Context
A-tranche gate of Epic #14560's horizon map v2 — the first post-veto leaf: HOW the fleet-manager module lives inside the one harness app (#14577 veto: apps/agentos includes FM as a module). Every A/B-group leaf mounts into the shape this record picks.
The Problem
apps/agentos offers two shipped module shapes: childapps (own index.html + neo-config.json per child — popup-window-capable, e.g. childapps/strategy|swarm|widget) and view-regions (views composed in the main Viewport's navigation). The FM module needs: the cockpit as a primary in-app surface (the SSOT's default view), pop-out capability for detail (#14610 — which childapps get structurally), and shell-agnostic mounting (#13033 hosts the app later). Picking wrong forks 60 leaves' worth of mounts.
The Architectural Reality
- Childapp shape: aligned with the pop-out story (own window on the shared heap is the childapp mechanic) but makes the COCKPIT a secondary window — inverted product weight (mission control IS the home view).
- View-region shape: the cockpit as the agentos main view (or its primary nav target) with pop-out via the multi-window primitive per-pane — matches the SSOT ("the cockpit's default view") + ADR 0029 §2.6 layout-blind cards.
- Hybrid (likely): cockpit = view-region primary; detail pop-out = the childapp/window mechanic per pane; the decision record names which mechanic each surface uses.
The Fix
A 30-minute decision record (comment-resolution on this ticket, cited by every A/B leaf): the module's mount shape, the nav integration point in the agentos Viewport, the route class (#fleet-style), and the per-surface window mechanic table. Design-authority review (Grace) per the epic rule.
Acceptance Criteria
Out of Scope
The mount implementation (next leaf) · pop-out behavior (#14610) · demolition of existing views (their own leaves).
Related
Epic #14560 (parent) · #14577 (the veto) · #14628/PR #14629 (the relocation) · ADR 0020/0029.
Sweeps: live window current (newest known #14628); no equivalent; A2A clean. Structure-map: N/A — decision record.
Origin Session ID: a28f1415-37a4-4a5a-a3e4-b56adf6274ec
Retrieval Hint: "FM module topology childapp vs view region mount decision"
Context
A-tranche gate of Epic #14560's horizon map v2 — the first post-veto leaf: HOW the fleet-manager module lives inside the one harness app (
#14577veto:apps/agentosincludes FM as a module). Every A/B-group leaf mounts into the shape this record picks.The Problem
apps/agentosoffers two shipped module shapes: childapps (ownindex.html+neo-config.jsonper child — popup-window-capable, e.g.childapps/strategy|swarm|widget) and view-regions (views composed in the main Viewport's navigation). The FM module needs: the cockpit as a primary in-app surface (the SSOT's default view), pop-out capability for detail (#14610 — which childapps get structurally), and shell-agnostic mounting (#13033 hosts the app later). Picking wrong forks 60 leaves' worth of mounts.The Architectural Reality
The Fix
A 30-minute decision record (comment-resolution on this ticket, cited by every A/B leaf): the module's mount shape, the nav integration point in the agentos Viewport, the route class (
#fleet-style), and the per-surface window mechanic table. Design-authority review (Grace) per the epic rule.Acceptance Criteria
Out of Scope
The mount implementation (next leaf) · pop-out behavior (#14610) · demolition of existing views (their own leaves).
Related
Epic #14560 (parent) ·
#14577(the veto) · #14628/PR #14629 (the relocation) · ADR 0020/0029.Sweeps: live window current (newest known #14628); no equivalent; A2A clean. Structure-map: N/A — decision record.
Origin Session ID: a28f1415-37a4-4a5a-a3e4-b56adf6274ec Retrieval Hint: "FM module topology childapp vs view region mount decision"