diff --git a/AGENTS.md b/AGENTS.md index 027901f..e464a8f 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -100,7 +100,8 @@ WaggleDance routes in the sidecar and `team-sync` in core, with the team server `packages/team-server` (P1.1); the single-mind to multi-mind migration in core (P1.2); Waggle Room tasks for external coding agents (`POST /api/tools/run`, the headless task contract in tool manifests, run tokens and the hook settings a launch used to set) and the router's -external executors. The Tool Launcher keeps detect, launch, live output and stop; the Room canvas +external executors; the Benchmarks page (`/benchmarks`), which showed LoCoMo scores of the old +memory engine. The Tool Launcher keeps detect, launch, live output and stop; the Room canvas (`/room`) still shows Waggle agents, groups and launched tools. The install audit trail moved from core to the server (`packages/server/src/local/install-audit.ts`, P1.2); other packages write to it through `InstallAuditRecorder` from `@waggle/shared`. diff --git a/CHANGELOG.md b/CHANGELOG.md index e908596..43565ab 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -31,6 +31,8 @@ lists what changed for users compared with that product. The reasons are recorde Launcher still detects these tools, opens them in the active workspace, shows their live output and stops them. "Best fit" in chat now proposes Waggle personas only, so no workspace memories are sent to an external agent. +- **The Benchmarks page** (`/benchmarks`, opened from the command palette). It showed LoCoMo + scores of the old memory engine, which this edition does not run. - **`install.sh`** and `scripts/waggle-server.sh`, replaced by the server installer. - The automatic start of a LiteLLM proxy, which needed Python. It is now opt-in (see Changed). diff --git a/apps/web/src/App.tsx b/apps/web/src/App.tsx index a4db485..b5ad724 100644 --- a/apps/web/src/App.tsx +++ b/apps/web/src/App.tsx @@ -34,7 +34,6 @@ const MissionControlRoute = lazy(() => import("@/routes/MissionControlRoute")); const TimelineRoute = lazy(() => import("@/routes/TimelineRoute")); const EventsRoute = lazy(() => import("@/routes/EventsRoute")); const UsageRoute = lazy(() => import("@/routes/UsageRoute")); -const BenchmarkRoute = lazy(() => import("@/routes/BenchmarkRoute")); const PlatformRoute = lazy(() => import("@/routes/PlatformRoute")); const WorkspacesRoute = lazy(() => import("@/routes/WorkspacesRoute")); const AuthRoute = lazy(() => import("@/routes/AuthRoute")); @@ -157,8 +156,7 @@ const App = () => ( )} /> )} /> )} /> - {/* ── PR6a: ⌘K-only static surfaces ── */} - )} /> + {/* ── PR6a: ⌘K-only static surface ── */} )} /> {/* ADD ALL CUSTOM ROUTES ABOVE THE CATCH-ALL "*" ROUTE */} } /> diff --git a/apps/web/src/components/os/apps/BenchmarkApp.test.tsx b/apps/web/src/components/os/apps/BenchmarkApp.test.tsx deleted file mode 100644 index cc7e955..0000000 --- a/apps/web/src/components/os/apps/BenchmarkApp.test.tsx +++ /dev/null @@ -1,73 +0,0 @@ -/** - * BenchmarkApp — screen 17. Pure-UI static showcase. The test renders the - * component, asserts the Capabilities view's framing + matrix, toggles to the - * Memory SOTA view, and asserts the LoCoMo numbers + stat chips. No mocks - * needed — the component has zero props and makes no calls. - */ -import { describe, it, expect, afterEach } from 'vitest'; -import { render, screen, cleanup, fireEvent, within } from '@testing-library/react'; - -import BenchmarkApp from './BenchmarkApp'; - -afterEach(() => cleanup()); - -describe('BenchmarkApp', () => { - it('renders the Capabilities view by default with the positioning framing', () => { - render(); - // Tablist toggle present, Capabilities tab selected. - expect(screen.getByRole('tab', { name: /vs competitors/i })).toHaveAttribute('aria-selected', 'true'); - expect(screen.getByRole('tab', { name: /memory sota/i })).toHaveAttribute('aria-selected', 'false'); - // Headline + the two framing cards. - expect(screen.getByText(/waggle remembers you/i)).toBeInTheDocument(); - expect(screen.getByRole('heading', { name: /task agents/i })).toBeInTheDocument(); - expect(screen.getByRole('heading', { name: /a memory layer \+ workspace/i })).toBeInTheDocument(); - // Honest disclaimer label. - expect(screen.getByText(/positioning view, not a lab benchmark/i)).toBeInTheDocument(); - }); - - it('renders the full 11-row capability matrix with the deliberate competitor win on deep coding', () => { - render(); - // Waggle column header + first/last capability rows. - expect(screen.getByRole('columnheader', { name: /^waggle$/i })).toBeInTheDocument(); - expect(screen.getByRole('rowheader', { name: /persistent memory across sessions/i })).toBeInTheDocument(); - const lastRow = screen.getByRole('rowheader', { name: /deep coding in the terminal/i }); - expect(lastRow).toBeInTheDocument(); - // 11 capability rows = 11 row headers. - expect(screen.getAllByRole('rowheader')).toHaveLength(11); - // The last row is honest: Waggle is only Partial on deep terminal coding. - const row = lastRow.closest('tr'); - expect(row).not.toBeNull(); - const cells = within(row as HTMLElement).getAllByRole('cell'); - expect(cells[0]).toHaveAttribute('aria-label', 'Partial'); // Waggle column - expect(cells[1]).toHaveAttribute('aria-label', 'Yes'); // Claude Code column - }); - - it('toggles to the Memory SOTA view and shows the LoCoMo numbers + stat chips', () => { - render(); - fireEvent.click(screen.getByRole('tab', { name: /memory sota/i })); - - // View switched. - expect(screen.getByRole('tab', { name: /memory sota/i })).toHaveAttribute('aria-selected', 'true'); - expect(screen.queryByRole('rowheader', { name: /persistent memory across sessions/i })).not.toBeInTheDocument(); - - // SOTA headline + the four LoCoMo bar values. - expect(screen.getByText(/best long-term memory/i)).toBeInTheDocument(); - expect(screen.getByText('86.49')).toBeInTheDocument(); - expect(screen.getByText('81.95')).toBeInTheDocument(); - expect(screen.getByText('78.05')).toBeInTheDocument(); - expect(screen.getByText('62.47')).toBeInTheDocument(); - - // Stat chips. - expect(screen.getByText('+4.54')).toBeInTheDocument(); - expect(screen.getByText('92.27%')).toBeInTheDocument(); - expect(screen.getByText('100%')).toBeInTheDocument(); - }); - - it('toggles back to Capabilities', () => { - render(); - fireEvent.click(screen.getByRole('tab', { name: /memory sota/i })); - fireEvent.click(screen.getByRole('tab', { name: /vs competitors/i })); - expect(screen.getByRole('tab', { name: /vs competitors/i })).toHaveAttribute('aria-selected', 'true'); - expect(screen.getByRole('rowheader', { name: /persistent memory across sessions/i })).toBeInTheDocument(); - }); -}); diff --git a/apps/web/src/components/os/apps/BenchmarkApp.tsx b/apps/web/src/components/os/apps/BenchmarkApp.tsx deleted file mode 100644 index 71a8b2b..0000000 --- a/apps/web/src/components/os/apps/BenchmarkApp.tsx +++ /dev/null @@ -1,409 +0,0 @@ -/** - * BenchmarkApp — Warm-Hive Benchmarks surface (screen 17). A pure-UI, ⌘K-only - * showcase with two views behind a segmented toggle: - * - * 1) Capabilities — honest positioning vs the field (task agents vs Waggle's - * memory layer), an 11-row capability matrix with the Waggle column - * honey-highlighted. Deliberately credits competitors on "deep terminal - * coding" — it's a positioning view, not a lab benchmark. - * 2) Memory SOTA — the one real head-to-head lab result: LoCoMo bars + stats. - * - * Static data only — no backend, no API, no props (default-export, zero-prop). - * Source fidelity: docs/design_handoff_waggle_app/design-files/screens/benchmark.html - */ -import { useState } from 'react'; -import { BarChart3, Info } from 'lucide-react'; - -// ─────────────────────────────────────────────────────────────────────────── -// Benchmark figures — manually refreshed from the hive-mind benchmark, -// 2026-06-18 — do not fabricate. Numbers MUST match benchmark.html exactly. -// ─────────────────────────────────────────────────────────────────────────── - -/** A capability mark: yes (●) / partial (◐) / no (○). */ -type Mark = 'Y' | 'P' | 'N'; - -/** Competitor columns, in matrix order after the honey Waggle column. */ -const COMPETITORS = [ - { name: 'Claude Code', sub: 'CLI' }, - { name: 'Codex', sub: 'CLI' }, - { name: 'Cowork', sub: 'Anthropic' }, - { name: 'Hermes', sub: 'agent' }, - { name: 'Odysseus', sub: 'agent' }, -] as const; - -/** 11-row capability matrix. marks order: Waggle, then the 5 COMPETITORS. */ -const CAPABILITY_MATRIX: ReadonlyArray<{ capability: string; marks: readonly Mark[] }> = [ - { capability: 'Persistent memory across sessions', marks: ['Y', 'P', 'N', 'P', 'P', 'P'] }, - { capability: 'Local-first & private by default', marks: ['Y', 'N', 'N', 'N', 'P', 'P'] }, - { capability: 'Any model (incl. small local models)', marks: ['Y', 'N', 'N', 'N', 'P', 'P'] }, - { capability: 'Built for non-technical experts', marks: ['Y', 'N', 'N', 'P', 'N', 'N'] }, - { capability: 'Runs your existing coding agents', marks: ['Y', 'N', 'N', 'N', 'N', 'N'] }, - { capability: 'Self-evolving skills', marks: ['Y', 'N', 'N', 'N', 'P', 'N'] }, - { capability: 'Skills shared across agents/workspaces', marks: ['Y', 'P', 'P', 'N', 'P', 'N'] }, - { capability: 'Multi-agent swarm', marks: ['Y', 'P', 'N', 'P', 'P', 'Y'] }, - { capability: 'Native desktop app · Windows, macOS, Linux', marks: ['Y', 'N', 'N', 'P', 'N', 'N'] }, - { capability: 'Audit-ready / EU AI Act', marks: ['Y', 'P', 'P', 'P', 'N', 'N'] }, - { capability: 'Deep coding in the terminal', marks: ['P', 'Y', 'Y', 'P', 'P', 'Y'] }, -]; - -/** LoCoMo memory bars — value drives both the label and the bar width. */ -const MEMORY_BARS: ReadonlyArray<{ name: string; note: string; value: number; us?: boolean }> = [ - { name: 'Waggle · Hive Mind', note: 'ours · local', value: 86.49, us: true }, - { name: 'Memori', note: 'prev. SOTA', value: 81.95 }, - { name: 'LangMem', note: 'corrected', value: 78.05 }, - { name: 'Mem0', note: 'baseline', value: 62.47 }, -]; - -/** Stat chips beneath the bars. */ -const MEMORY_STATS: ReadonlyArray<{ value: string; label: React.ReactNode }> = [ - { value: '+4.54', label: <>points over the prior best · z = 4.64, p < 10⁻⁵ }, - { value: '92.27%', label: <>single-hop recall, ~1pt off the full-context ceiling }, - { value: '100%', label: <>local: warm recalls in 58–83 ms, on-device }, -]; - -// ─────────────────────────────────────────────────────────────────────────── - -type View = 'caps' | 'memory'; - -const VIEW_LABELS: Record = { - caps: <>Capabilities: where Waggle sits vs the field, - memory: <>Memory SOTA: the one head-to-head lab result, -}; - -const MARK_GLYPH: Record = { Y: '●', P: '◐', N: '○' }; - -/** Colour a mark by kind. The Waggle (first) column reads honey. */ -function markClass(mark: Mark, isUs: boolean): string { - if (isUs && mark === 'Y') return 'font-bold'; - if (mark === 'Y') return ''; - if (mark === 'P') return ''; - return 'opacity-60'; -} - -function markStyle(mark: Mark, isUs: boolean): React.CSSProperties { - if (isUs) return { color: 'var(--honey)' }; - if (mark === 'Y') return { color: 'var(--healthy)' }; - if (mark === 'P') return { color: 'var(--attention)' }; - return { color: 'var(--text-2)' }; -} - -/** One disclaimer/caveat block — shared between the two views. */ -function Disclaimer({ children }: { children: React.ReactNode }) { - return ( -
- -

{children}

-
- ); -} - -const BenchmarkApp = () => { - const [view, setView] = useState('caps'); - - return ( -
- {/* Controls bar — segmented toggle + live view label */} -
- - - Benchmarks · view - -
- {(['caps', 'memory'] as View[]).map(v => ( - - ))} -
- {VIEW_LABELS[view]} -
- - {/* Stage */} -
-
- {view === 'caps' ? : } -
-
-
- ); -}; - -function CapabilitiesView() { - return ( -
-
-
-
-

- They automate tasks. Waggle remembers you. -

-

- The strong agents today are terminal coding tools for engineers. - Waggle plays a different game: a persistent, local-first memory layer for - knowledge workers that even runs those agents inside it. Here's the - honest lay of the land. -

-
- - {/* Two framing cards */} -
-
-
The field
-

Task agents

-

- Powerful, mostly terminal-based, model-locked, and built for developers. Each session starts fresh; the - intelligence lives in the model, not in a memory of you. -

-
- Claude Code · Codex · Claude Cowork · Hermes · Odysseus -
-
- -
-
- Our category -
-

- A memory layer + workspace -

-

- Knows you and your work across every session, runs on any model (even local), is built for non-technical - experts, and can launch the task agents into its shared memory. -

-
- Waggle: complements them, doesn't compete head-on -
-
-
- - {/* Capability matrix */} -
- - - - - - - {COMPETITORS.map(c => ( - - ))} - - - - {CAPABILITY_MATRIX.map((row, rowIdx) => { - const isLast = rowIdx === CAPABILITY_MATRIX.length - 1; - return ( - - - {row.marks.map((mark, colIdx) => { - const isUs = colIdx === 0; - return ( - - ); - })} - - ); - })} - -
Capability comparison: Waggle versus task agents
- Capability - - Waggle - - {c.name} - {c.sub} -
- {row.capability} - - - {MARK_GLYPH[mark]} - -
-
- - {/* Legend */} -
- - ●Yes - - - ◐Partial / limited - - - ○No / not the focus - -
- - - This is a positioning view, not a lab benchmark: capabilities as - understood June 2026, focused on the dimensions Waggle is built around. Coding agents are genuinely excellent at - deep code work (the last row), which is exactly why Waggle launches them{' '} - rather than replacing them. The one head-to-head lab result is the memory benchmark → see “Memory SOTA”. - -
- ); -} - -function MemorySotaView() { - const maxValue = Math.max(...MEMORY_BARS.map(b => b.value)); - - return ( -
-
-
-
-

- The best long-term memory on record. -

-

- On LoCoMo (the standard test for long-term conversational memory), - Waggle's open-source substrate scores 86.49%, a new state of the - art, measured under the prior leader's own protocol and judge. -

-
- - {/* LoCoMo bars */} -
- {MEMORY_BARS.map(bar => ( -
-
- {bar.name} - {bar.note} -
-
-
-
-
- {bar.value} -
-
- ))} -
- - {/* Stat chips */} -
- {MEMORY_STATS.map((stat, i) => ( -
-
- {stat.value} -
-
{stat.label}
-
- ))} -
- - - LoCoMo, N = 1,540, GPT-4.1-mini as answerer & judge: the prior - SOTA's exact published protocol, reproduced in-harness to 0.03 points before comparison. The intelligence lives in - the memory layer, not the model, so it travels onto a small local - model too. Honest caveat: higher token use per question than the leanest systems; efficiency work underway. - -
- ); -} - -export default BenchmarkApp; diff --git a/apps/web/src/components/os/apps/RoomApp.tsx b/apps/web/src/components/os/apps/RoomApp.tsx index f560bf6..738f58f 100644 --- a/apps/web/src/components/os/apps/RoomApp.tsx +++ b/apps/web/src/components/os/apps/RoomApp.tsx @@ -40,7 +40,7 @@ interface RoomAppProps { // Warm-token status styling (D21): the live SSE roster maps to semantic vars, // not raw emerald/sky/violet. Inline style is the project convention for the -// --work/--intel/--healthy/--risk semantic palette (see BenchmarkApp). +// --work/--intel/--healthy/--risk semantic palette. const STATUS_DOT_STYLE: Record = { pending: { background: 'var(--text-dim)' }, running: { background: 'var(--honey)' }, diff --git a/apps/web/src/lib/command-catalog.ts b/apps/web/src/lib/command-catalog.ts index 619b6e8..7c14fe4 100644 --- a/apps/web/src/lib/command-catalog.ts +++ b/apps/web/src/lib/command-catalog.ts @@ -4,7 +4,7 @@ import { Plus, Rocket, Settings, Server, Plug, Store, Package, Shield, Clock, FolderOpen, Lock, Activity, History, BarChart3, Users, Radio, - Gauge, Monitor, LayoutGrid, Eye, + Monitor, LayoutGrid, Eye, } from "lucide-react"; /** @@ -85,7 +85,6 @@ export function buildCommandCatalog(ctx: CatalogContext): CatalogGroup[] { { id: "timeline", group: "power", name: "Timeline", subtitle: "activity history", icon: History, to: "/settings/timeline" }, { id: "events", group: "power", name: "Events & logs", subtitle: "live activity feed", icon: Radio, to: "/settings/events" }, { id: "usage", group: "power", name: "Usage & cost", subtitle: "tokens · spend", icon: BarChart3, to: "/settings/usage" }, - { id: "benchmarks", group: "power", name: "Benchmarks", subtitle: "capabilities · memory SOTA", icon: Gauge, to: "/benchmarks" }, { id: "platform", group: "power", name: "Platform & roadmap", subtitle: "desktop · what's coming", icon: Monitor, to: "/platform" }, ]; diff --git a/apps/web/src/lib/motion/route-transition.ts b/apps/web/src/lib/motion/route-transition.ts index ff14824..5ea1f6f 100644 --- a/apps/web/src/lib/motion/route-transition.ts +++ b/apps/web/src/lib/motion/route-transition.ts @@ -45,7 +45,6 @@ const ROUTE_LABELS: Record = { marketplace: 'Marketplace', launcher: 'Launcher', settings: 'Settings', - benchmarks: 'Benchmarks', platform: 'Platform', }; diff --git a/apps/web/src/routes/BenchmarkRoute.tsx b/apps/web/src/routes/BenchmarkRoute.tsx deleted file mode 100644 index bee0ea2..0000000 --- a/apps/web/src/routes/BenchmarkRoute.tsx +++ /dev/null @@ -1,11 +0,0 @@ -/** PR6a route wrapper — `/benchmarks` → BenchmarkApp (static, ⌘K-only surface). */ -import BenchmarkApp from '@/components/os/apps/BenchmarkApp'; -import SurfaceBoundary from './SurfaceBoundary'; - -const BenchmarkRoute = () => ( - - - -); - -export default BenchmarkRoute; diff --git a/apps/web/src/routes/index.ts b/apps/web/src/routes/index.ts index 80cd683..f4d0f12 100644 --- a/apps/web/src/routes/index.ts +++ b/apps/web/src/routes/index.ts @@ -49,8 +49,7 @@ export { default as MissionControlRoute } from './MissionControlRoute'; export { default as TimelineRoute } from './TimelineRoute'; export { default as EventsRoute } from './EventsRoute'; export { default as UsageRoute } from './UsageRoute'; -// PR6a — ⌘K-only static surfaces (Benchmarks · Platform) -export { default as BenchmarkRoute } from './BenchmarkRoute'; +// PR6a — ⌘K-only static surface (Platform) export { default as PlatformRoute } from './PlatformRoute'; export { default as WorkspacesRoute } from './WorkspacesRoute'; // PR7b — pre-shell /auth route (sibling outside AppShell) diff --git a/scripts/ux-gates/color-guard-baseline.json b/scripts/ux-gates/color-guard-baseline.json index 7d85f35..1f9e6ec 100644 --- a/scripts/ux-gates/color-guard-baseline.json +++ b/scripts/ux-gates/color-guard-baseline.json @@ -87,12 +87,6 @@ "snippet": "text-muted-foreground/30", "kind": "low-opacity-token" }, - { - "file": "apps/web/src/components/os/apps/BenchmarkApp.tsx", - "line": 129, - "kind": "inline-hex", - "snippet": "color:#1a1407" - }, { "file": "apps/web/src/components/os/apps/CapabilitiesApp.tsx", "line": 497, diff --git a/tests/e2e/runtime-a11y.spec.ts b/tests/e2e/runtime-a11y.spec.ts index ef94b82..2020b93 100644 --- a/tests/e2e/runtime-a11y.spec.ts +++ b/tests/e2e/runtime-a11y.spec.ts @@ -42,7 +42,6 @@ const ROUTES = [ '/files', '/approvals', '/artifacts', - '/benchmarks', '/platform', ] as const; diff --git a/tests/e2e/user-journeys.spec.ts b/tests/e2e/user-journeys.spec.ts index 0085750..bc615a5 100644 --- a/tests/e2e/user-journeys.spec.ts +++ b/tests/e2e/user-journeys.spec.ts @@ -842,14 +842,11 @@ test.describe('User Journey Tests', () => { }); test('J-route-coverage: thin utility routes render or redirect clearly', async ({ page }) => { - await gotoApp(page, '/benchmarks'); - await expect(page.locator('body')).toContainText(/benchmark|capability|score|memory/i); - await gotoApp(page, '/platform'); await expect(page.locator('body')).toContainText(/platform|local|governance|memory|agent/i); - // Payment pages are gone in the community edition (ADR-0001). - for (const route of ['/payment-cancelled', '/payment-success']) { + // Payment pages (ADR-0001) and the Benchmarks page are gone in the community edition. + for (const route of ['/payment-cancelled', '/payment-success', '/benchmarks']) { await gotoApp(page, route); await expect(page.locator('body')).toContainText(/page you.re after doesn.t exist/i); }