paddle ocr suport

backend/internal/ocr/client.go

@@ -8,6 +8,8 @@ import (
 	"io"
 	"net/http"
 	"time"
+
+	"github.com/ledongthuc/pdf"
 )
 
 type Client struct {
@@ -25,6 +27,18 @@ func NewClient(endpoint string) *Client {
 }
 
 func (c *Client) Recognize(imageData []byte) ([]TextBlock, error) {
+
+	docType := DetectDocumentType(imageData)
+
+	switch docType {
+	case PDFScanned:
+		fmt.Println("PDFScanned -> PaddleOCR")
+	case Image:
+		fmt.Println("Image -> PaddleOCR")
+	default:
+		return []TextBlock{}, fmt.Errorf("DOCTYPE not supported => %s", docType)
+	}
+
 	b64 := base64.StdEncoding.EncodeToString(imageData)
 
 	reqBody, err := json.Marshal(OCRRequest{Image: b64})
@@ -63,6 +77,59 @@ func (c *Client) Recognize(imageData []byte) ([]TextBlock, error) {
 	return flattenResults(ocrResp.Result), nil
 }
 
+type DocType string
+
+const (
+	Image      DocType = "image"
+	PDFText    DocType = "pdf_text"
+	PDFScanned DocType = "pdf_scanned"
+	Unknown    DocType = "unknown"
+)
+
+func isPDFText(data []byte) bool {
+	reader := bytes.NewReader(data)
+
+	r, err := pdf.NewReader(reader, int64(len(data)))
+	if err != nil {
+		return false
+	}
+
+	for i := 1; i <= r.NumPage(); i++ {
+		page := r.Page(i)
+		if page.V.IsNull() {
+			continue
+		}
+
+		text, _ := page.GetPlainText(nil)
+		if len(text) > 20 {
+			return true
+		}
+	}
+
+	return false
+}
+
+func DetectDocumentType(data []byte) DocType {
+	mime := http.DetectContentType(data)
+
+	switch {
+	case mime == "application/pdf":
+		if isPDFText(data) {
+			return PDFText
+		}
+		return PDFScanned
+
+	case bytes.HasPrefix(data, []byte{0xFF, 0xD8}): // JPEG
+		return Image
+
+	case bytes.HasPrefix(data, []byte{0x89, 0x50, 0x4E, 0x47}): // PNG
+		return Image
+
+	default:
+		return Unknown
+	}
+}
+
 func (c *Client) HealthCheck() error {
 	resp, err := c.httpClient.Get(c.endpoint + "/health")
 	if err != nil {

backend/internal/service/presignedurl.go

@@ -22,7 +22,7 @@ func GenerateFileDownloadUrl(fileID string) string {
 	sig := sign(fileID, expires, secret)
 
 	url := fmt.Sprintf(
-		"http://192.168.1.142:8080/api/v1/files/%s?expires=%d&sig=%s",
+		"http://192.168.1.17:8080/api/v1/files/%s?expires=%d&sig=%s",
 		fileID,
 		expires,
 		sig,