From f7a5e3272bf181ecd8a85c8e8ca9fb7a5e6f8893 Mon Sep 17 00:00:00 2001 From: d-robotics Date: Wed, 28 Oct 2026 00:49:09 +0800 Subject: [PATCH] =?UTF-8?q?=F0=9F=97=82=EF=B8=8F=20Add=20task=20card=20for?= =?UTF-8?q?=20remaining=20work-buffer=20poll=20cost?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Co-Authored-By: Claude Opus 5.5 --- .scratch/workbench-poll-cost/work.md | 8 ++++++++ 1 file changed, 8 insertions(+) create mode 100644 .scratch/workbench-poll-cost/work.md diff --git a/.scratch/workbench-poll-cost/work.md b/.scratch/workbench-poll-cost/work.md new file mode 100644 index 00000000..3cbc7318 --- /dev/null +++ b/.scratch/workbench-poll-cost/work.md @@ -0,0 +1,8 @@ +# Cut the remaining work-buffer poll cost and keep dashboard caches warm + +Goal: after #63, the 3s work-buffer poll still spent ~126ms re-resolving display labels (full run documents, up to 6MB) and ~136ms in the store; dashboard caches were flushed by every autosave +Agent: Claude Code (Opus 5.5), desktop session 2026-10-28 +Base: origin/dev @ 8b96632e +Scope: internal/api/annotation_work_buffer.go (2-minute display cache), internal/store/annotation_work_buffer.go (skip rereads for slots the active scan already counted), internal/api/server.go (draft, review-session, review-draft and work-buffer saves keep dashboard caches), tests +Out of scope: run document size, other list endpoints' enrichment, poll interval, dashboard computations themselves +Risk: no API, data or index change; pool labels may lag a project/workflow rename by up to 2 minutes; dashboards may lag a draft save by up to the existing 45s TTL