Update app.py

app.py

@@ -20,7 +20,6 @@ logging.basicConfig(level=logging.INFO, format='%(asctime)s %(levelname)s: %(mes
 FONT_PATH = os.path.join(os.path.dirname(__file__), "NanumGothicCoding.ttf")
 MIN_FONT_SIZE = 8
 
-
 def get_vision_client():
     b64 = os.getenv("GCP_SERVICE_ACCOUNT_JSON")
     if not b64:
@@ -37,6 +36,20 @@ def get_vision_client():
 vision_client = get_vision_client()
 checker = spell_init()
 
+# ──────────────────────────────── KoSpellPy 긴 텍스트 안전 처리 ────────────────────────────────
+def chunk_text(text, max_len=500):
+    return [text[i:i+max_len] for i in range(0, len(text), max_len)]
+
+def safe_kospell_check(text):
+    parts = chunk_text(text)
+    corrected = []
+    for part in parts:
+        try:
+            corrected.append(checker(part))
+        except Exception as e:
+            logging.warning(f"[Spell] 일부 텍스트 교정 실패: {e}")
+            corrected.append(part)  # 오류 발생 시 해당 부분은 원문 사용
+    return ' '.join(corrected)
 
 def normalize_text(text: str) -> str:
     return unicodedata.normalize('NFC', text)
@@ -58,7 +71,6 @@ def preprocess_with_adaptive_threshold(img: Image.Image) -> Image.Image:
     bgr = cv2.cvtColor(adap, cv2.COLOR_GRAY2BGR)
     return Image.fromarray(cv2.cvtColor(bgr, cv2.COLOR_BGR2RGB))
 
-
 def ocr_overlay_and_correct_text(img: Image.Image):
     corrected_text = ""
     overlay = None
@@ -73,9 +85,14 @@ def ocr_overlay_and_correct_text(img: Image.Image):
         )
         ann = res.full_text_annotation
         raw = ann.text.replace('\n', ' ').strip()
-        logging.info(f"[OCR] Raw: {raw}")
-        corrected_text = checker(raw)
-        logging.info(f"[Spell] Corrected: {corrected_text}")
+        logging.info(f"[OCR] Raw length: {len(raw)} / Raw: {raw}")
+
+        try:
+            corrected_text = safe_kospell_check(raw)
+            logging.info(f"[Spell] Corrected: {corrected_text}")
+        except Exception as e:
+            logging.error(f"[Spell] 교정 중 오류 발생: {e}")
+            corrected_text = raw  # 오류 시 원문 반환
 
         syms = []
         for pg in ann.pages:
@@ -142,10 +159,13 @@ def ocr_overlay_and_correct_text(img: Image.Image):
 
     return overlay, corrected_text
 
-
 def text_correct_fn(text):
     raw = normalize_text(text.strip())
-    corrected = checker(raw)
+    try:
+        corrected = safe_kospell_check(raw)
+    except Exception as e:
+        logging.error(f"[Spell/TextInput] 교정 중 오류 발생: {e}")
+        corrected = raw
     return None, corrected
 
 def img_correct_fn(blob):
@@ -154,7 +174,6 @@ def img_correct_fn(blob):
         img = Image.open(io.BytesIO(blob)).convert('RGB')
     return ocr_overlay_and_correct_text(img)
 
-
 with gr.Blocks(
     css="""
         .gradio-container {background-color: #fafaf5}
@@ -201,6 +220,5 @@ with gr.Blocks(
         return None, gr.update(label="사진 촬영 및 업로드", interactive=True, value=None), '', gr.update(interactive=False), None, ''
     clear_btn.click(on_clear, None, [state, upload, text_in, img_check_btn, img_out, txt_out])
 
-
 if __name__ == '__main__':
     demo.launch(server_name="0.0.0.0", server_port=int(os.getenv("PORT", 7860)))