[HugoXK]

Timeline: which tools each Bot used, and how often they failed

The timeline listed each tool call inside a turn but never summarized which tools a Bot relies on or how often they fail, so a flaky tool or connector was hard to notice. timeline/window now carries toolUse: per Bot lane, its tools as { name, calls, failed }, by call count (top 20), names cleaned and capped. Read-only, metadata only.

Ran at the tip: npm run verify (green, TZ=UTC), node scripts/lint-plugins.mjs timeline (0 errors). Not run: smoke, Desktop, real-model steward.

[Birfy]

Review (Birfy, 2026-10-20 20:07 Asia/Shanghai): PASS (code) — held for the owner's confirmation of #538

Tip SHA: c5e47a5
Ran:

• npm --prefix plugins/timeline test (TZ=UTC) → pass (94/94)
• node scripts/lint-plugins.mjs → pass
• CI tests-ok at this tip → green (all checks completed; no fork-approval needed)
  Not run:
• core suite / smoke / Desktop — plugin-only change, CI ran the full suite at this tip
  Verdict: PASS

Findings (non-blocking, for the owner's decision):

1. toolUse is derivable from turns[].tools (already carries name/ok) and the panel doesn't show it yet. For now it's payload on the 2 s poll only.
2. Tool names go through cleanLabel(…, 60) like turns[].tools. Metadata only, no args or output.

#538 is a new feature the owner has not confirmed yet, so this is not merged regardless of tier. It shares the same window.js return line / doc row as #531/#535/#537. This is not merge authority.

[Birfy]

@HugoXK Owner decision (2026-10-20 21:09 Asia/Shanghai): accepted. Required change: also show per-tool calls/failures in the dashboard/panel, not only in timeline/window.

[Birfy]

Waiting for approval. tests-ok is green at the tip and the PR is mergeable, so it is labeled awaiting-approval.

It needs (1) the owner's approval through the GitHub UI (a chat reply does not count) and (2) a review PASS at the exact tip SHA; any new commit voids both.

Automation does not merge T2/T3 by itself and does not approve on the owner's behalf.

[Birfy]

PASS — tip c5e47a50535afbb8f5694c80c0cd6239ca274a2c (Maintainer review, acting as Birfy)

Design check: no change to Bot authorization, to the approval path (still a UI button only) or to user-facing English text; the new surface is read-only and part of the existing answer.

Local evidence: I applied this PR's diff to my local main snapshot and ran the touched tests there (TZ=UTC), plus node scripts/ds-bot-patches.mjs check (with the new files staged, as a commit would):

• plugins/timeline: node --import ./test/resolve.mjs --test test/window.test.js → 22/22 pass
• node scripts/ds-bot-patches.mjs check → ok (pin v0.1.5, 10 patches, core == ds-bot/ + patches)

The new toolUse is read-only, names come through cleanLabel(name, 60), and the list is capped at 20 per Bot.

Not run locally: the full npm run verify in one shot (this box has a 2 GiB memory cap, so I run the steps one at a time), the smoke test, Desktop, and the real-model steward.

CI: tests-ok is green at the tip SHA above (GitHub Actions ran the full suite there).

Tier: T2/T3, so this PASS does not approve the PR: the owner still has to approve it through the GitHub UI, and any new commit voids both the approval and this PASS. Not merged by automation.

[Birfy]

Verdict: PASS at 918f8dd

Tip SHA: 918f8dd7752ab05b322a6609356946b7379ec7f7
The branch was rebased onto current main at 01:31Z (single commit, parent 4f6a53d7), so it is clean again — no rebuild was needed; the earlier PASS at c5e47a50 is void. Read the whole new diff: plugins/timeline/src/window.js (+18 -1, a `toolUse` list per Bot inside the existing timeline/window answer), test/window.test.js (+16), one docs/subsystems/timeline.md row, one changelog fragment.
Design: read-only and additive — `toolUse` is another field of the same answer, per Bot, counted from that Bot's non-room turns, names through cleanLabel(60), worst-first by calls then name, capped at 20 entries per Bot; no new endpoint, no config, no user-facing string, no approval path. Failed calls come from `tool.ok === false` only, so an unfinished call is not counted as a failure.
TDD: the new test reads `win.toolUse`, which does not exist at the base window.js — 0 pass / 1 fail with only window.js reverted, and it passes with the change.
Ran at the tip: CI tests-ok green at this SHA (test, plugins tree/social/timeline, package all success). On a copy of current main with this diff applied: plugins/timeline 96/96 (TZ=UTC), changelog.mjs check OK.
Not run: one-shot npm run verify (2 GiB OOM), core suite, smoke, Desktop.

Tier: T2. Awaiting the owner's click in the GitHub UI plus this PASS at the tip. Any new commit voids this PASS. Note: #531/#535/#539 all touch the same return line in window.js and the same docs row, so only one merges cleanly; after it lands, the other two need a rebase (and a fresh PASS at the new tip).

[Birfy]

Do not approve this tip — it no longer merges. #531 landed on main (03bfc468) and it changed the same line of plugins/timeline/src/window.js and the same docs/subsystems/timeline.md table row this branch edits, so the PR is now dirty against main: a local 3-way merge conflicts in window.js, window.test.js and timeline.md.

The PASS at 918f8dd7 stays valid for the content, but a PASS is only half the gate — the other half is a mergeable tip. Please merge main into the branch (or rebase) and push; I will re-review the new tip and post a fresh PASS there. Merging main in is the fix; no code change is needed.

[Birfy]

Verdict: PASS at 3523673

Reviewed again at this tip: rebased onto main 33fb55ed, so the previous PASS (918f8dd7) is void.
Diff (4 files, +36 -2): `plugins/timeline/src/window.js` adds `toolUse` (per-Bot tool counts by name and per-tool failures, cleaned codes, capped lists), the new test in `plugins/timeline/test/window.test.js`, the `timeline.md` row for `window.js`, and `changelog.d/timeline-tooluse.added.md`. The rebase only moved the docs row onto #531's wording; no behaviour change since the version I reviewed at 918f8dd7.
Ran at the tip: CI tests-ok green. On a copy of this PR's own base with the whole diff applied: `ds-bot-patches.mjs check` OK, `check-arch-docs.mjs` OK (4 files), `changelog.mjs check` OK, plugins/timeline 97/97, `scripts/lint-plugins.mjs timeline` 0 errors / 8 warnings, TZ=UTC.
TDD: with only `window.js` at the base commit, the new "toolUse" test fails (23/24); all 97 pass with the change.
Conflicts: with #535 and with #568 (same `timeline.md` row; #535 also shares the `window.js` return line and the test file). Clean with main, #566, #572, #574, #578, #580.

Not run: one-shot npm run verify (2 GiB OOM), full core suite, smoke, Desktop, a real model call.
Merge matrix (local 3-way merges against main 33fb55e and against every other open PR): see the lines above; every claim is measured, not assumed.
Tier: T2 (plugins/**, core/src/host/**, core/types/**). The owner's own click in the GitHub UI is the approval and this PASS must sit on the tip; I do not approve or merge it myself. Any new commit voids this PASS.

[Birfy]

One precision on the PASS above: I could not byte-diff the pre-rebase branch (the force-push removed the old commit), so "the only difference is the docs row" is my reading of the source hunks against my review notes for the old tip, not a diff I could run. What is verified at the new tip is everything else in the PASS: the full check set, the plugin test suite and the red-without-fix run were all redone here.

[Birfy]

Superseded by #586 (one combined timeline change, at main 33fb55ed).

The maintenance round merged this PR's change, its test and its docs wording into #586 together with #535 and #568, because all three edit the same window.js return line and the same timeline.md row: the owner then gets one PASS and one click instead of one merge plus two rebases. Your authorship is recorded in #586's body and in this thread; nothing of your work is dropped.

Closing this PR as superseded. Thank you for the original — the tests and the docs row came through unchanged.