A fast, helpful, and open-source document parser
Honor OCR rotation option in Tesseract engine
Max Headley4fdf3c9358c2
diff --git a/.changeset/tame-mirrors-bake.md b/.changeset/tame-mirrors-bake.mdnew file mode 100644index 0000000..8255808--- /dev/null+++ b/.changeset/tame-mirrors-bake.md@@ -0,0 +1,5 @@+---+"@llamaindex/liteparse": patch+---++Honor the OCR rotation-correction option in the built-in Tesseract engine by mapping it to Tesseract.js auto-rotation.diff --git a/src/engines/ocr/tesseract.ts b/src/engines/ocr/tesseract.tsindex 87b7abd..ca344c1 100644--- a/src/engines/ocr/tesseract.ts+++ b/src/engines/ocr/tesseract.ts@@ -117,7 +117,12 @@ export class TesseractEngine implements OcrEngine {// In tesseract.js v6+, we need to enable blocks output to get word-level dataconst {data: { blocks },- } = await this.scheduler.addJob("recognize", image, {}, { blocks: true });+ } = await this.scheduler.addJob(+ "recognize",+ image,+ options.correctRotation ? { rotateAuto: true } : {},+ { blocks: true }+ );// Extract words from hierarchical blocks structure: blocks → paragraphs → lines → wordsconst results: OcrResult[] = [];