update

2026-05-20 03:25:35 +08:00
parent 5e3dac3020
commit 7f5350db62
30 changed files with 3370 additions and 1752 deletions
--- a/internal/api/classifier.go
+++ b/internal/api/classifier.go
@@ -0,0 +1,309 @@
+package api
+
+import (
+	"encoding/json"
+	"fmt"
+	"io"
+	"net/http"
+	"net/url"
+	"strings"
+	"time"
+)
+
+// CompanyIndustryClassifier 企业行业分类器
+type CompanyIndustryClassifier struct {
+	httpClient *http.Client
+	llmClient  *LLMClient
+}
+
+// CompanyData 企业数据
+type CompanyData struct {
+	Description   string   `json:"description"`
+	Business      string   `json:"business"`
+	BiddingTitles []string `json:"bidding_titles"`
+}
+
+// IndustryClassification 行业分类结果
+type IndustryClassification struct {
+	CompanyName string      `json:"company_name"`
+	Primary     PrimaryInfo `json:"primary"`
+	Secondary   []Secondary `json:"secondary"`
+}
+
+// PrimaryInfo 主要行业信息
+type PrimaryInfo struct {
+	Code   string `json:"code"`
+	Name   string `json:"name"`
+	Reason string `json:"reason"`
+}
+
+// Secondary 次要行业信息
+type Secondary struct {
+	Code   string `json:"code"`
+	Name   string `json:"name"`
+	Reason string `json:"reason"`
+}
+
+// NewCompanyIndustryClassifier 创建企业行业分类器
+func NewCompanyIndustryClassifier(llmClient *LLMClient) *CompanyIndustryClassifier {
+	client := &http.Client{
+		Timeout: 30 * time.Second,
+	}
+
+	return &CompanyIndustryClassifier{
+		httpClient: client,
+		llmClient:  llmClient,
+	}
+}
+
+// getDescriptionText 获取简介文本
+func getDescriptionText(description string) string {
+	if description == "" {
+		return "(无简介)"
+	}
+	return description
+}
+
+// getBusinessText 获取经营范围文本
+func getBusinessText(business string) string {
+	if business == "" {
+		return "(无经营范围)"
+	}
+	return business
+}
+
+// FetchCompanyData 从cmdp API获取企业工商和招投标数据
+func (c *CompanyIndustryClassifier) FetchCompanyData(companyName string) (map[string]interface{}, error) {
+	apiURL := "https://a.cmdp.cn/cmds/v1/company/info-rest/custom"
+
+	params := url.Values{}
+	params.Set("name", companyName)
+	params.Set("api", "p_info,p_biddings_all")
+
+	fullURL := apiURL + "?" + params.Encode()
+
+	req, err := http.NewRequest("GET", fullURL, nil)
+	if err != nil {
+		return nil, fmt.Errorf("创建请求失败: %v", err)
+	}
+
+	resp, err := c.httpClient.Do(req)
+	if err != nil {
+		return nil, fmt.Errorf("发送请求失败: %v", err)
+	}
+	defer resp.Body.Close()
+
+	body, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return nil, fmt.Errorf("读取响应失败: %v", err)
+	}
+
+	if resp.StatusCode != http.StatusOK {
+		return nil, fmt.Errorf("API请求失败，状态码: %d, 响应: %s", resp.StatusCode, string(body))
+	}
+
+	var result map[string]interface{}
+	err = json.Unmarshal(body, &result)
+	if err != nil {
+		return nil, fmt.Errorf("解析响应JSON失败: %v", err)
+	}
+
+	return result, nil
+}
+
+// ExtractFields 从API响应中提取关键字段
+func (c *CompanyIndustryClassifier) ExtractFields(data map[string]interface{}) (string, string, []string) {
+	description := ""
+	business := ""
+	var biddingTitles []string
+
+	// 兼容不同的响应结构
+	payload := data
+	if data["data"] != nil {
+		if payloadData, ok := data["data"].(map[string]interface{}); ok {
+			payload = payloadData
+		}
+	}
+
+	// p_info -> description, business
+	if pInfo := payload["p_info"]; pInfo != nil {
+		var pInfoObj map[string]interface{}
+
+		switch v := pInfo.(type) {
+		case []interface{}:
+			if len(v) > 0 {
+				if item, ok := v[0].(map[string]interface{}); ok {
+					pInfoObj = item
+				}
+			}
+		case map[string]interface{}:
+			pInfoObj = v
+		}
+
+		if pInfoObj != nil {
+			if desc, ok := pInfoObj["description"].(string); ok {
+				description = desc
+			}
+			if bus, ok := pInfoObj["business"].(string); ok {
+				business = bus
+			}
+		}
+	}
+
+	// p_biddings_all -> title list
+	if biddingsRaw := payload["p_biddings_all"]; biddingsRaw != nil {
+		var biddingsList []interface{}
+
+		switch v := biddingsRaw.(type) {
+		case []interface{}:
+			biddingsList = v
+		case map[string]interface{}:
+			// 尝试从data或list字段获取
+			if dataField, ok := v["data"].([]interface{}); ok {
+				biddingsList = dataField
+			} else if listField, ok := v["list"].([]interface{}); ok {
+				biddingsList = listField
+			}
+		}
+
+		for _, item := range biddingsList {
+			if itemMap, ok := item.(map[string]interface{}); ok {
+				if title, ok := itemMap["title"].(string); ok && strings.TrimSpace(title) != "" {
+					biddingTitles = append(biddingTitles, strings.TrimSpace(title))
+				}
+			}
+		}
+	}
+
+	return description, business, biddingTitles
+}
+
+// BuildPrompt 构建提示词
+func (c *CompanyIndustryClassifier) BuildPrompt(companyName, description, business string, biddingTitles []string) string {
+	titlesText := "（无招投标记录）"
+	if len(biddingTitles) > 0 {
+		var sb strings.Builder
+		for i, title := range biddingTitles {
+			sb.WriteString(fmt.Sprintf("%d. %s\n", i+1, title))
+		}
+		titlesText = sb.String()
+	}
+
+	prompt := fmt.Sprintf(`你是国民经济行业分类专家，请根据以下企业信息，依据《国民经济行业分类》（GB/T 4754-2017）进行行业大类判断。
+
+## 企业名称
+%s
+
+## 企业简介（description）
+%s
+
+## 经营范围（business）
+%s
+
+## 招投标项目标题（bidding titles）
+%s
+
+## 任务要求
+1. 判断该企业**主要**所属的行业大类（两位数字编号 + 名称）
+2. 如有次要业务涉及其他大类，也一并列出（最多2个）
+3. 给出简要判断依据（不超过100字）
+
+## 输出格式（严格 JSON，不要 Markdown 代码块）
+{
+  "company_name": "%s",
+  "primary": {
+    "code": "65",
+    "name": "软件和信息技术服务业",
+    "reason": "判断依据..."
+  },
+  "secondary": [
+    {
+      "code": "72",
+      "name": "商务服务业",
+      "reason": "轻微涉及原因..."
+    }
+  ]
+}
+
+注意：secondary 可以为空列表 []；只输出 JSON，不要任何额外说明。`,
+		companyName,
+		getDescriptionText(description),
+		getBusinessText(business),
+		titlesText,
+		companyName)
+
+	return prompt
+}
+
+// ClassifyCompany 完整流程：获取数据 → 提取字段 → 构建提示词 → 调用 LLM → 返回 JSON
+func (c *CompanyIndustryClassifier) ClassifyCompany(companyName string) (*IndustryClassification, error) {
+	// 1. 获取企业数据
+	rawData, err := c.FetchCompanyData(companyName)
+	if err != nil {
+		return nil, fmt.Errorf("获取企业数据失败: %v", err)
+	}
+
+	// 2. 提取关键字段
+	description, business, biddingTitles := c.ExtractFields(rawData)
+
+	// 3. 构建提示词
+	prompt := c.BuildPrompt(companyName, description, business, biddingTitles)
+	// 4. 调用LLM并解析JSON
+	response, err := c.llmClient.Chat([]ChatMessage{
+		{Role: "user", Content: prompt},
+	})
+	if err != nil {
+		return nil, fmt.Errorf("调用LLM失败: %v", err)
+	}
+
+	// 提取JSON部分（如果LLM返回了markdown代码块）
+	jsonStr := extractJSONFromResponse(response.Choices[0].Message.Content)
+
+	var result IndustryClassification
+	err = json.Unmarshal([]byte(jsonStr), &result)
+	if err != nil {
+		return nil, fmt.Errorf("解析LLM返回的JSON失败: %v", err)
+	}
+
+	return &result, nil
+}
+
+// extractJSONFromResponse 从LLM响应中提取JSON内容
+// 查找并提取 ```json {...} ``` 或 ``` {...} ``` 中的内容
+func extractJSONFromResponse(response string) string {
+	// 首先尝试匹配 ```json 开头的代码块
+	start := strings.Index(response, "```json")
+	if start != -1 {
+		start += 7 // 跳过 "```json"
+		end := strings.Index(response[start:], "```")
+		if end != -1 {
+			jsonContent := strings.TrimSpace(response[start : start+end])
+			return jsonContent
+		}
+	}
+
+	// 然后尝试匹配 ``` 开头的代码块（假设是JSON）
+	start = strings.Index(response, "```")
+	if start != -1 {
+		start += 3 // 跳过 "```"
+		end := strings.Index(response[start:], "```")
+		if end != -1 {
+			jsonContent := strings.TrimSpace(response[start : start+end])
+			// 检查是否看起来像JSON
+			jsonContent = strings.TrimSpace(jsonContent)
+			if strings.HasPrefix(jsonContent, "{") || strings.HasPrefix(jsonContent, "[") {
+				return jsonContent
+			}
+		}
+	}
+
+	// 如果没有找到代码块，尝试直接解析整个响应是否为JSON
+	response = strings.TrimSpace(response)
+	if strings.HasPrefix(response, "{") && strings.HasSuffix(response, "}") ||
+		strings.HasPrefix(response, "[") && strings.HasSuffix(response, "]") {
+		return response
+	}
+
+	// 如果以上都失败，返回原内容
+	return response
+}