fix(web): clean up the provider settings list and editor

[rynfar]

Problem

The Providers settings page carried a "Provider / On" header row, a bottom refresh bar, 11px status text that overflowed, a fixed 64px row height, status dots on every row regardless of health, and an Advanced section that could not even be opened on a read-only session.

Fix

Manual port of upstream #8504 (5e63aea2d) from the f94a0d646..1f8ed54ad T3 review. PR #172 had already taken its connected-dot hunk; this brings the rest: header row and bottom bar removed, "Checked Xs ago" plus refresh and add-provider as labelled icon buttons in the section header, Configuration tab first, 13px two-line-clamped status text with the probe error inline (and unclamped in the editor), rounded bg-foreground/8 selection, status dots only for warning and error, Advanced openable read-only, one 16px gutter.

ProviderSettingsPanel.tsx is byte-identical to upstream's ported region. ProviderInstanceCard.tsx was re-implemented on Pylon's card rather than merged, because Pylon's version carries drain-order chevrons, environmentId-driven in-app sign-in, and the redacted "Authenticated as" header with reveal — all of which survive and are tested.

Three deliberate deviations, each commented in source:

• min-h-19 instead of upstream's fixed h-19, because Pylon rows can also carry ProviderUsageSummary, which a hard height would clip.
• The Sign in button gets its own inert fence. Upstream leaves the whole status line outside the read-only fence because its status line holds no write action; Pylon's does, and a literal port would have made a write clickable on a read-only session.
• Pylon's motion-reduce-aware animated update marker is kept over upstream's new static advisory icon.

Dropping width="expanded" narrows the page from max-w-6xl to max-w-4xl, as upstream did.

Verification

• vp test run on ProviderInstanceCard.test.ts, ProviderSettingsPanel.environment.test.tsx, SettingsPanels.logic.test.ts — 37 passed initially, then a review round added read-only coverage: the drain-order chevrons and Sign in each sit inside an inert fence while the email reveal does not; the Advanced trigger is outside the panel's fence; chevron disabled state is pinned to handler presence rather than list edges. Every new assertion was mutation-checked (each of four source mutations fails exactly one case).
• @t3tools/web typecheck, lint, format, and git diff --check clean.
• Independent review traced every interactive control for read-only leakage and found none; confirmed both header icon buttons carry aria-labels and the error state stays textual with the dots removed; confirmed ProviderSettingsPanel has a single caller and no prop change.
• Not done: a visual pass. Highest-value things to look at: light/dark selection tint, a long probe error wrapping to two lines, a row with both status and usage summary, the read-only editor header, and whether the narrower page still fits the 20rem list beside the editor.

---

Claude Fable 5 in Pylon.

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.5 KiB	+194 B (+1.4%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−7 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.6 KiB	+201 B (+3.1%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	57.0 KiB	+1.5 KiB (+2.7%)	66.4 KiB	✅
Codex	Live turn messages	9	11	+2 (+22.2%)	21	✅
Claude	Total thread wire	13.3 KiB	13.3 KiB	−68 B (−0.5%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−9 B (−0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.4 KiB	−59 B (−0.9%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	56.3 KiB	−44 B (−0.1%)	66.4 KiB	✅
Claude	Live turn messages	10	9	−1 (−10.0%)	21	✅

Baseline: abe70a7 · PR result: 16d0c75 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.