perf(usage): the Usage page shows numbers in under a second and dims only what is still loading

[t3dotgg]

The Usage page waited on every source before it showed anything. Cursor's account API alone took 19–32s, and the only sign of loading was a tiny spinner, so it was hard to tell the numbers were stale.

Takeover of #16376 by @bmdavis419, rebased onto current main. His work is imported as one commit with his authorship. My fixes are in a separate commit.

What it does (from #16376)

• Cursor account records are cached per login for 90+ days. The server answers from that cache right away and marks Cursor refreshing. The client then sends one follow-up with awaitRefresh to get the fresh Cursor numbers. Older servers ignore the flag.
• Transcript, OpenCode, Antigravity and Cursor sources scan at the same time. The scan cache is written after the response.
• Only the figures still loading are dimmed: a provider's row, its models, and the totals. Switching period keeps the previous period on screen, dimmed, until the new one answers.
• Shared loading rules live in client-runtime. Mobile gets the server speedups and keeps its current loading state.

Revisits drop from ~19s to 60–80ms. After a server restart, the page answers in ~120ms from the Cursor cache.

What I changed

• Resolved conflicts with the Antigravity scan cache (#17139) and the provider filter (#16970).
• Dropped perf(usage): the usage page loads in a fraction of a second and dims only what is still updating #16376's parallel Cursor pager. It kept requesting pages after a failure (open High finding). #17140, now on main, speeds up paging with a smaller change.
• A hidden provider that is still refreshing no longer adds its row or dims the totals.
• The period label names the period on screen while the previous one is kept.
• The kept-period test now really switches periods.

Screenshots (from #16376)

Before: a small spinner after "All environments" was the only sign of loading.

Cursor still loading: only Cursor's row and models, plus the totals, are dimmed.

Every figure updating, e.g. while a new period loads:

Verification

Usage tests (248) pass. Typecheck passes in server, web, mobile, client-runtime and contracts. I did not run the app in a browser for this rebase.

Original work by @bmdavis419 with Claude Opus 5.5 in T3 Code via Claude Code. Takeover by Claude Opus 5.5 in T3 Code via Claude Code.

🤖 Generated with Claude Code

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This is a large production behavior change introducing persistent Cursor account caching, background refresh coordination, concurrent scanning, shared query-runtime changes, and new stale/provisional Usage-page rendering. Its broad runtime and data-persistence impact goes beyond a self-contained performance fix and warrants human review.

Notes:

• No code objects were reviewed. Approvability was decided on eligibility alone.

No code changes detected at 23a4c36. Prior analysis still applies.

^{You can add or adjust custom eligibility rules. Learn more.}

[coderabbitai]

✨ Finishing Touches

📝 Generate docstrings

• Commit to this branch
• Create a new PR

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

• Autopilot · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

---

_{Comment @coderabbitai help to get the list of available commands.}

[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	5.0 KiB	5.0 KiB	0 B (0.0%)	6.8 KiB	✅
Codex	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Codex	Live turn WebSocket wire	1.2 KiB	1.2 KiB	0 B (0.0%)	2.0 KiB	✅
Codex	Live turn WebSocket decoded	20.9 KiB	20.9 KiB	0 B (0.0%)	29.3 KiB	✅
Codex	Live turn messages	2	2	0 (0.0%)	8	✅
Claude	Total thread wire	5.0 KiB	5.0 KiB	0 B (0.0%)	6.8 KiB	✅
Claude	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Claude	Live turn WebSocket wire	1.2 KiB	1.2 KiB	0 B (0.0%)	2.0 KiB	✅
Claude	Live turn WebSocket decoded	21.2 KiB	21.2 KiB	0 B (0.0%)	29.3 KiB	✅
Claude	Live turn messages	2	2	0 (0.0%)	8	✅

Baseline: 0ce6b9d · PR result: 23a4c36 · 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: 108.5 KiB
• Claude decoded thread snapshot: 108.8 KiB

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