Skip to content

Sync-in · Sovereign platform for file storage, sharing, synchronization, and collaboration.

fix(backend:files): add support for page rotation during OCR extraction

johaven6837cc4ea8f1
diff --git a/backend/src/applications/files/utils/doc-textify/adapters/pdf.ts b/backend/src/applications/files/utils/doc-textify/adapters/pdf.ts
index 370231c2..fb22c750 100644
--- a/backend/src/applications/files/utils/doc-textify/adapters/pdf.ts
+++ b/backend/src/applications/files/utils/doc-textify/adapters/pdf.ts
@@ -61,7 +61,7 @@ function extractText(items: unknown[], newlineDelimiter: string): string {
return fragments.join('')
}
-async function extractTextFromImages(images: ExtractedImages, options: DocTextifyOptions): Promise<string[]> {
+async function extractTextFromImages(images: ExtractedImages, options: DocTextifyOptions, pageRotation = 0): Promise<string[]> {
const contents: string[] = []
const worker: DocTextifyOCRWorkerLike | undefined = options.ocrWorker
@@ -87,6 +87,10 @@ async function extractTextFromImages(images: ExtractedImages, options: DocTextif
imageProcessor = imageProcessor.flatten({ background: '#ffffff' })
}
+ if (pageRotation) {
+ imageProcessor = imageProcessor.rotate(pageRotation)
+ }
+
imageBuffer = await imageProcessor
.resize({
width: 1800,
@@ -156,7 +160,8 @@ export async function parsePdf(filePath: string, options: DocTextifyOptions): Pr
if (!pageHasImages) {
continue
}
- const ocrContents = await extractTextFromImages(images, options)
+ const viewport = page.getViewport({ scale: 1 })
+ const ocrContents = await extractTextFromImages(images, options, viewport.rotation)
if (ocrContents.length > 0) {
contents.push(...ocrContents)
}