feat(document): repair pdf with libreoffic

operator/document/v0/convert_to_images.go

@@ -45,8 +45,18 @@ func ConvertDocumentToImage(inputStruct *ConvertDocumentToImagesInput) (*Convert
 		base64PDF = strings.Split(inputStruct.Document, ",")[1]
 	}
 
+	var base64PDFWithoutMime string
+	if RequiredToRepair(base64PDF) {
+		base64PDFWithoutMime, err = RepairPDF(base64PDF)
+		if err != nil {
+			return nil, fmt.Errorf("failed to repair PDF: %w", err)
+		}
+	} else {
+		base64PDFWithoutMime = base.TrimBase64Mime(base64PDF)
+	}
+
 	paramsJSON := map[string]interface{}{
-		"PDF":      base.TrimBase64Mime(base64PDF),
+		"PDF":      base64PDFWithoutMime,
 		"filename": inputStruct.Filename,
 	}
 

operator/document/v0/execution/pdf_checker.py

@@ -0,0 +1,21 @@
+from io import BytesIO
+import json
+import base64
+import sys
+
+# TODO: Deal with the import error when running the code in the docker container
+# from pdf_to_markdown import PDFTransformer
+
+if __name__ == "__main__":
+    json_str   = sys.stdin.buffer.read().decode('utf-8')
+    params     = json.loads(json_str)
+    pdf_string = params["PDF"]
+
+    decoded_bytes = base64.b64decode(pdf_string)
+    pdf_file_obj = BytesIO(decoded_bytes)
+    pdf = PDFTransformer(x=pdf_file_obj)
+    pages = pdf.raw_pages
+    output = {
+        "required": len(pages) == 0,
+    }
+    print(json.dumps(output))

operator/document/v0/helper.go

@@ -0,0 +1,43 @@
+package document
+
+import (
+	"encoding/json"
+	"log"
+
+	"github.com/instill-ai/component/base"
+	"github.com/instill-ai/component/internal/util"
+)
+
+func RequiredToRepair(pdfBase64 string) bool {
+
+	paramsJSON := map[string]interface{}{
+		"PDF": base.TrimBase64Mime(pdfBase64),
+	}
+
+	pythonCode := pdfTransformer + pdfChecker
+
+	outputBytes, err := util.ExecutePythonCode(pythonCode, paramsJSON)
+
+	if err != nil {
+		// It shouldn't block the original process.
+		log.Println("failed to run python script: %w", err)
+		return false
+	}
+
+	var output struct {
+		Repair bool `json:"required"`
+	}
+
+	err = json.Unmarshal(outputBytes, &output)
+
+	if err != nil {
+		// It shouldn't block the original process.
+		log.Println("failed to unmarshal output: %w", err)
+	}
+
+	return output.Repair
+}
+
+func RepairPDF(pdfBase64 string) (string, error) {
+	return ConvertToPDF(pdfBase64, "pdf")
+}

operator/document/v0/main.go

@@ -36,6 +36,9 @@ var (
 	//go:embed execution/task_convert_to_images.py
 	taskConvertToImagesExecution string
 
+	//go:embed execution/pdf_checker.py
+	pdfChecker string
+
 	once sync.Once
 	comp *component
 )

operator/document/v0/pdf_to_markdown_converter.go

@@ -19,8 +19,19 @@ type converterOutput struct {
 
 func convertPDFToMarkdownWithPDFPlumber(base64Text string, displayImageTag bool, displayAllPage bool) (converterOutput, error) {
 
+	var pdfBase64 string
+	var err error
+	if RequiredToRepair(base64Text) {
+		pdfBase64, err = RepairPDF(base64Text)
+		if err != nil {
+			return converterOutput{}, fmt.Errorf("failed to repair PDF: %w", err)
+		}
+	} else {
+		pdfBase64 = base.TrimBase64Mime(base64Text)
+	}
+
 	paramsJSON, err := json.Marshal(map[string]interface{}{
-		"PDF":                    base.TrimBase64Mime(base64Text),
+		"PDF":                    pdfBase64,
 		"display-image-tag":      displayImageTag,
 		"display-all-page-image": displayAllPage,
 	})