Update app.py

app.py

@@ -4,61 +4,54 @@ from fastapi.middleware.cors import CORSMiddleware
 import shutil
 import os
 from uuid import uuid4
+from concurrent.futures import ThreadPoolExecutor
 from marker.converters.pdf import PdfConverter
 from marker.models import create_model_dict
 from marker.output import text_from_rendered
-from threading import Lock
+import asyncio
 
 app = FastAPI()
 
-# Enable CORS for all origins
+# CORS
 app.add_middleware(
     CORSMiddleware,
-    allow_origins=["*"],  # Allow all origins
+    allow_origins=["*"],
     allow_credentials=True,
-    allow_methods=["*"],  # Allow all HTTP methods
-    allow_headers=["*"],  # Allow all headers
+    allow_methods=["*"],
+    allow_headers=["*"],
 )
 
-# Singleton class for PdfConverter
-class PdfConverterSingleton:
-    _instance = None
-    _lock = Lock()
+# Initialize converter artifact once
+ARTIFACT_DICT = create_model_dict()
 
-    def __new__(cls):
-        if cls._instance is None:
-            with cls._lock:
-                if cls._instance is None:
-                    instance = super().__new__(cls)
-                    instance._initialize()
-                    cls._instance = instance
-        return cls._instance
+# ThreadPoolExecutor: use 2 * number of vCPUs for I/O-bound tasks
+EXECUTOR = ThreadPoolExecutor(max_workers=8)  # Tune based on PDF size
 
-    def _initialize(self):
-        self.converter = PdfConverter(artifact_dict=create_model_dict())
+def _extract_text_sync(pdf_path: str) -> str:
+    """Blocking function to extract text from a PDF using PdfConverter."""
+    converter = PdfConverter(artifact_dict=ARTIFACT_DICT)  # fresh instance
+    rendered = converter(pdf_path)
+    text, _, _ = text_from_rendered(rendered)
+    return str(text)
 
-    def get_text(self, pdf_path: str) -> str:
-        rendered = self.converter(pdf_path)
-        text, _, _ = text_from_rendered(rendered)
-        return str(text)
+async def extract_text_from_pdf(pdf_path: str) -> str:
+    """Async wrapper for the blocking PDF extraction."""
+    loop = asyncio.get_running_loop()
+    return await loop.run_in_executor(EXECUTOR, _extract_text_sync, pdf_path)
 
-# API function to call converter
-def extract_text_from_pdf(pdf_path: str) -> str:
-    return PdfConverterSingleton().get_text(pdf_path)
-
-# Endpoint to upload a file and extract markdown text
 @app.post("/extract-pdf-text")
 async def extract_pdf_text(file: UploadFile = File(...)):
     if file.content_type != "application/pdf":
         raise HTTPException(status_code=400, detail="Only PDF files are supported.")
 
     temp_filename = f"/tmp/{uuid4().hex}.pdf"
+
     try:
         with open(temp_filename, "wb") as buffer:
             shutil.copyfileobj(file.file, buffer)
 
-        text = extract_text_from_pdf(temp_filename)
-        return JSONResponse(content={"markdown_text": text})
+        markdown_text = await extract_text_from_pdf(temp_filename)
+        return JSONResponse(content={"markdown_text": markdown_text})
 
     except Exception as e:
         raise HTTPException(status_code=500, detail=str(e))