support: image ocr.

client/builders.go

@@ -1146,6 +1146,32 @@ func (c *QQClient) buildGroupFileDownloadReqPacket(groupCode int64, fileId strin
 	return seq, packet
 }
 
+// OidbSvc.0xe07_0
+func (c *QQClient) buildImageOcrRequestPacket(url, md5 string, size, weight, height int32) (uint16, []byte) {
+	seq := c.nextSeq()
+	body := &oidb.DE07ReqBody{
+		Version:  1,
+		Entrance: 3,
+		OcrReqBody: &oidb.OCRReqBody{
+			ImageUrl:              url,
+			OriginMd5:             md5,
+			AfterCompressMd5:      md5,
+			AfterCompressFileSize: size,
+			AfterCompressWeight:   weight,
+			AfterCompressHeight:   height,
+			IsCut:                 false,
+		},
+	}
+	b, _ := proto.Marshal(body)
+	req := &oidb.OIDBSSOPkg{
+		Command:    3591,
+		Bodybuffer: b,
+	}
+	payload, _ := proto.Marshal(req)
+	packet := packets.BuildUniPacket(c.Uin, seq, "OidbSvc.0xe07_0", 1, c.OutGoingPacketSessionId, EmptyBytes, c.sigInfo.d2Key, payload)
+	return seq, packet
+}
+
 // PttCenterSvr.ShortVideoDownReq
 func (c *QQClient) buildPttShortVideoDownReqPacket(uuid, md5 []byte) (uint16, []byte) {
 	seq := c.nextSeq()

client/client.go

@@ -142,6 +142,7 @@ func NewClientMd5(uin int64, passwordMd5 [16]byte) *QQClient {
 			"MultiMsg.ApplyDown":                           decodeMultiApplyDownResponse,    // 长消息/合并转发请求下载包
 			"OidbSvc.0x6d6_2":                              decodeOIDB6d6Response,           // 群文件操作包
 			"OidbSvc.0x88d_0":                              decodeGroupInfoResponse,         // 获取群资料包
+			"OidbSvc.0xe07_0":                              decodeImageOcrResponse,          // 图片OCR请求包
 			"SummaryCard.ReqSummaryCard":                   decodeSummaryCardResponse,       // 获取用户卡片资料包
 			"PttCenterSvr.ShortVideoDownReq":               decodePttShortVideoDownResponse, // 短视频下载请求包
 			"LightAppSvc.mini_app_info.GetAppInfoById":     decodeAppInfoResponse,           // 获取小程序资料包
@@ -624,6 +625,24 @@ func (c *QQClient) uploadPrivateImage(target int64, img []byte, count int) (*mes
 	return e, nil
 }
 
+func (c *QQClient) ImageOcr(img interface{}) (*OcrResponse, error) {
+	switch e := img.(type) {
+	case *message.GroupImageElement:
+		rsp, err := c.sendAndWait(c.buildImageOcrRequestPacket(e.Url, strings.ToUpper(hex.EncodeToString(e.Md5)), e.Size, e.Width, e.Height))
+		if err != nil {
+			return nil, err
+		}
+		return rsp.(*OcrResponse), nil
+	case *message.ImageElement:
+		rsp, err := c.sendAndWait(c.buildImageOcrRequestPacket(e.Url, strings.ToUpper(hex.EncodeToString(e.Md5)), e.Size, e.Width, e.Height))
+		if err != nil {
+			return nil, err
+		}
+		return rsp.(*OcrResponse), nil
+	}
+	return nil, errors.New("image error")
+}
+
 func (c *QQClient) UploadGroupPtt(groupCode int64, voice []byte) (*message.GroupVoiceElement, error) {
 	h := md5.Sum(voice)
 	seq, pkt := c.buildGroupPttStorePacket(groupCode, h[:], int32(len(voice)), 0, int32(len(voice)))

client/decoders.go

@@ -1009,6 +1009,39 @@ func decodeOIDB6d6Response(c *QQClient, _ uint16, payload []byte) (interface{},
 	return fmt.Sprintf("http://%s/ftn_handler/%s/", ip, url), nil
 }
 
+func decodeImageOcrResponse(_ *QQClient, _ uint16, payload []byte) (interface{}, error) {
+	pkg := oidb.OIDBSSOPkg{}
+	rsp := oidb.DE07RspBody{}
+	if err := proto.Unmarshal(payload, &pkg); err != nil {
+		return nil, err
+	}
+	if err := proto.Unmarshal(pkg.Bodybuffer, &rsp); err != nil {
+		return nil, err
+	}
+	if rsp.Wording != "" {
+		return nil, errors.New(rsp.Wording)
+	}
+	var texts []*TextDetection
+	for _, text := range rsp.OcrRspBody.TextDetections {
+		var points []*Coordinate
+		for _, c := range text.Polygon.Coordinates {
+			points = append(points, &Coordinate{
+				X: c.X,
+				Y: c.Y,
+			})
+		}
+		texts = append(texts, &TextDetection{
+			Text:        text.DetectedText,
+			Confidence:  text.Confidence,
+			Coordinates: points,
+		})
+	}
+	return &OcrResponse{
+		Texts:    texts,
+		Language: rsp.OcrRspBody.Language,
+	}, nil
+}
+
 func decodePttShortVideoDownResponse(c *QQClient, _ uint16, payload []byte) (interface{}, error) {
 	rsp := pttcenter.ShortVideoRspBody{}
 	if err := proto.Unmarshal(payload, &rsp); err != nil {

client/entities.go

@@ -183,6 +183,22 @@ type (
 		Friend *FriendInfo
 	}
 
+	OcrResponse struct {
+		Texts    []*TextDetection `json:"'texts'"`
+		Language string           `json:"language"`
+	}
+
+	TextDetection struct {
+		Text        string        `json:"text"`
+		Confidence  int32         `json:"confidence"`
+		Coordinates []*Coordinate `json:"coordinates"`
+	}
+
+	Coordinate struct {
+		X int32 `json:"x"`
+		Y int32 `json:"y"`
+	}
+
 	groupMemberListResponse struct {
 		NextUin int64
 		list    []*GroupMemberInfo