feature: dashboard paper filter, read trend + digest arXiv refs

[Jammy2211]

Overview

Follow-up polish to #42/PR#43, human-approved from the ship-summary suggestions: make the 231-paper queue navigable at speed (filter box), make new queue entries link straight to their arXiv abstract page (digest refs), and surface whether the queue is draining (read trend).

Plan

• Filter box: self-contained JS <input> above the reading queue in the html render; live-filters paper rows across sections, auto-expands sections with matches, restores collapsed state when cleared.
• Digest arXiv refs (PyAutoMind, direct-to-main like the original pipeline task): the nightly digest's paste-ready append block writes Title — <arXiv id> so the dashboard's existing — <ref> parse links each new paper to its abstract page.
• Read trend: from the DONE <date> history — "N read last 7 / M last 30 days" in the Reading queue header plus a small 8-week bar strip, computed relative to the snapshot generated timestamp (pure, testable).

Detailed implementation plan

Affected Repositories

• PyAutoMemory (primary — scripts/board.py, tests/test_board.py)
• PyAutoMind (.github/workflows/arxiv_papers.yml — direct main commit)

Branch Survey

Repository	Current Branch	Dirty?
./PyAutoMemory	main (synced)	clean
./PyAutoMind	main (synced)	clean

Suggested branch: feature/dashboard-filter-trend
Worktree: ~/Code/PyAutoLabs-wt/dashboard-filter-trend/

Implementation Steps

1. board.py html render: <input id="pfilter" type="search"> + flt() (vanilla JS, ES5) filtering ul.papers li text across details.qsec; sections with no match hide; non-empty query opens matching sections, empty query restores.
2. board.py: pure helper _read_trend(snapshot) parsing done_date across queue papers against snapshot["generated"] → counts for 7/30 days + 8 weekly buckets; render counts in the Reading queue <h2> meta and a mini bar strip (existing .bar/.seg styling pattern); mirror a one-line count in the md render.
3. tests/test_board.py: fixture gains DONE lines near a pinned generated date; pin the trend counts, the filter markup, and that badge/strip stay byte-identical.
4. PyAutoMind arxiv_papers.yml: append-block format becomes Title — <arXiv id>; YAML validated before push; digest still never commits.

Testing

pytest tests/ + make validate in PyAutoMemory; render --html and check filter behaviour by hand; YAML parse-check the Mind workflow.

Original Prompt

Click to expand starting prompt

See PyAutoMind/draft/feature/pyautomemory/dashboard_filter_trend_digest_refs.md (filed 2026-08-21; human approval: "ok do all 4" on the ship-summary suggestion list).