From 9f95ede409e232fc4f96a43ef8cf5dac4b4230de Mon Sep 17 00:00:00 2001 From: NotYuSheng Date: Tue, 7 Oct 2025 11:03:42 +0800 Subject: [PATCH] fix: timeout issues --- frontend/main.py | 10 +---- frontend/my_pages/1_upload_UI.py | 50 ++++++++++++--------- pdf_extraction_service/main.py | 4 ++ pdf_extraction_service/routers/extractor.py | 16 ++++--- pdf_extraction_service/utils/caption.py | 4 +- 5 files changed, 47 insertions(+), 37 deletions(-) diff --git a/frontend/main.py b/frontend/main.py index 70efa14c..07ccb346 100644 --- a/frontend/main.py +++ b/frontend/main.py @@ -14,7 +14,7 @@ # Page configuration st.set_page_config( page_title="OmniPDF", - page_icon="🦸", + page_icon="🔍", layout="wide", initial_sidebar_state="expanded", ) @@ -110,12 +110,6 @@ icon="📄", ) - settings_UI = st.Page( - page="my_pages/10_settings_UI.py", - title="Settings", - icon="⚙️" - ) - # To go between the different pages pg = st.navigation( pages=[ @@ -125,12 +119,10 @@ wordcloud_UI, translate_UI, metadata_UI, - settings_UI, ] ) # Add version number to sidebar footer - st.sidebar.markdown("---") st.sidebar.markdown( f"
v{html.escape(__version__)}
", unsafe_allow_html=True diff --git a/frontend/my_pages/1_upload_UI.py b/frontend/my_pages/1_upload_UI.py index a2302932..883fb4de 100644 --- a/frontend/my_pages/1_upload_UI.py +++ b/frontend/my_pages/1_upload_UI.py @@ -18,8 +18,6 @@ if "uploaded_files" not in st.session_state: st.session_state.uploaded_files = None -client = httpx.AsyncClient(cookies=st.session_state.httpx_cookies, timeout=300.0) # 5 minute timeout for long translations - # Language options for translation LANGUAGES = { "auto": "Auto-detect", @@ -38,7 +36,7 @@ } -async def check_job_status(doc_id: str, endpoint: str): +async def check_job_status(doc_id: str, endpoint: str, client: httpx.AsyncClient): """Check the status of a specific job.""" try: response = await client.get(f"{PDF_PROCESSOR_URL}/{endpoint}/{doc_id}") @@ -56,7 +54,7 @@ async def check_job_status(doc_id: str, endpoint: str): return "error" -async def poll_processing_status(doc_id: str, status_container, progress_text, source_lang="", target_lang="", max_attempts=120, delay=2): +async def poll_processing_status(doc_id: str, status_container, progress_text, client: httpx.AsyncClient, source_lang="", target_lang="", delay=2): """ Poll the backend for processing status and update the UI. @@ -93,24 +91,24 @@ async def poll_processing_status(doc_id: str, status_container, progress_text, s translation_triggered = False renderer_triggered = False - for attempt in range(max_attempts): + while True: all_complete = True status_lines = [] # STAGE 1: Check extraction status - extraction_status = await check_job_status(doc_id, "extractor") + extraction_status = await check_job_status(doc_id, "extractor", client) stages["extraction"]["status"] = extraction_status # STAGE 2: After extraction completes, check embedding and translation concurrently if extraction_status == "completed": # Check embedding status - embedding_status = await check_job_status(doc_id, "embed/sentence") + embedding_status = await check_job_status(doc_id, "embed/sentence", client) stages["embedding"]["status"] = embedding_status # Check/trigger translation (concurrent with embedding) if "translation" in stages: if not translation_triggered: - translation_status = await check_job_status(doc_id, "translation") + translation_status = await check_job_status(doc_id, "translation", client) # If translation not started, trigger it if translation_status == "not_started": @@ -143,13 +141,13 @@ async def poll_processing_status(doc_id: str, status_container, progress_text, s stages["translation"]["status"] = translation_status else: # Already triggered, just check status - translation_status = await check_job_status(doc_id, "translation") + translation_status = await check_job_status(doc_id, "translation", client) stages["translation"]["status"] = translation_status # STAGE 3: Concurrent processing after stage 2 completes # Metadata: Check after embedding completes (translation can still be running) if embedding_status == "completed": - metadata_status = await check_job_status(doc_id, "metadata") + metadata_status = await check_job_status(doc_id, "metadata", client) stages["metadata"]["status"] = metadata_status # Renderer: Check/trigger after translation completes (embedding can still be running) @@ -157,7 +155,7 @@ async def poll_processing_status(doc_id: str, status_container, progress_text, s translation_status = stages["translation"]["status"] if translation_status == "completed": if not renderer_triggered: - renderer_status = await check_job_status(doc_id, "renderer") + renderer_status = await check_job_status(doc_id, "renderer", client) # If renderer not started, trigger it if renderer_status == "not_started": @@ -189,7 +187,7 @@ async def poll_processing_status(doc_id: str, status_container, progress_text, s stages["renderer"]["status"] = renderer_status else: # Already triggered, just check status - renderer_status = await check_job_status(doc_id, "renderer") + renderer_status = await check_job_status(doc_id, "renderer", client) stages["renderer"]["status"] = renderer_status # Build status display grouped by stage @@ -261,30 +259,37 @@ async def poll_processing_status(doc_id: str, status_container, progress_text, s await asyncio.sleep(delay) - status_container.warning("⚠️ Processing timeout. Some stages may still be running.") - progress_text.empty() - return False - async def process_pdf(uploaded_file, file_expander, source_lang="", target_lang=""): """ Uploads PDF to backend and stores document metadata in session state. Optionally triggers translation if languages are specified. """ - # Process pdf through PDF_processor endpoint try: - # Upload the PDF document + # Upload the PDF document with longer timeout for file upload logger.info(f"Uploading PDF: {uploaded_file}") bytes_data = uploaded_file.getvalue() # bytes files = {"file": (uploaded_file.name, bytes_data, "application/pdf")} - upload_response = await client.post( + upload_client = httpx.AsyncClient( + cookies=st.session_state.httpx_cookies, + timeout=httpx.Timeout(connect=10.0, read=120.0, write=120.0, pool=10.0) + ) + upload_response = await upload_client.post( f"{PDF_PROCESSOR_URL}/documents/", files=files ) + await upload_client.aclose() if upload_response.cookies: st.session_state.httpx_cookies.update(upload_response.cookies) + # Create polling client AFTER upload to ensure it has the latest session cookies + # Use longer read timeout (120s) to handle translation/renderer trigger requests for large PDFs + poll_client = httpx.AsyncClient( + cookies=st.session_state.httpx_cookies, + timeout=httpx.Timeout(connect=10.0, read=120.0, write=10.0, pool=10.0) + ) + logger.info(f"Upload PDF response: {upload_response.text}") upload_data = upload_response.json() @@ -312,17 +317,20 @@ async def process_pdf(uploaded_file, file_expander, source_lang="", target_lang= # Poll for processing status with spinner with st.spinner("Processing document..."): - await poll_processing_status(doc_id, status_container, progress_text, source_lang, target_lang) + await poll_processing_status(doc_id, status_container, progress_text, poll_client, source_lang, target_lang) except Exception as e: with file_expander: st.error(f"❌ Error processing PDF: {uploaded_file.name}") logger.error(f"Error processing PDF: {e}") + finally: + # Close the client after processing + await poll_client.aclose() -st.markdown('

🦸 OmniPDF

', unsafe_allow_html=True) +st.markdown('

🔍 OmniPDF

', unsafe_allow_html=True) st.header("📁 Upload PDF") uploaded_files = st.file_uploader( "Choose a PDF file", diff --git a/pdf_extraction_service/main.py b/pdf_extraction_service/main.py index bb08848c..bfa9d9ef 100644 --- a/pdf_extraction_service/main.py +++ b/pdf_extraction_service/main.py @@ -8,6 +8,10 @@ level=logging.INFO, format="%(asctime)s [%(levelname)s] %(name)s: %(message)s" ) +# Enable DEBUG logging for docling to see more detailed processing information +logging.getLogger("docling").setLevel(logging.DEBUG) +logging.getLogger("docling_core").setLevel(logging.DEBUG) + app = FastAPI(root_path="/pdf_extraction") # Initialize Prometheus metrics diff --git a/pdf_extraction_service/routers/extractor.py b/pdf_extraction_service/routers/extractor.py index ba9a2519..142bd52e 100644 --- a/pdf_extraction_service/routers/extractor.py +++ b/pdf_extraction_service/routers/extractor.py @@ -1,5 +1,5 @@ import asyncio -from fastapi import APIRouter, BackgroundTasks, HTTPException +from fastapi import APIRouter, BackgroundTasks, HTTPException, Response import logging import time import io @@ -30,7 +30,7 @@ USE_GPU = os.getenv("USE_GPU", "false").lower() == "true" logger.info(f"GPU acceleration enabled: {USE_GPU}") -async def process_pdf(doc_id: str, presign_url: str, img_scale: float = 2.0): +async def process_pdf(doc_id: str, presign_url: str, img_scale: float = 1.0): start_time = time.time() opts = PdfPipelineOptions() @@ -164,18 +164,24 @@ async def submit_pdf(doc_id: str, download_url: str, background_tasks: Backgroun @router.get("/{doc_id}", response_model=ExtractResponse) -async def get_status(doc_id: str): +async def get_status(doc_id: str, response: Response): job = load_job(doc_id=doc_id, job_type=JobType.EXTRACTION) if not job: raise HTTPException(status_code=404, detail="Document ID not found") - if job.get("status") == "failed": + job_status = job.get("status", "unknown") + + if job_status == "failed": error_message = job.get("data", {}).get("message", "Processing failed") raise HTTPException(status_code=500, detail=error_message) job_data = job.get("data", {}) result = job_data.get("result", None) + # Return 202 Accepted if still processing, 200 OK if completed + if job_status == "processing": + response.status_code = 202 + return ExtractResponse( - doc_id=doc_id, status=job.get("status", "unknown"), result=result + doc_id=doc_id, status=job_status, result=result ) diff --git a/pdf_extraction_service/utils/caption.py b/pdf_extraction_service/utils/caption.py index 7434fa92..fa05fdff 100644 --- a/pdf_extraction_service/utils/caption.py +++ b/pdf_extraction_service/utils/caption.py @@ -13,14 +13,14 @@ async def get_caption(doc_id: str, image_id: str, image_url: str) -> str: "image_url": image_url } - async with httpx.AsyncClient() as client: + async with httpx.AsyncClient(timeout=httpx.Timeout(connect=60.0, read=120.0, write=60.0, pool=60.0)) as client: try: response = await client.post(f"{IMAGE_CAPTIONER_URL}/caption/", json=payload) response.raise_for_status() data = response.json() return data.get("caption", "") except (httpx.HTTPStatusError, httpx.RequestError) as e: - logger.error(f"HTTP error retrieving caption: {e}") + logger.error(f"HTTP error retrieving caption for image {image_id}: {type(e).__name__}: {e}", exc_info=True) return "" except Exception as e: logger.error(f"Unexpected error retrieving caption: {e}", exc_info=True)