fix: add post markdown

app/services/ocr_service.py

@@ -85,6 +85,8 @@ def _split_glued_command_token(token: str) -> str:
 
 def _postprocess_math(expr: str) -> str:
     """Postprocess a *math* expression (already inside $...$ or $$...$$)."""
+    # stage0: fix OCR number errors (digits with spaces)
+    expr = _fix_ocr_number_errors(expr)
     # stage1: split glued command tokens (e.g. \cdotdS)
     expr = _COMMAND_TOKEN_PATTERN.sub(lambda m: _split_glued_command_token(m.group(0)), expr)
     # stage2: normalize differentials (keep conservative)
@@ -93,6 +95,42 @@ def _postprocess_math(expr: str) -> str:
     return expr
 
 
+def _fix_ocr_number_errors(expr: str) -> str:
+    """Fix common OCR errors in LaTeX math expressions.
+
+    OCR often splits numbers incorrectly, especially decimals:
+    - "2 2. 2" should be "22.2"
+    - "3 0. 4" should be "30.4"
+    - "1 5 0" should be "150"
+
+    This function merges digit sequences that are separated by spaces.
+
+    Args:
+        expr: LaTeX math expression.
+
+    Returns:
+        LaTeX expression with number errors fixed.
+    """
+    # Fix pattern 1: "digit space digit(s). digit(s)" → "digit digit(s).digit(s)"
+    # Example: "2 2. 2" → "22.2"
+    expr = re.sub(r'(\d)\s+(\d+)\.\s*(\d+)', r'\1\2.\3', expr)
+    
+    # Fix pattern 2: "digit(s). space digit(s)" → "digit(s).digit(s)"
+    # Example: "22. 2" → "22.2"
+    expr = re.sub(r'(\d+)\.\s+(\d+)', r'\1.\2', expr)
+    
+    # Fix pattern 3: "digit space digit" (no decimal point, within same number context)
+    # Be careful: only merge if followed by decimal point or comma/end
+    # Example: "1 5 0" → "150" when followed by comma or end
+    expr = re.sub(r'(\d)\s+(\d)(?=\s*[,\)]|$)', r'\1\2', expr)
+    
+    # Fix pattern 4: Multiple spaces in decimal numbers
+    # Example: "2  2  .  2" → "22.2"
+    expr = re.sub(r'(\d)\s+(\d)(?=\s*\.)', r'\1\2', expr)
+    
+    return expr
+
+
 def _postprocess_markdown(markdown_content: str) -> str:
     """Apply LaTeX postprocessing only within $...$ / $$...$$ segments."""
     if not markdown_content: