LearnNewsExamplesServices
Frontmatter
id14560
titleEpic: Fleet Manager cockpit UI/UX — the design-led product surface (Lane B of the cockpit plan)
stateOpen
labels
enhancementepicdeveloper-experienceai
assigneesneo-opus-vega
createdAtJul 4, 2026, 2:26 AM
updatedAtJul 10, 2026, 10:17 PM
githubUrlhttps://github.com/neomjs/neo/issues/14560
authorneo-opus-vega
commentsCount12
parentIssuenull
subIssues
14577 FM cockpit target-app decision: evolve apps/agentos vs fresh app
14578 FM cockpit theme/token layer from the cockpit-plan design SSOT
14592 FM cockpit density evidence spike: real fleet-scale numbers before layout freezes
14593 FM cockpit state-dot, pulse and health primitives (reduced-motion honored)
14594 FM cockpit event-kind chip system (pr / a2a / review / alert classes)
14598 FM cockpit AgentCard component — the resident card at the artifact bar
14599 FM cockpit fleet grid + health summary bar (density-informed layout)
14605 FM cockpit card contract: consolidate the fleet-card definition under the design SSOT
14606 FM cockpit ActivityStream component: bounded, real-time, backpressure-aware
14607 FM cockpit stream burst e2e: bound holds live under real feed pressure
14608 FM cockpit agent detail view: drill-in panes with freshness-labeled claims
14610 FM cockpit pop-out: agent detail to its own OS window on the shared heap
14611 FM cockpit per-agent control cluster: start / stop / restart with honest round-trip states
14612 FM cockpit fleet-level morning start: one action, staged bring-up, honest partial results
14613 FM cockpit drill e2e: card → detail → pop-out → reattach round-trip
14614 FM cockpit connect and add-agent entry: operable-cold onboarding surface
14615 FM cockpit shell layout: compose the surface via the docking container contract
14616 FM cockpit perspective presets: named layouts over the saved-layout substrate
14617 FM cockpit auto-hide rails: secondary panes on the edge-rail chain
14618 FM cockpit visual-regression baseline harness
14619 FM cockpit accessibility pass: focus order, contrast, reduced motion
14620 FM cockpit catch-up view: what happened since you last looked
14631 FM module topology record: childapp vs view-region inside apps/agentos
14632 FM module mount: navigation integration + route wiring in the agentos Viewport
14635 FM module FamilyRail primitive: era-attribute data-binding (ADR 0032 §2.3.3)
14637 FM module HealthSwatch primitive: the five-category legend unit
14639 FM module event-kind map: the data-driven kind→token registry
14641 FM module card name slot: display-name binding with provenance/assent trail
14643 FM module stale-wire rendering: the not-wired and low-confidence card states
14646 FM cockpit walkthrough demo: the mission-control tour, NL-driven and recordable
14657 FM module cockpit dock document: the default layout as dockZone.v1 data
14658 FM module cockpit projection wiring: adapter render + resize commit loop
14681 Light + dark theme system: first-class dual-mode tokens for the agentos module and demos
14728 Harden fleet closed-set token resolvers against prototype-shaped keys (hasOwn)
subIssuesCompleted13
subIssuesTotal34
contentTrust
projected
quarantined0
signals[]
blockedBy[]
blocking[]
milestonev13.2

Epic: Fleet Manager cockpit UI/UX — the design-led product surface (Lane B of the cockpit plan)

Open Backlog/active-chunk-2 enhancementepicdeveloper-experienceai
neo-opus-vega
neo-opus-vega commented on Jul 4, 2026, 2:26 AM

Provenance: Operator-directed (@tobiu, 2026-07-04): the FM design/UX direction gets its own epic. Design SSOT: apps/agentos/design/fleet-manager-cockpit-plan.html (authored @neo-opus-grace, committed via #14512) — "Mission control for a cross-model AI engineering team — not a settings form." This epic elevates that plan's Lane B (the cockpit — UI/UX) to a dedicated, design-led epic per the operator's call. Grace holds design authority over the SSOT artifact.

Problem scope

The fleet-services spine is largely done Brain-side (define/start/stop/provision exist as services — the closed leaves under #13015), but the product surface is ~0% planned and the current apps/agentos UI is brittle: feature-first accretion, no design system, no UX bar. The artifact's own diagnosis: "Done right it's the flagship demo. Done as it is now, it undercuts the whole story."

The cockpit is the exact app the engine is built for — live multi-agent state, high-frequency streams, views splitting across OS windows on one shared App-Worker heap, object permanence. A weak surface here doesn't just cost adoption; it publicly falsifies the engine story.

Why an epic, not a ticket: a token/theme layer, the cockpit shell + fleet grid, the live activity stream, agent detail + pop-out, fleet controls, and NL-driven e2e coverage are separable one-PR leaves that must converge on ONE visual/interaction system across multiple weeks.

Intended solution shape

  • The artifact is the design SSOT. Its visual vocabulary (family rails, state dots + pulse, health summary, activity-event kinds), its three product principles (state reads at a glance · real-time is the spine · operable cold), and its §04 PoC falsifier (@tobiu starts one agent from the UI instead of a terminal) govern this epic's subs. Deltas from the artifact are recorded design decisions on subs — never silent drift.
  • Neo-native, engine-strengths-first: theme/token layer before views; real components (no hand-rolled per-view DOM); the activity stream as the real-time showcase; pop-out via the multi-window/shared-heap primitive; docking/perspectives consumed from #13158 as the container contract.
  • NL-driven e2e is the verification spine — every shipped view lands with Neural-Link/whitebox e2e coverage; the cockpit doubles as the natural Neural-Link demo surface.
  • Evolve-apps/agentos-vs-fresh-shell-app is the first gated design decision (a sub), not assumed here.

Out of scope

  • Lane A (Electron shell — #13033), Lane C (NL/MCP wiring — e.g. #14537), Lane D (credentials/onboarding) — siblings from the same plan, separately owned.
  • FM capability semantics + the service spine — #13015 (parent context #13012).
  • Docking engine internals — #13158 (consumed, not modified here).
  • v14 Institution Cockpit / COP render-model — #13444 / #14445.
  • #13448's keeper-views/nav definition overlaps this surface; reconciliation belongs to that epic's owner (flagged via A2A), not silently absorbed here.

Avoided traps

No second brittle app (the evolve-vs-fresh call is an explicit gated sub) · no pixel-guessing outside the SSOT · no UX-without-e2e · no bundling separable leaves into one sub. The epic is Refs-only — never a PR close-target.

Structure-map gate: N/A — app-layer (apps/) design/UX epic; no new ai/ substrate placement.

tobiu referenced in commit f9dabd0 - "feat(agentos): fleet cockpit ActivityStream — bounded, backpressure-aware feed (#14606) (#14827) on Jul 5, 2026, 7:39 AM
tobiu referenced in commit 7eb145a - "feat(agentos): FM cockpit control cluster — per-agent power toggle + whole-fleet + honest-state render (#14611) (#14892) on Jul 6, 2026, 6:56 PM