update

app.py

@@ -13,8 +13,7 @@ load_dotenv()
 API_KEY = os.getenv("API_KEY")
 MODEL_ID = "scb10x/typhoon-ocr-7b"
 
-# Load pipeline (require torch)
-ocr_pipeline = pipeline("document-question-answering", model=MODEL_ID)
+ocr_pipeline = pipeline("image-to-text", model="scb10x/typhoon-ocr-7b")
 
 # FastAPI app init
 app = FastAPI()
@@ -31,11 +30,8 @@ def pdf_to_image(file_bytes: bytes) -> Image.Image:
     return images[0]  # Only first page for now
 
 def run_ocr(image: Image.Image) -> str:
-    result = ocr_pipeline({
-        "image": image,
-        "question": "อ่านข้อความทั้งหมดในภาพนี้"
-    })
-    return result[0]["answer"] if isinstance(result, list) else str(result)
+    result = ocr_pipeline(image)
+    return result[0]["generated_text"]
 
 def preprocess_text(text: str) -> str:
     text = re.sub(r"</?(figure|table|tr|td|th|b|i|u|p|div|span)[^>]*>", "\n", text)