YogeLiu/doc_ai_backed
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: add mathpixel

Browse Source
This commit is contained in:
yoge
2025-12-20 21:42:58 +08:00
parent aa7fb1c7ca
commit 203c2b64c0
7 changed files with 416 additions and 93 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
config/config.go
View File

@@ -13,6 +13,12 @@ type Config struct {
UploadDir string `mapstructure:"upload_dir"`
Limit LimitConfig `mapstructure:"limit"`
Aliyun AliyunConfig `mapstructure:"aliyun"`
Mathpix MathpixConfig `mapstructure:"mathpix"`
}
type MathpixConfig struct {
AppID string `mapstructure:"app_id"`
AppKey string `mapstructure:"app_key"`
}
type LimitConfig struct {

4
config/config_dev.yaml
View File

@@ -44,3 +44,7 @@ aliyun:
access_key_id: LTAI5t8qXhow6NCdYDtu1saF
access_key_secret: qZ2SwYsNCEBckCVSOszH31yYwXU44A
bucket_name: texpixel-doc
mathpix:
app_id: "576157025515297792"
app_key: "683f7133391a1039383161653531396234343536393263346632613437343332"

5
config/config_prod.yaml
View File

@@ -43,3 +43,8 @@ aliyun:
access_key_id: LTAI5t8qXhow6NCdYDtu1saF
access_key_secret: qZ2SwYsNCEBckCVSOszH31yYwXU44A
bucket_name: texpixel-doc
mathpix:
app_id: "576157025515297792"
app_key: "fb72d251e33ac85c929bfd4eec40d78368d08d82fb2ee1cffb04a8bb967d1db5"

379
internal/service/recognition_service.go
View File

@@ -166,9 +166,20 @@ func (s *RecognitionService) GetFormualTask(ctx context.Context, taskNo string)
log.Error(ctx, "func", "GetFormualTask", "msg", "查询任务结果失败", "error", err, "task_no", taskNo)
return nil, common.NewError(common.CodeDBError, "查询任务结果失败", err)
}
latex := taskRet.NewContentCodec().GetContent().(string)
markdown := fmt.Sprintf("$$%s$$", latex)
return &formula.GetFormulaTaskResponse{TaskNo: taskNo, Latex: latex, Markdown: markdown, Status: int(task.Status)}, nil
// 构建 Markdown 格式
markdown := taskRet.Markdown
if markdown == "" {
markdown = fmt.Sprintf("$$%s$$", taskRet.Latex)
}
return &formula.GetFormulaTaskResponse{
TaskNo: taskNo,
Latex: taskRet.Latex,
Markdown: markdown,
MathML: taskRet.MathML,
Status: int(task.Status),
}, nil
}
func (s *RecognitionService) handleFormulaRecognition(ctx context.Context, taskID int64) error {
@@ -209,6 +220,185 @@ func (s *RecognitionService) processVLFormula(ctx context.Context, taskID int64)
log.Info(ctx, "func", "processVLFormulaQueue", "msg", "处理任务成功", "task_id", taskID)
}
// MathpixRequest Mathpix API /v3/text 完整请求结构
type MathpixRequest struct {
// 图片源：URL 或 base64 编码
Src string `json:"src"`
// 元数据键值对
Metadata map[string]interface{} `json:"metadata"`
// 标签列表，用于标识结果
Tags []string `json:"tags"`
// 异步请求标志
Async bool `json:"async"`
// 回调配置
Callback *MathpixCallback `json:"callback"`
// 输出格式列表：text, data, html, latex_styled
Formats []string `json:"formats"`
// 数据选项
DataOptions *MathpixDataOptions `json:"data_options"`
// 返回检测到的字母表
IncludeDetectedAlphabets bool `json:"include_detected_alphabets"`
// 允许的字母表
AlphabetsAllowed *MathpixAlphabetsAllowed `json:"alphabets_allowed"`
// 指定图片区域
Region *MathpixRegion `json:"region"`
// 蓝色HSV过滤模式
EnableBlueHsvFilter bool `json:"enable_blue_hsv_filter"`
// 置信度阈值
ConfidenceThreshold float64 `json:"confidence_threshold"`
// 符号级别置信度阈值，默认0.75
ConfidenceRateThreshold float64 `json:"confidence_rate_threshold"`
// 包含公式标签
IncludeEquationTags bool `json:"include_equation_tags"`
// 返回逐行信息
IncludeLineData bool `json:"include_line_data"`
// 返回逐词信息
IncludeWordData bool `json:"include_word_data"`
// 化学结构OCR
IncludeSmiles bool `json:"include_smiles"`
// InChI数据
IncludeInchi bool `json:"include_inchi"`
// 几何图形数据
IncludeGeometryData bool `json:"include_geometry_data"`
// 图表文本提取
IncludeDiagramText bool `json:"include_diagram_text"`
// 页面信息，默认true
IncludePageInfo *bool `json:"include_page_info"`
// 自动旋转置信度阈值，默认0.99
AutoRotateConfidenceThreshold float64 `json:"auto_rotate_confidence_threshold"`
// 移除多余空格，默认true
RmSpaces *bool `json:"rm_spaces"`
// 移除字体命令，默认false
RmFonts bool `json:"rm_fonts"`
// 使用aligned/gathered/cases代替array，默认false
IdiomaticEqnArrays bool `json:"idiomatic_eqn_arrays"`
// 移除不必要的大括号，默认false
IdiomaticBraces bool `json:"idiomatic_braces"`
// 数字始终为数学模式，默认false
NumbersDefaultToMath bool `json:"numbers_default_to_math"`
// 数学字体始终为数学模式，默认false
MathFontsDefaultToMath bool `json:"math_fonts_default_to_math"`
// 行内数学分隔符，默认 ["\\(", "\\)"]
MathInlineDelimiters []string `json:"math_inline_delimiters"`
// 行间数学分隔符，默认 ["\\[", "\\]"]
MathDisplayDelimiters []string `json:"math_display_delimiters"`
// 高级表格处理，默认false
EnableTablesFallback bool `json:"enable_tables_fallback"`
// 全角标点，null表示自动判断
FullwidthPunctuation *bool `json:"fullwidth_punctuation"`
}
// MathpixCallback 回调配置
type MathpixCallback struct {
URL string `json:"url"`
Headers map[string]string `json:"headers"`
}
// MathpixDataOptions 数据选项
type MathpixDataOptions struct {
IncludeAsciimath bool `json:"include_asciimath"`
IncludeMathml bool `json:"include_mathml"`
IncludeLatex bool `json:"include_latex"`
IncludeTsv bool `json:"include_tsv"`
}
// MathpixAlphabetsAllowed 允许的字母表
type MathpixAlphabetsAllowed struct {
En bool `json:"en"`
Hi bool `json:"hi"`
Zh bool `json:"zh"`
Ja bool `json:"ja"`
Ko bool `json:"ko"`
Ru bool `json:"ru"`
Th bool `json:"th"`
Vi bool `json:"vi"`
}
// MathpixRegion 图片区域
type MathpixRegion struct {
TopLeftX int `json:"top_left_x"`
TopLeftY int `json:"top_left_y"`
Width int `json:"width"`
Height int `json:"height"`
}
// MathpixResponse Mathpix API /v3/text 完整响应结构
type MathpixResponse struct {
// 请求ID，用于调试
RequestID string `json:"request_id"`
// Mathpix Markdown 格式文本
Text string `json:"text"`
// 带样式的LaTeX（仅单个公式图片时返回）
LatexStyled string `json:"latex_styled"`
// 置信度 [0,1]
Confidence float64 `json:"confidence"`
// 置信度比率 [0,1]
ConfidenceRate float64 `json:"confidence_rate"`
// 行数据
LineData []map[string]interface{} `json:"line_data"`
// 词数据
WordData []map[string]interface{} `json:"word_data"`
// 数据对象列表
Data []MathpixDataItem `json:"data"`
// HTML输出
HTML string `json:"html"`
// 检测到的字母表
DetectedAlphabets []map[string]interface{} `json:"detected_alphabets"`
// 是否打印内容
IsPrinted bool `json:"is_printed"`
// 是否手写内容
IsHandwritten bool `json:"is_handwritten"`
// 自动旋转置信度
AutoRotateConfidence float64 `json:"auto_rotate_confidence"`
// 几何数据
GeometryData []map[string]interface{} `json:"geometry_data"`
// 自动旋转角度 {0, 90, -90, 180}
AutoRotateDegrees int `json:"auto_rotate_degrees"`
// 图片宽度
ImageWidth int `json:"image_width"`
// 图片高度
ImageHeight int `json:"image_height"`
// 错误信息
Error string `json:"error"`
// 错误详情
ErrorInfo *MathpixErrorInfo `json:"error_info"`
// API版本
Version string `json:"version"`
}
// MathpixDataItem 数据项
type MathpixDataItem struct {
Type string `json:"type"`
Value string `json:"value"`
}
// MathpixErrorInfo 错误详情
type MathpixErrorInfo struct {
ID string `json:"id"`
Message string `json:"message"`
}
// GetMathML 从响应中获取MathML
func (r *MathpixResponse) GetMathML() string {
for _, item := range r.Data {
if item.Type == "mathml" {
return item.Value
}
}
return ""
}
// GetAsciiMath 从响应中获取AsciiMath
func (r *MathpixResponse) GetAsciiMath() string {
for _, item := range r.Data {
if item.Type == "asciimath" {
return item.Value
}
}
return ""
}
func (s *RecognitionService) processFormulaTask(ctx context.Context, taskID int64, fileURL string) (err error) {
// 为整个任务处理添加超时控制
ctx, cancel := context.WithTimeout(ctx, 45*time.Second)
@@ -308,20 +498,12 @@ func (s *RecognitionService) processFormulaTask(ctx context.Context, taskID int6
log.Error(ctx, "func", "processFormulaTask", "msg", "解析响应JSON失败", "error", err)
return err
}
// katex := utils.ToKatex(formulaResp.Result)
katex := formulaResp.Result
content := &dao.FormulaRecognitionContent{Latex: katex}
b, _ := json.Marshal(content)
// Save recognition result
result := &dao.RecognitionResult{
TaskID: taskID,
TaskType: dao.TaskTypeFormula,
Content: b,
}
if err := resultDao.Create(tx, *result); err != nil {
err = resultDao.Create(tx, dao.RecognitionResult{TaskID: taskID, TaskType: dao.TaskTypeFormula, Latex: formulaResp.Result})
if err != nil {
log.Error(ctx, "func", "processFormulaTask", "msg", "保存任务结果失败", "error", err)
return err
}
isSuccess = true
return nil
}
@@ -426,39 +608,22 @@ func (s *RecognitionService) processVLFormulaTask(ctx context.Context, taskID in
}
resultDao := dao.NewRecognitionResultDao()
var formulaRes *dao.FormulaRecognitionContent
var formulaRes *dao.RecognitionResult
result, err := resultDao.GetByTaskID(dao.DB.WithContext(ctx), taskID)
if err != nil {
log.Error(ctx, "func", "processVLFormulaTask", "msg", "获取任务结果失败", "error", err)
return err
}
if result == nil {
formulaRes = &dao.FormulaRecognitionContent{EnhanceLatex: latex}
b, err := formulaRes.Encode()
if err != nil {
log.Error(ctx, "func", "processVLFormulaTask", "msg", "编码任务结果失败", "error", err)
return err
}
err = resultDao.Create(dao.DB.WithContext(ctx), dao.RecognitionResult{TaskID: taskID, TaskType: dao.TaskTypeFormula, Content: b})
formulaRes = &dao.RecognitionResult{TaskID: taskID, TaskType: dao.TaskTypeFormula, Latex: latex}
err = resultDao.Create(dao.DB.WithContext(ctx), *formulaRes)
if err != nil {
log.Error(ctx, "func", "processVLFormulaTask", "msg", "创建任务结果失败", "error", err)
return err
}
} else {
formulaRes = result.NewContentCodec().(*dao.FormulaRecognitionContent)
err = formulaRes.Decode()
if err != nil {
log.Error(ctx, "func", "processVLFormulaTask", "msg", "解码任务结果失败", "error", err)
return err
}
formulaRes.EnhanceLatex = latex
b, err := formulaRes.Encode()
if err != nil {
log.Error(ctx, "func", "processVLFormulaTask", "msg", "编码任务结果失败", "error", err)
return err
}
err = resultDao.Update(dao.DB.WithContext(ctx), result.ID, map[string]interface{}{"content": b})
formulaRes.Latex = latex
err = resultDao.Update(dao.DB.WithContext(ctx), result.ID, map[string]interface{}{"latex": latex})
if err != nil {
log.Error(ctx, "func", "processVLFormulaTask", "msg", "更新任务结果失败", "error", err)
return err
@@ -504,8 +669,7 @@ func (s *RecognitionService) processOneTask(ctx context.Context) {
ctx = context.WithValue(ctx, utils.RequestIDKey, task.TaskUUID)
log.Info(ctx, "func", "processFormulaQueue", "msg", "获取任务成功", "task_id", taskID)
// 处理任务
err = s.processFormulaTask(ctx, taskID, task.FileURL)
err = s.processMathpixTask(ctx, taskID, task.FileURL)
if err != nil {
log.Error(ctx, "func", "processFormulaQueue", "msg", "处理任务失败", "error", err)
return
@@ -513,3 +677,142 @@ func (s *RecognitionService) processOneTask(ctx context.Context) {
log.Info(ctx, "func", "processFormulaQueue", "msg", "处理任务成功", "task_id", taskID)
}
// processMathpixTask 使用 Mathpix API 处理公式识别任务（用于增强识别）
func (s *RecognitionService) processMathpixTask(ctx context.Context, taskID int64, fileURL string) error {
isSuccess := false
logDao := dao.NewRecognitionLogDao()
defer func() {
if !isSuccess {
err := dao.NewRecognitionTaskDao().Update(dao.DB.WithContext(ctx), map[string]interface{}{"id": taskID}, map[string]interface{}{"status": dao.TaskStatusFailed})
if err != nil {
log.Error(ctx, "func", "processMathpixTask", "msg", "更新任务状态失败", "error", err)
}
return
}
err := dao.NewRecognitionTaskDao().Update(dao.DB.WithContext(ctx), map[string]interface{}{"id": taskID}, map[string]interface{}{"status": dao.TaskStatusCompleted})
if err != nil {
log.Error(ctx, "func", "processMathpixTask", "msg", "更新任务状态失败", "error", err)
}
}()
// 下载图片
imageUrl, err := oss.GetDownloadURL(ctx, fileURL)
if err != nil {
log.Error(ctx, "func", "processMathpixTask", "msg", "获取图片URL失败", "error", err)
return err
}
// 创建 Mathpix API 请求
mathpixReq := MathpixRequest{
Src: imageUrl,
Formats: []string{
"text",
"latex_styled",
"data",
"html",
},
DataOptions: &MathpixDataOptions{
IncludeMathml: true,
IncludeAsciimath: true,
IncludeLatex: true,
IncludeTsv: true,
},
}
jsonData, err := json.Marshal(mathpixReq)
if err != nil {
log.Error(ctx, "func", "processMathpixTask", "msg", "JSON编码失败", "error", err)
return err
}
headers := map[string]string{
"Content-Type": "application/json",
"app_id": config.GlobalConfig.Mathpix.AppID,
"app_key": config.GlobalConfig.Mathpix.AppKey,
}
endpoint := "https://api.mathpix.com/v3/text"
resp, err := s.httpClient.RequestWithRetry(ctx, http.MethodPost, endpoint, bytes.NewReader(jsonData), headers)
if err != nil {
log.Error(ctx, "func", "processMathpixTask", "msg", "Mathpix API 请求失败", "error", err)
return err
}
defer resp.Body.Close()
body := &bytes.Buffer{}
if _, err = body.ReadFrom(resp.Body); err != nil {
log.Error(ctx, "func", "processMathpixTask", "msg", "读取响应体失败", "error", err)
return err
}
// 创建日志记录
recognitionLog := &dao.RecognitionLog{
TaskID: taskID,
Provider: dao.ProviderMathpix,
RequestBody: string(jsonData),
ResponseBody: body.String(),
}
// 解析响应
var mathpixResp MathpixResponse
if err := json.Unmarshal(body.Bytes(), &mathpixResp); err != nil {
log.Error(ctx, "func", "processMathpixTask", "msg", "解析响应失败", "error", err)
return err
}
// 检查错误
if mathpixResp.Error != "" {
errMsg := mathpixResp.Error
if mathpixResp.ErrorInfo != nil {
errMsg = fmt.Sprintf("%s: %s", mathpixResp.ErrorInfo.ID, mathpixResp.ErrorInfo.Message)
}
log.Error(ctx, "func", "processMathpixTask", "msg", "Mathpix API 返回错误", "error", errMsg)
return fmt.Errorf("mathpix error: %s", errMsg)
}
// 保存日志
err = logDao.Create(dao.DB.WithContext(ctx), recognitionLog)
if err != nil {
log.Error(ctx, "func", "processMathpixTask", "msg", "保存日志失败", "error", err)
}
// 更新或创建识别结果
resultDao := dao.NewRecognitionResultDao()
result, err := resultDao.GetByTaskID(dao.DB.WithContext(ctx), taskID)
if err != nil {
log.Error(ctx, "func", "processMathpixTask", "msg", "获取任务结果失败", "error", err)
return err
}
if result == nil {
// 创建新结果
err = resultDao.Create(dao.DB.WithContext(ctx), dao.RecognitionResult{
TaskID: taskID,
TaskType: dao.TaskTypeFormula,
Latex: mathpixResp.LatexStyled,
Markdown: mathpixResp.Text,
MathML: mathpixResp.GetMathML(),
})
if err != nil {
log.Error(ctx, "func", "processMathpixTask", "msg", "创建任务结果失败", "error", err)
return err
}
} else {
// 更新现有结果
err = resultDao.Update(dao.DB.WithContext(ctx), result.ID, map[string]interface{}{
"latex": mathpixResp.LatexStyled,
"markdown": mathpixResp.Text,
"mathml": mathpixResp.GetMathML(),
})
if err != nil {
log.Error(ctx, "func", "processMathpixTask", "msg", "更新任务结果失败", "error", err)
return err
}
}
isSuccess = true
return nil
}

5
internal/service/task.go
View File

@@ -89,9 +89,12 @@ func (svc *TaskService) GetTaskList(ctx context.Context, req *task.TaskListReque
var markdown string
recognitionResult := recognitionResultMap[item.ID]
if recognitionResult != nil {
latex = recognitionResult.NewContentCodec().GetContent().(string)
latex = recognitionResult.Latex
markdown = recognitionResult.Markdown
if markdown == "" {
markdown = fmt.Sprintf("$$%s$$", latex)
}
}
originURL, err := oss.GetDownloadURL(ctx, item.FileURL)
if err != nil {
log.Error(ctx, "func", "GetTaskList", "msg", "get origin url failed", "error", err)

52
internal/storage/dao/recognition_log.go Normal file
View File

@@ -0,0 +1,52 @@
package dao
import (
"gorm.io/gorm"
)
// RecognitionLogProvider 第三方服务提供商
type RecognitionLogProvider string
const (
ProviderMathpix RecognitionLogProvider = "mathpix"
ProviderSiliconflow RecognitionLogProvider = "siliconflow"
ProviderTexpixel RecognitionLogProvider = "texpixel"
)
// RecognitionLog 识别调用日志表，记录第三方API调用请求和响应
type RecognitionLog struct {
BaseModel
TaskID int64 `gorm:"column:task_id;bigint;not null;default:0;index;comment:关联任务ID" json:"task_id"`
Provider RecognitionLogProvider `gorm:"column:provider;varchar(32);not null;comment:服务提供商" json:"provider"`
RequestBody string `gorm:"column:request_body;type:longtext;comment:请求体" json:"request_body"`
ResponseBody string `gorm:"column:response_body;type:longtext;comment:响应体" json:"response_body"`
}
func (RecognitionLog) TableName() string {
return "recognition_log"
}
type RecognitionLogDao struct{}
func NewRecognitionLogDao() *RecognitionLogDao {
return &RecognitionLogDao{}
}
// Create 创建日志记录
func (d *RecognitionLogDao) Create(tx *gorm.DB, log *RecognitionLog) error {
return tx.Create(log).Error
}
// GetByTaskID 根据任务ID获取日志
func (d *RecognitionLogDao) GetByTaskID(tx *gorm.DB, taskID int64) ([]*RecognitionLog, error) {
var logs []*RecognitionLog
err := tx.Where("task_id = ?", taskID).Order("created_at DESC").Find(&logs).Error
return logs, err
}
// GetByProvider 根据提供商获取日志
func (d *RecognitionLogDao) GetByProvider(tx *gorm.DB, provider RecognitionLogProvider, limit int) ([]*RecognitionLog, error) {
var logs []*RecognitionLog
err := tx.Where("provider = ?", provider).Order("created_at DESC").Limit(limit).Find(&logs).Error
return logs, err
}

56
internal/storage/dao/result.go
View File

@@ -1,66 +1,16 @@
package dao
import (
"encoding/json"
"gorm.io/gorm"
)
type JSON []byte
// ContentCodec 定义内容编解码接口
type ContentCodec interface {
Encode() (JSON, error)
Decode() error
GetContent() interface{} // 更明确的方法名
}
type FormulaRecognitionContent struct {
content JSON
Latex string `json:"latex"`
AdjustLatex string `json:"adjust_latex"`
EnhanceLatex string `json:"enhance_latex"`
}
func (c *FormulaRecognitionContent) Encode() (JSON, error) {
b, err := json.Marshal(c)
if err != nil {
return nil, err
}
return b, nil
}
func (c *FormulaRecognitionContent) Decode() error {
return json.Unmarshal(c.content, c)
}
// GetPreferredContent 按优先级返回公式内容
func (c *FormulaRecognitionContent) GetContent() interface{} {
c.Decode()
if c.EnhanceLatex != "" {
return c.EnhanceLatex
} else if c.AdjustLatex != "" {
return c.AdjustLatex
} else {
return c.Latex
}
}
type RecognitionResult struct {
BaseModel
TaskID int64 `gorm:"column:task_id;bigint;not null;default:0;comment:任务ID" json:"task_id"`
TaskType TaskType `gorm:"column:task_type;varchar(16);not null;comment:任务类型;default:''" json:"task_type"`
Content JSON `gorm:"column:content;type:json;not null;comment:识别内容" json:"content"`
}
// NewContentCodec 创建对应任务类型的内容编解码器
func (r *RecognitionResult) NewContentCodec() ContentCodec {
switch r.TaskType {
case TaskTypeFormula:
return &FormulaRecognitionContent{content: r.Content}
default:
return nil
}
Latex string `json:"latex"`
Markdown string `json:"markdown"` // Mathpix Markdown 格式
MathML string `json:"mathml"` // MathML 格式
}
type RecognitionResultDao struct {