Improve OCR and add file retention toggle

Author: TheQuantifier

api/src/config/env.js

@@ -50,6 +50,7 @@ const env = {
     ? path.resolve(process.env.OCR_WORKER_SCRIPT)
     : path.resolve(__dirname, "..", "..", "worker", "ocr_demo.py"),
   pythonBin: process.env.PYTHON_BIN || null,
+  keepReceiptFiles: boolFromEnv(process.env.RECEIPT_KEEP_FILES, true),
 
   // AI Parser
   aiProvider: (process.env.AI_PROVIDER || "gemini").toLowerCase(),

api/src/controllers/receipts.controller.js

@@ -3,6 +3,7 @@ import asyncHandler from "../middleware/async.js";
 
 import { parseReceiptText } from "../services/aiParser.service.js";
 import { runOcrBuffer } from "../services/ocr.service.js";
+import env from "../config/env.js";
 import { parseDateOnly } from "./records.controller.js";
 
 import { query } from "../config/db.js";
@@ -150,6 +151,17 @@ export const confirmUpload = asyncHandler(async (req, res) => {
     });
   }
 
+  // 7) Optional: remove uploaded file after processing (useful for testing)
+  if (!env.keepReceiptFiles) {
+    try {
+      if (receipt.object_key) {
+        await deleteObject({ key: receipt.object_key });
+      }
+    } catch (err) {
+      console.error("Error deleting R2 object after OCR", receipt.id, err);
+    }
+  }
+
   res.status(200).json({
     receipt: updatedReceipt,
     autoRecord,

api/src/services/ocr.service.js

@@ -44,6 +44,9 @@ export function runOcrBuffer(buffer) {
 
       try {
         const parsed = JSON.parse(stdout);
+        if (parsed?.error) {
+          return reject(new Error(`OCR worker error: ${parsed.error}`));
+        }
         return resolve(parsed);
       } catch {
         return reject(new Error(`Failed to parse OCR output: ${stdout}`));

worker/ocr_demo.py

@@ -3,19 +3,65 @@
 import json
 import fitz          # PyMuPDF
 import pytesseract
-from PIL import Image, ImageFile
+from PIL import Image, ImageFile, ImageOps, ImageFilter
 import io
 
 # Prevent truncated-image crashes in Pillow
 ImageFile.LOAD_TRUNCATED_IMAGES = True
 
 
+TESSERACT_CONFIG = "--oem 3 --psm 6 -l eng"
+
+
+def _preprocess_image(img: Image.Image) -> Image.Image:
+    # Normalize orientation, boost contrast, and improve OCR readability.
+    try:
+        img = ImageOps.exif_transpose(img)
+    except Exception:
+        pass
+
+    img = img.convert("L")
+    img = ImageOps.autocontrast(img)
+
+    # Scale up small images for better OCR, cap large images to avoid huge memory use.
+    max_side = max(img.size)
+    if max_side < 1000:
+        scale = 1000 / max_side
+        img = img.resize((int(img.size[0] * scale), int(img.size[1] * scale)), Image.BICUBIC)
+    elif max_side > 3000:
+        scale = 3000 / max_side
+        img = img.resize((int(img.size[0] * scale), int(img.size[1] * scale)), Image.BICUBIC)
+
+    img = img.filter(ImageFilter.SHARPEN)
+    return img
+
+
+def _ocr_image(img: Image.Image) -> str:
+    img = _preprocess_image(img)
+    text = pytesseract.image_to_string(img, config=TESSERACT_CONFIG)
+    return text or ""
+
+
+def _render_page_to_image(page) -> Image.Image:
+    # Render PDF page to raster image for OCR fallback.
+    pix = page.get_pixmap(dpi=200)
+    mode = "RGB" if pix.alpha == 0 else "RGBA"
+    return Image.frombytes(mode, [pix.width, pix.height], pix.samples)
+
+
 def process_pdf(buffer: bytes) -> str:
     try:
         pdf = fitz.open(stream=buffer, filetype="pdf")
         text = ""
         for page in pdf:
             text += page.get_text() or ""
+        if text.strip():
+            return text
+
+        # Fallback: render pages and OCR if no embedded text exists.
+        for page in pdf:
+            img = _render_page_to_image(page)
+            text += _ocr_image(img)
         return text
     except Exception as e:
         # Output to stderr so Node can capture useful debugging info
@@ -26,8 +72,7 @@ def process_pdf(buffer: bytes) -> str:
 def process_image(buffer: bytes) -> str:
     try:
         img = Image.open(io.BytesIO(buffer))
-        text = pytesseract.image_to_string(img)
-        return text or ""
+        return _ocr_image(img)
     except Exception as e:
         print(f"Image processing error: {e}", file=sys.stderr)
         return ""
@@ -49,17 +94,22 @@ def main():
     # Detect PDFs safely
     is_pdf = buffer.startswith(b"%PDF") or buffer[:4] == b"\x25\x50\x44\x46"
 
-    if is_pdf:
-        text = process_pdf(buffer)
-    else:
-        text = process_image(buffer)
+    error = ""
+    try:
+        if is_pdf:
+            text = process_pdf(buffer)
+        else:
+            text = process_image(buffer)
+    except Exception as e:
+        error = str(e)
+        text = ""
 
     # Always output valid JSON
     try:
-        print(json.dumps({"text": text}))
+        print(json.dumps({"text": text, "error": error}))
     except Exception as e:
         print(f"JSON output error: {e}", file=sys.stderr)
-        print('{"text": ""}')
+        print('{"text": "", "error": "json_output_error"}')
 
 
 if __name__ == "__main__":