[leeaky]

• An arXiv link (abs, pdf or html form) is read from its abstract page for title and summary, as before — but its word count comes from the full paper's PDF, so papers land in Long with their real minutes. The PDF text is not stored and never reaches the model; the safety scan is unchanged (owner's choice).
• If the PDF can't be fetched or read (or is over 40 MB), the abstract's count is kept.
• One-off command readurlist-arxiv-lengths fixes papers already saved; pauses 3 s between papers, as arXiv asks.
• New dependency: pypdf. No database change.

Checks: uv run pytest -q 136 passed. Local database, real link arxiv.org/abs/1706.03762: abstract 346 words → stored 6,111 (Long · 27 min). One-off fix: reset to 346 → back to 6,111.

Deploy (owner): merge → .\scripts\live-worker-update.ps1 (its uv sync installs pypdf) → in C:\apps\ReadURList: .venv\Scripts\readurlist-arxiv-lengths.exe → check an arXiv paper on All shows Long.

The map still needs stage 11's check-log row (owner's explain-back) before map-check says ready.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
read-ur-list	Ready	Preview	Nov 1, 2026 5:26am UTC