，每个用户一条配置'
|
||||
);
|
||||
|
||||
-- AI 使用记录表：记录每次调用 DeepSeek 的用量与结果
|
||||
CREATE TABLE IF NOT EXISTS t_ai_usage (
|
||||
id BIGINT AUTO_INCREMENT PRIMARY KEY COMMENT '记录ID，主键自增',
|
||||
user_id BIGINT NOT NULL COMMENT '所属用户ID',
|
||||
type VARCHAR(20) NOT NULL COMMENT '调用类型：SUGGEST-生成建议，TEST-测试连接',
|
||||
model VARCHAR(50) COMMENT '模型名称',
|
||||
transaction_count INT NOT NULL DEFAULT 0 COMMENT '本次分析的记录数',
|
||||
prompt_tokens INT NOT NULL DEFAULT 0 COMMENT '输入 token 数',
|
||||
completion_tokens INT NOT NULL DEFAULT 0 COMMENT '输出 token 数',
|
||||
total_tokens INT NOT NULL DEFAULT 0 COMMENT '总 token 数',
|
||||
cached_tokens INT NOT NULL DEFAULT 0 COMMENT '缓存命中的输入 token 数',
|
||||
success TINYINT(1) NOT NULL DEFAULT 1 COMMENT '是否成功：1-成功，0-失败',
|
||||
error_message VARCHAR(500) COMMENT '失败原因',
|
||||
duration_ms BIGINT COMMENT '调用耗时（毫秒）',
|
||||
created_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP COMMENT '创建时间',
|
||||
INDEX idx_user_created (user_id, created_at) COMMENT '用户ID+创建时间联合索引'
|
||||
);
|
||||
|
||||
Reference in New Issue
Block a user