perf(control): finish streaming and virtualization overhaul

[Th0rgal]

Summary

• finish the remaining /control perf overhaul items: chat/thought virtualization, slice stores, worker replay reducer, inactive delta summarization, WS streaming, negotiated text_op deltas, canonical assistant rows, and @Perf Playwright coverage
• document the updated streaming contract and final PERF_OVERHAUL status/measurements
• keep fresher Grok workspace OAuth tokens instead of overwriting them with stale host auth

Local validation

• cargo fmt --all
• cargo check -q
• cargo test -q text_op_stream_transform --lib
• cargo test -q inactive_stream_summary --lib
• cargo test -q finalized_text_ops_collapse_to_canonical_assistant_row --lib
• cd dashboard && npx tsc --noEmit
• cd dashboard && bun run build
• cd dashboard && PLAYWRIGHT_PORT=3001 PLAYWRIGHT_BASE_URL=http://localhost:3001 bunx playwright test tests/control-perf.spec.ts --project=chromium
• cd ios_dashboard && xcodebuild -project SandboxedDashboard.xcodeproj -scheme SandboxedDashboard -destination 'platform=iOS Simulator,name=iPhone 17 Pro,OS=26.4' build

Deploy / iOS smoke

• redeployed sandboxed-sh-dev after backend changes and verified /api/health
• simulator rendered historical goal fixture and connected SSE state against dev backend

---

Note

Medium Risk
Medium risk because it changes the /control streaming transport contract (adds WebSocket + negotiated text_op) and introduces new client-side execution paths (virtualized lists and a Web Worker) that could affect real-time rendering and history replay behavior.

Overview
Completes the remaining /control perf overhaul items by virtualizing the transcript/thoughts UI (via @tanstack/react-virtual) and offloading large history replays to a new events-worker.ts Web Worker with a synchronous fallback.

Updates the streaming contract documentation to include a WebSocket transport (/api/control/ws) and capability-negotiated text_op deltas plus finalized canonical assistant rows, and refreshes the perf log/measurements and local validation commands.

Adds perf regression coverage by introducing an @perf Playwright test (control-perf.spec.ts) that loads a large fixture mission and asserts DOM/heap/longtask budgets, and makes Playwright’s baseURL/port configurable via env vars.

^{Reviewed by Cursor Bugbot for commit e8cbb30. Bugbot is set up for automated code reviews on this repo. Configure here.}

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
sandboxed-dashboard	Ready	Preview, Comment	May 19, 2026 9:46am
sandboxed-sh	Ready	Preview, Comment	May 19, 2026 9:46am

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit e8cbb30. Configure here.}

[cursor]

Playwright port/baseURL mismatch when only one env var set

Medium Severity

When PLAYWRIGHT_BASE_URL is set without PLAYWRIGHT_PORT (e.g. PLAYWRIGHT_BASE_URL=http://localhost:3001), port defaults to '3099', so webServer.command starts Next on 3099 while webServer.url polls 3001. The test run will hang waiting for a server that never appears on the expected port. Deriving port from the URL, or requiring both vars together, would fix this.

 

^{Reviewed by Cursor Bugbot for commit e8cbb30. Configure here.}