Spaces:
Running
Running
Commit History
-
sync: update pdf-script from github.com/peyajm29/pdf-script@488b4f6e90561a9c6cce3f0166f0690793886b59 Running
github-actions[bot]c85cca0 -
sync: update pdf-script from github.com/peyajm29/pdf-script@3538c5b239a9e0e76c810f18b40bae862f5b4ab2
github-actions[bot]a6407e9 -
sync: update backend/deploy files from github.com/peyajm29/pdf-script-web@066eb2df655864389b0829938de5f3c4ad519054
github-actions[bot]037a9f2 -
sync: update pdf-script from github.com/peyajm29/pdf-script@e01da77cdffd02bb939e97990a9f77de211d8c4d
github-actions[bot]7b3e462 -
sync: update pdf-script from github.com/peyajm29/pdf-script@70bff5d4e8a7b6412ca4848ad01dca86cf32103c
github-actions[bot]1678f56
-
Update backend/routes/download.py
Verified signature7af94c3 -
Create stats.py
Verified signature3f2e280 -
Update backend/main.py
Verified signature549db0f -
Update backend/database.py
Verified signature6dedb20
-
sync: fix PDF candidate selection to consider all PDFs and prefer main.pdf/paper.pdf at root
peyajm3088c6f
-
Fix #33: restore dc:creator XMP field to real authors after setCreator()
peyajm +1: Claude Sonnet 4.6iText auto-syncs xmp:CreatorTool from setCreator(), which overwrites dc:creator (shown as "Authors" in macOS Finder) with "iTagPDF". Explicitly re-set dc:creator in XMP to the extracted author names after the sync. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>5d94592 -
Fix #32: set PDF Creator field to iTagPDF (www.itagpdf.com)
peyajm +1: Claude Sonnet 4.6Sync from peyajm29/pdf-script@8c08e3a Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>6264d44
-
Fix #27: generate one LI per list item, not per line
peyajm +1: Claude Sonnet 4.6Sync from peyajm29/pdf-script@1e7a6ce Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>b858f16
-
fix: downscale large images before Tesseract OCR to prevent hangs
peyajm +1: Claude Sonnet 4.6Poster-sized PDFs at 200 DPI produce images up to ~4677x6610px which causes Tesseract to take several minutes per page on CPU-only hardware. Downscale to max 1800px on the longest side before OCR, then scale bboxes back up to original coordinates so downstream YOLO mapping is unaffected. Removes the timeout entirely since the root cause is fixed. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>a00cf0b -
fix: replace executor on cancel to unblock new pipeline runs
peyajm +1: Claude Sonnet 4.6Cancelling the asyncio task left the blocking thread (Tesseract/YOLO) still running in the single-worker ThreadPoolExecutor, causing any subsequent pipeline run to queue behind it and appear stuck on stage 1. On cancel, swap in a fresh executor so the next run starts immediately. Also adds per-page OCR progress logging and a 120s Tesseract timeout to surface hangs instead of silently blocking forever. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>e083911 -
debug: add trace prints to upload route to find hang point
peyajm +1: Claude Sonnet 4.6Also fix pdfinfo_from_path blocking the async event loop by wrapping it in run_in_threadpool. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>3d24b16 -
fix: log google auth error reason; export pipeline env vars in start.sh
peyajm3f24feb
-
Add itagpdf.com to CORS allowed origins
peyajm +1: Claude Sonnet 4.6Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>5c529ab -
fix: commit pdftagger.jar via LFS (was excluded by .gitignore)
peyajm +1: Claude Sonnet 4.6The JAR was listed in pdf-script/.gitignore so it was never committed to the repo. Removing that exclusion so it is tracked via Git LFS and available inside the Docker container at /app/pdf-script/pdftagger.jar. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>3542f67
-
Sync latest backend and pipeline code
peyajm- Add auth system: Google OAuth, JWT, SQLite database - Add sessions route, dependencies, database module - Update config with GOOGLE_CLIENT_ID, JWT_SECRET, DB_PATH - Update requirements: google-auth, python-jose, httpx - Update all backend routes and services to latest - Dockerfile: remove maven build step (use pre-built JAR) - pdf-script: update labeling.py and latex_processor.py - Include pre-built pdftagger.jarfb7b1a4
-
fix: remove pipeline cancellation so concurrent users queue instead of cancel
peyajm87cf1c9 -
fix: mvn shade outputs pdftagger.jar to project root, no cp needed
peyajm9f3ffaf -
exclude large notebook from repo
peyajm72da2ec -
add pdf-script as plain directory
peyajmc6c9042 -
remove pdf-script submodule reference
peyajm8125b1e -
build pdftagger.jar from source using maven
peyajm97aa9e5 -
add pdftagger.jar via LFS
peyajme798bd7 -
fix: use default-jdk (openjdk-17 not available in Debian trixie)
peyajm21d7c29 -
Initial deployment
peyajm42d4c60