v3.3.4-rc3 v1.15.0
vercel-labs/agent-browserAug 15, 2026 → Aug 17, 20262 days between releases
What's New
Switched from ACME/Lego infrastructure updates to a new selective OCR pipeline using PDFium and PP-OCRv6.
New Features
- Selective OCR for scanned and mixed PDFs using PDFium and PP-OCRv6.
- Recovers credible PDFium text layers to route only affected pages.
- Reuses OCR sessions and preserves row boundaries for faster extraction.
- Adds pinned runtime and model caching for offline operation.
- Exposes OCR pipeline through Rust, pdf2md, Node.js, and Python.
Improvements
- Reuses OCR sessions and preserves row boundaries for faster extraction.
- Adds pinned runtime and model caching for offline operation.
- Recovers credible PDFium text layers to route only affected pages.
Bug Fixes
- Recovers credible PDFium text layers to route only affected pages (fixes incomplete text extraction).
Upgrade Notes
Configure model caching directories for offline operation and ensure pinned runtime versions are installed.
The Two Releases
v3.3.4-rc3
Aug 15, 2026· 2 months ago
This release candidate focuses on upgrading the ACME client to Lego v5 for better infrastructure support and fixes critical connectivity issues like SSO redirect loops and virtual directory conflicts.
Full release notes1.15.0 — Selective OCR for scanned and mixed PDFs
Aug 17, 2026· 2 months ago
This release introduces selective OCR capabilities for scanned and mixed PDFs using PDFium rendering and PP-OCRv6. It focuses on recovering existing text layers before applying OCR, reusing sessions for performance, and exposing the OCR pipeline across Rust, Python, and Node.js.
Full release notes