gaschz/trilium
1
0
Fork 0
mirror of https://github.com/zadam/trilium.git synced 2025-12-06 23:44:25 +01:00
Code Issues Actions 30 Packages Projects Releases Wiki Activity

feat(ocr): automatically process images

Browse Source
This commit is contained in:
Elian Doran 2025-07-26 14:00:35 +03:00
parent 08ca86c68a
commit 72cea245f1
No known key found for this signature in database
2 changed files with 44 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
apps/server/src/services/handlers.ts
View File

@ -140,7 +140,8 @@ eventService.subscribe(eventService.ENTITY_CREATED, ({ entityName, entity }) =>
} else if (entityName === "notes") {
runAttachedRelations(entity, "runOnNoteCreation", entity);
// Automatically process OCR for file notes if OCR is enabled
// Note: OCR processing for images is now handled in image.ts during image processing
// OCR processing for files remains here since they don't go through image processing
if (entity.type === 'file' && ocrService.isOCREnabled()) {
// Check if the file MIME type is supported by any OCR processor
const supportedMimeTypes = ocrService.getAllSupportedMimeTypes();

45
apps/server/src/services/image.ts
View File

@ -12,8 +12,9 @@ import sanitizeFilename from "sanitize-filename";
import isSvg from "is-svg";
import isAnimated from "is-animated";
import htmlSanitizer from "./html_sanitizer.js";
import ocrService, { type OCRResult } from "./ocr/ocr_service.js";
async function processImage(uploadBuffer: Buffer, originalName: string, shrinkImageSwitch: boolean) {
async function processImage(uploadBuffer: Buffer, originalName: string, shrinkImageSwitch: boolean, noteId?: string) {
const compressImages = optionService.getOptionBool("compressImages");
const origImageFormat = await getImageType(uploadBuffer);
@ -24,6 +25,24 @@ async function processImage(uploadBuffer: Buffer, originalName: string, shrinkIm
shrinkImageSwitch = false;
}
// Process OCR on the original (uncompressed) image for best quality
let ocrResult: OCRResult | null = null;
if (noteId && ocrService.isOCREnabled() && origImageFormat) {
const imageMime = getImageMimeFromExtension(origImageFormat.ext);
const supportedMimeTypes = ocrService.getAllSupportedMimeTypes();
if (supportedMimeTypes.includes(imageMime)) {
try {
ocrResult = await ocrService.extractTextFromFile(uploadBuffer, imageMime);
if (ocrResult) {
log.info(`Successfully processed OCR for image ${noteId} (${originalName})`);
}
} catch (error) {
log.error(`Failed to process OCR for image ${noteId}: ${error}`);
}
}
}
let finalImageBuffer;
let imageFormat;
@ -39,7 +58,8 @@ async function processImage(uploadBuffer: Buffer, originalName: string, shrinkIm
return {
buffer: finalImageBuffer,
imageFormat
imageFormat,
ocrResult
};
}
@ -72,12 +92,17 @@ function updateImage(noteId: string, uploadBuffer: Buffer, originalName: string)
note.setLabel("originalFileName", originalName);
// resizing images asynchronously since JIMP does not support sync operation
processImage(uploadBuffer, originalName, true).then(({ buffer, imageFormat }) => {
processImage(uploadBuffer, originalName, true, noteId).then(({ buffer, imageFormat, ocrResult }) => {
sql.transactional(() => {
note.mime = getImageMimeFromExtension(imageFormat.ext);
note.save();
note.setContent(buffer);
// Store OCR result if available
if (ocrResult && note.blobId) {
ocrService.storeOCRResult(note.blobId, ocrResult);
}
});
});
}
@ -108,7 +133,7 @@ function saveImage(parentNoteId: string, uploadBuffer: Buffer, originalName: str
note.addLabel("originalFileName", originalName);
// resizing images asynchronously since JIMP does not support sync operation
processImage(uploadBuffer, originalName, shrinkImageSwitch).then(({ buffer, imageFormat }) => {
processImage(uploadBuffer, originalName, shrinkImageSwitch, note.noteId).then(({ buffer, imageFormat, ocrResult }) => {
sql.transactional(() => {
note.mime = getImageMimeFromExtension(imageFormat.ext);
@ -120,6 +145,11 @@ function saveImage(parentNoteId: string, uploadBuffer: Buffer, originalName: str
}
note.setContent(buffer, { forceSave: true });
// Store OCR result if available
if (ocrResult && note.blobId) {
ocrService.storeOCRResult(note.blobId, ocrResult);
}
});
});
@ -159,7 +189,7 @@ function saveImageToAttachment(noteId: string, uploadBuffer: Buffer, originalNam
}, 5000);
// resizing images asynchronously since JIMP does not support sync operation
processImage(uploadBuffer, originalName, !!shrinkImageSwitch).then(({ buffer, imageFormat }) => {
processImage(uploadBuffer, originalName, !!shrinkImageSwitch, attachment.attachmentId).then(({ buffer, imageFormat, ocrResult }) => {
sql.transactional(() => {
// re-read, might be changed in the meantime
if (!attachment.attachmentId) {
@ -175,6 +205,11 @@ function saveImageToAttachment(noteId: string, uploadBuffer: Buffer, originalNam
}
attachment.setContent(buffer, { forceSave: true });
// Store OCR result if available
if (ocrResult && attachment.blobId) {
ocrService.storeOCRResult(attachment.blobId, ocrResult);
}
});
});