Auto-update via watch script

app.py

@@ -220,6 +220,66 @@ def _spinner(msg: str, color: str = "#3b82f6", bg: str = "rgba(59,130,246,0.08)"
 # Main processing functions
 # ---------------------------------------------------------------------------
 
+def _preprocess_image(image: Image.Image) -> Image.Image:
+    img = image.convert("L")
+    w, h = img.size
+    if w < 1800:
+        scale = 1800 / w
+        img = img.resize((int(w * scale), int(h * scale)), getattr(Image, "Resampling", Image).LANCZOS)
+    img = ImageEnhance.Contrast(img).enhance(1.8)
+    img = ImageEnhance.Sharpness(img).enhance(2.0)
+    return img.convert("RGB")
+
+
+def _column_aware_ocr(image: Image.Image) -> str:
+    data = pytesseract.image_to_data(image, config="--psm 11 -l eng", output_type=pytesseract.Output.DICT)
+    img_width = image.size[0]
+
+    words = []
+    for i, text in enumerate(data["text"]):
+        text = text.strip()
+        if not text or data["word_num"][i] == 0 or int(data["conf"][i]) < 30:
+            continue
+        words.append({"text": text, "x": data["left"][i], "y": data["top"][i]})
+
+    if not words:
+        return ""
+
+    gap_threshold = img_width * 0.15
+    sorted_x = sorted(set(w["x"] for w in words))
+    col_starts = [0]
+    for idx in range(1, len(sorted_x)):
+        if sorted_x[idx] - sorted_x[idx - 1] > gap_threshold:
+            col_starts.append(sorted_x[idx])
+
+    def assign_col(x):
+        col = 0
+        for j, start in enumerate(col_starts):
+            if x >= start:
+                col = j
+        return col
+
+    for word in words:
+        word["col"] = assign_col(word["x"])
+
+    result_parts = []
+    for col_idx in range(len(col_starts)):
+        col_words = sorted([w for w in words if w["col"] == col_idx], key=lambda w: (w["y"], w["x"]))
+        if not col_words:
+            continue
+        lines, current = [], [col_words[0]]
+        for word in col_words[1:]:
+            if abs(word["y"] - current[0]["y"]) <= 8:
+                current.append(word)
+            else:
+                lines.append(current)
+                current = [word]
+        lines.append(current)
+        result_parts.append("\n".join(" ".join(w["text"] for w in line) for line in lines))
+
+    return "\n\n".join(result_parts)
+
+
 def extract_full_article(image: Image.Image, hf_token: str) -> tuple[str, str]:
     """
     Hybrid extraction: