PRD-72: Activity Command Centre v2

Version: 2.0 Status: Draft Priority: P1 Author: Gar Kavanagh + Auto CTO Created: 2026-03-07 Updated: 2026-03-11 Dependencies: PRD-10 (Workflow Engine — COMPLETE), PRD-55 (Agent Heartbeats — COMPLETE), PRD-06 (Monitoring & Analytics — COMPLETE), PRD-76 (Agent Reporting — COMPLETE), PRD-77 (Agent Self-Scheduling — IN PROGRESS) Branch: feat/activity-command-centre-v2 Inspiration: crshdn/mission-control — Kanban patterns, @hello-pangea/dnd, SSE real-time

---

Executive Summary

The Activity page is the nerve centre of Automatos — but v1 was a list-based feed that required clicking through multiple pages to understand what's happening. Users couldn't track task progress, manage workload, review agent output, or see their schedule without navigating 3-4 different views.

v2 rebuilds Activity as a five-tab operational dashboard that replaces the need for Jira, combines Feed + Reports into a Kanban board, adds a Calendar for scheduled tasks, enhances Memory with a daily journal view, and introduces a Projects page for multi-agent initiatives.

What's Changing

v1 Tab	v2 Tab	What Changed
Dashboard	Summary	Rename + new analytics widgets (status donut, priority breakdown, workload, types of work)
Feed	Board	List → Kanban board with drag-and-drop, agent sidebar, unified task viewer
Reports	(merged into Board)	Reports are now viewable inside task cards at Review/Done stages
Memory	Memory	Enhanced: two-panel layout, daily journal, organized by day/agent/type
Missions	Projects	Evolved from placeholder into project cards with progress tracking
(new)	Calendar	Full scheduler view: always-running tasks, week grid, next-up list

What's New (Cross-Cutting)

• Global Search (Cmd+K) — spotlight-style overlay searching tasks, memories, documents, agents

• SSE real-time updates — Board receives live task status changes via Server-Sent Events instead of polling

• Unified Task Viewer — one slide-over component replaces separate Feed detail, Reports viewer, and Execution detail

What We're NOT Building

• New execution engine (existing recipe executor + heartbeat service are sufficient)

• Agent Zero / Orchestrator rename (separate effort, out of scope)

• Chat entries on the Board (chat history stays at /chat)

• Analytics replacement (Analytics page stays for cost/performance trends)

---

1. Naming Convention

User-Facing Terminology

UI displays "Task" — backend remains "Recipe". No backend rename.

Technical Concept	User-Facing Name	Icon	Description
Recipe execution (single/multi-step)	Task	CheckSquare	A unit of work — one step or many
Agent heartbeat	Routine	RefreshCw	Recurring background check
Complex dynamic workflow	Project	FolderKanban	Multi-task initiative with progress tracking
Chat + tool calls	Chat	MessageCircle	Conversations (not shown on Board)

Task Status Labels (Board Columns)

Status	Column	Badge Colour	CSS Variable
inbox	Inbox	--muted	bg-muted/10 text-muted-foreground
assigned	Assigned	--agent	bg-[hsl(var(--agent))]/10 text-[hsl(var(--agent))]
in_progress	In Progress	--info	bg-[hsl(var(--info))]/10 text-[hsl(var(--info))]
review	Review	--warning	bg-[hsl(var(--warning))]/10 text-[hsl(var(--warning))]
done	Done	--success	bg-[hsl(var(--success))]/10 text-[hsl(var(--success))]
failed	Done (with error)	--destructive	bg-destructive/10 text-destructive

Priority Levels

Priority	Colour	Indicator
urgent	--destructive	Red dot / border
high	--warning	Amber badge
medium	--info	Blue badge
low	--muted	Gray badge

---

2. Design System Compliance

All components follow the established Automatos design language from globals.css. No custom colours or one-off styling.

Required Patterns

Pattern	Class / Token	Used By
Page header	<PageHeader> shared component	Title with gradient-text accent
Stat cards	<StatsBar> shared component	4-up hero stats, glass-card + card-glow
Tab navigation	<FilterTabs> shared component	Icon + label tabs, bg-secondary/40 backdrop-blur
Content cards	glass-card class	All card surfaces
Drill-down panels	glass-panel class	Task viewer slide-over
Status badges	Semantic Badge variants	Using CSS variable palette above
Charts	Recharts with MODEL_COLORS palette	Donut, bar charts on Summary
Motion	framer-motion staggered entrance	initial → animate with delay: index * 0.08
Empty states	Centered icon + two-line text	Muted icon (30% opacity) + description
Loading	<Skeleton> matching card layout	Skeleton grid matching real content shape
Mobile	Reduced backdrop-blur(8px)	Per globals.css @media (max-width: 767px)

Typography Scale

• Page title: text-2xl md:text-3xl font-bold

• Stat value: text-2xl font-bold leading-none

• Stat label: text-sm text-muted-foreground

• Card title: CardTitle with icon + gap-2

• Table header: text-[11px] font-medium text-muted-foreground uppercase tracking-wider

• Body text: text-sm

• Timestamps: text-xs text-muted-foreground

Colour Semantics (DO NOT hardcode hex)

Primary/Orange:  hsl(var(--primary))       — active states, CTAs, accents
Success/Green:   hsl(var(--success))        — completed, healthy
Info/Blue:       hsl(var(--info))           — running, in-progress
Warning/Amber:   hsl(var(--warning))        — review, attention
Destructive/Red: hsl(var(--destructive))    — failed, errors
Agent/Purple:    hsl(var(--agent))          — agent-related, assigned
Muted:           hsl(var(--muted-foreground)) — inactive, secondary text

---

3. Page Layout

Route

/activity (unchanged from v1)

Tab Structure

┌──────────────────────────────────────────────────────────────┐
│  <PageHeader                                                  │
│    title="Command"                                            │
│    titleAccent="Centre"                                       │
│    subtitle="Your AI workforce at a glance"                   │
│    actions={<PeriodToggle /> <RefreshButton />}               │
│  />                                                           │
├──────────────────────────────────────────────────────────────┤
│  <StatsBar stats={[                                           │
│    { label: "Working Now", value: "3", icon: Activity }       │
│    { label: "Agents Active", value: "11", icon: Users }       │
│    { label: "Tasks in Queue", value: "35", icon: ListTodo }   │
│    { label: "Needs Attention", value: "1", icon: AlertTriangle│
│  ]} />                                                        │
├──────────────────────────────────────────────────────────────┤
│  <FilterTabs tabs={[                                          │
│    { value: "summary",  label: "Summary",  icon: LayoutDashboard }│
│    { value: "board",    label: "Board",    icon: Columns3 }   │
│    { value: "calendar", label: "Calendar", icon: Calendar }   │
│    { value: "memory",   label: "Memory",   icon: Brain }      │
│    { value: "projects", label: "Projects", icon: FolderKanban }│
│  ]}>                                                          │
│    <TabsContent value="summary">   → Section 4              │
│    <TabsContent value="board">     → Section 5              │
│    <TabsContent value="calendar">  → Section 6              │
│    <TabsContent value="memory">    → Section 7              │
│    <TabsContent value="projects">  → Section 8              │
│  </FilterTabs>                                                │
└──────────────────────────────────────────────────────────────┘

Deep-link support: ?tab=board&task_id=123 opens Board with task viewer for that task.

---

4. Summary Tab (Default)

Renamed from "Dashboard". Draggable widget grid using react-grid-layout (already installed).

4.1 Existing Widgets (Keep)

Widget	Component	Description
Active Now	active-now-widget.tsx	Top 5 currently running tasks with progress bars
Schedule	schedule-widget.tsx	Week calendar dots + upcoming 5 items. Click "View All" → Calendar tab
Agent Reports	agent-reports-widget.tsx	Pinned agent report cards
Recent Activity	recent-activity-widget.tsx	Last 5 completed + 3 failed. Click "View All" → Board tab

4.2 New Widgets

Status Overview (Donut Chart)

┌─ glass-card ──────────────────────────────┐
│  Status Overview              [View all →] │
│                                            │
│         ┌─────────┐                        │
│        ╱    77     ╲    ● In Review: 23    │
│       │  Total work │   ● Done: 17         │
│        ╲   items   ╱    ● To Do: 20        │
│         └─────────┘     ● In Progress: 17  │
│                                            │
└────────────────────────────────────────────┘

• Recharts PieChart with innerRadius for donut

• Counts from Board data (GET /api/activity/board/stats)

• Click segment → navigates to Board tab filtered by that status

• Click "View all" → Board tab

Priority Breakdown (Bar Chart)

┌─ glass-card ──────────────────────────────┐
│  Priority Breakdown                        │
│                                            │
│  60 ┤                                      │
│  40 ┤            ██                        │
│  20 ┤      ██    ██                        │
│   0 ┤  ██  ██    ██    ██    ██            │
│     Urgent High Medium  Low  None          │
└────────────────────────────────────────────┘

• Recharts BarChart

• Shows task count by priority level

• Colour-coded bars using priority colours from Section 1

Types of Work (Horizontal Bar)

┌─ glass-card ──────────────────────────────┐
│  Types of Work                 [View all →]│
│                                            │
│  Type          Distribution                │
│  ⚡ Routine    ████████████████████  65%    │
│  ✅ Task       ███████████  28%             │
│  📁 Project    ███  7%                      │
└────────────────────────────────────────────┘

• Shows distribution of work items by type

• Horizontal stacked bars

Team Workload (Agent Distribution)

┌─ glass-card ──────────────────────────────┐
│  Team Workload                             │
│                                            │
│  Assignee        Work distribution         │
│  🤖 Sentinel     ████████████████  12      │
│  📝 Loki         ██████████  8             │
│  🔍 Vision       ████████  6               │
│  ⚡ Pepper       ██████  4                 │
│  ○ Unassigned    ████  3                   │
└────────────────────────────────────────────┘

• Horizontal bars per agent

• Click agent row → Board filtered by that agent

4.3 Widget Grid

All 8 widgets rendered via react-grid-layout with drag-to-reorder. Layout persisted in localStorage (automatos:dashboard-layout). "Customize" toggle + "Reset Layout" button (existing pattern from v1).

Default layout (2-column on desktop):

Row 1: [Active Now          ] [Status Overview    ]
Row 2: [Schedule             ] [Priority Breakdown ]
Row 3: [Recent Activity      ] [Types of Work      ]
Row 4: [Agent Reports        ] [Team Workload      ]

---

5. Board Tab (Replaces Feed + Reports)

The centrepiece of v2. A Kanban board with agent sidebar, drag-and-drop task management, and a unified task viewer that replaces the separate Feed detail and Reports pages.

5.1 Layout

┌──────────────────────────────────────────────────────────────┐
│  [Filter: agent | priority | type | date]   11 agents | 35 tasks │
├──────────┬───────────────────────────────────────────────────┤
│ ● AGENTS │  ● TASK BOARD                                     │
│    (12)  │                                                    │
│          │  ● INBOX  ● ASSIGNED  ● IN PROGRESS  ● REVIEW  ● DONE │
│ Bhanu    │    (11)     (10)        (7)           (5)      (0) │
│ LEAD     │                                                    │
│ ●WORKING │  ┌────┐   ┌────┐     ┌────┐        ┌────┐        │
│          │  │card│   │card│     │card│        │card│        │
│ Friday   │  │    │   │    │     │    │        │    │        │
│ INT      │  └────┘   └────┘     └────┘        └────┘        │
│ ●WORKING │  ┌────┐   ┌────┐     ┌────┐        ┌────┐        │
│          │  │card│   │card│     │card│        │card│        │
│ Fury     │  └────┘   └────┘     └────┘        └────┘        │
│ SPC      │                                                    │
│ ●WORKING │           ┌────┐     ┌────┐                       │
│          │           │card│     │card│                       │
│ ...      │           └────┘     └────┘                       │
└──────────┴───────────────────────────────────────────────────┘

5.2 Agent Sidebar

Left panel showing all workspace agents:

┌─ AGENTS (12) ────────────────┐
│                               │
│  🤖 Bhanu          LEAD      │
│     Founder        ● WORKING │
│                               │
│  ⚔️ Friday         INT       │
│     Developer      ● WORKING │
│                               │
│  🔥 Fury           SPC       │
│     Customer Rese… ● WORKING │
│                               │
│  ...                          │
└───────────────────────────────┘

• Agent avatar, name, role badge (LEAD/INT/SPC), status dot

• Click agent → filters board to show only that agent's tasks

• Active filter shown as highlighted agent row

• Click again to clear filter

• Collapsible on mobile (hamburger toggle)

Data source: GET /api/agents?workspace_id=X (existing endpoint)

5.3 Board Columns

5 columns using @hello-pangea/dnd (fork of react-beautiful-dnd, actively maintained):

Column	Status	Header Colour	Description
Inbox	inbox	--muted	New tasks not yet assigned
Assigned	assigned	--agent	Assigned to an agent, waiting to start
In Progress	in_progress	--info	Currently executing
Review	review	--warning	Awaiting human or LLM review
Done	done	--success	Completed (success or failure)

Each column header shows: ● COLUMN NAME (count)

Columns are scrollable vertically. Cards within a column are ordered by priority (urgent first), then by updated_at DESC.

5.4 Board Card

Each task is a draggable card:

┌─ glass-card draggable ────────────────────┐
│  ↕ Explore SiteGPT Dashboard &            │
│    Document All Features                   │
│                                            │
│    Thoroughly explore the entire            │
│    SiteGPT dashboard...                    │
│                                            │
│  🤖 Loki              1 day ago            │
│  [research] [documentation] [sitegpt]      │
│                                    [→]     │
└────────────────────────────────────────────┘

Card fields:

• Task name (line-clamp-2)

• Description snippet (line-clamp-2, text-sm text-muted-foreground)

• Agent avatar + name (bottom-left)

• Time ago (bottom-right)

• Tags as pills (text-xs badges)

• Priority indicator: left border colour or small dot next to title

• Arrow button → opens Task Viewer

• For "In Progress": mini progress bar showing step X of Y

Drag behaviour:

• @hello-pangea/dnd <DragDropContext> wrapping the board

• Each column is a <Droppable>

• Each card is a <Draggable>

• On drag end: optimistic UI update → PATCH /api/activity/tasks/{id}/status

• On API failure: revert to previous position + toast error

• Cards in "In Progress" with running backend status: draggable (user can force-move to Review/Done)

• Cards in "Done" with completed status: draggable back to Inbox for re-run

5.5 Task Viewer (Slide-Over)

Replaces: activity-feed-item.tsx detail, report-viewer.tsx, execution-detail.tsx

Opens as a right slide-over panel when clicking a card's arrow button or double-clicking a card.

Content adapts based on the task's current column:

Inbox / Assigned View

┌─ Task Viewer ─────────────────────────────────────┐
│  ← Back                                    [Edit] │
│                                                    │
│  Explore SiteGPT Dashboard                        │
│  Created by Henry · 1 day ago                      │
│                                                    │
│  Description:                                      │
│  Thoroughly explore the entire SiteGPT dashboard   │
│  and document all features...                      │
│                                                    │
│  ┌─ Details ──────────────────────────────────┐   │
│  │ Assignee:  [🤖 Loki ▼]                     │   │
│  │ Priority:  [● High ▼]                      │   │
│  │ Due Date:  [Mar 15, 2026]                   │   │
│  │ Tags:      [research] [documentation] [+]   │   │
│  │ Steps:     3 steps defined                  │   │
│  │ Review:    [Human ▼] (Human / LLM / Auto)  │   │
│  └────────────────────────────────────────────┘   │
│                                                    │
│  [Start Task →]                                    │
└────────────────────────────────────────────────────┘

• Editable fields: assignee (agent dropdown), priority, due date, tags, review mode

• "Start Task" moves to In Progress and triggers execution

In Progress View

┌─ Task Viewer ─────────────────────────────────────┐
│  ← Back                                   [Stop]  │
│                                                    │
│  Explore SiteGPT Dashboard                        │
│  🤖 Loki · Working... · Step 2 of 3              │
│                                                    │
│  ┌─ Progress ─────────────────────────────────┐   │
│  │ ✓ Step 1: Pull dashboard data (45s)         │   │
│  │ ● Step 2: Analyze features...               │   │
│  │ ○ Step 3: Write documentation               │   │
│  └────────────────────────────────────────────┘   │
│                                                    │
│  ┌─ Live Logs ────────────────────────────────┐   │
│  │ 09:15:00  Started execution                 │   │
│  │ 09:15:02  Step 1: Fetching dashboard...     │   │
│  │ 09:15:47  Step 1: Complete (45s)            │   │
│  │ 09:15:48  Step 2: Analyzing features...     │   │
│  │ ● (streaming...)                            │   │
│  └────────────────────────────────────────────┘   │
│                                                    │
│  [Move to Review]                                  │
└────────────────────────────────────────────────────┘

• Reuses internals from existing execution-detail.tsx

• Live log streaming via SSE

• Step pipeline with stage-completed, stage-active, stage-pending CSS classes

• "Stop" button cancels execution

• "Move to Review" manually advances (or auto-advances on completion)

Review View

┌─ Task Viewer ─────────────────────────────────────┐
│  ← Back                              Review Mode: │
│                                      [Human ▼]    │
│  Explore SiteGPT Dashboard                        │
│  🤖 Loki · Completed in 2m 34s · Awaiting Review │
│                                                    │
│  ┌─ Report ───────────────────────────────────┐   │
│  │ ## SiteGPT Dashboard Analysis              │   │
│  │                                             │   │
│  │ Key findings:                               │   │
│  │ - Feature A does X                          │   │
│  │ - Feature B does Y                          │   │
│  │ ...                                         │   │
│  └────────────────────────────────────────────┘   │
│                                                    │
│  ┌─ Grade ────────────────────────────────────┐   │
│  │ Rating: ★★★★☆                              │   │
│  │ Notes:  [                                ]  │   │
│  └────────────────────────────────────────────┘   │
│                                                    │
│  [✗ Reject → Inbox] [✓ Approve → Done]            │
└────────────────────────────────────────────────────┘

• Renders report content as markdown (reuses report-viewer.tsx internals)

• Grade form with star rating + notes (reuses report-grade-form.tsx)

• Review mode selector: Human / LLM / Auto

  • Human: requires manual approve/reject

  • LLM: another agent reviews and auto-approves/rejects based on quality threshold

  • Auto: moves to Done automatically on completion (no review step)

• Reject sends back to Inbox with rejection notes

• Approve moves to Done with grade saved

Done View

┌─ Task Viewer ─────────────────────────────────────┐
│  ← Back                                           │
│                                                    │
│  Explore SiteGPT Dashboard          ✓ Completed   │
│  🤖 Loki · 2m 34s · Mar 7, 09:17                 │
│  Grade: ★★★★☆                                     │
│                                                    │
│  ┌─ Results ──────────────────────────────────┐   │
│  │ ## SiteGPT Dashboard Analysis              │   │
│  │ (full report content)                       │   │
│  └────────────────────────────────────────────┘   │
│                                                    │
│  ┌─ Execution Summary ────────────────────────┐   │
│  │ Duration: 2m 34s                            │   │
│  │ Tokens:   4,521                             │   │
│  │ Steps:    3/3 completed                     │   │
│  │ Trigger:  Manual                            │   │
│  └────────────────────────────────────────────┘   │
│                                                    │
│  [↻ Re-run] [📥 Download Report] [📋 Copy]       │
└────────────────────────────────────────────────────┘

• Full report content rendered

• Execution metadata: duration, tokens, steps, trigger

• Re-run creates a new execution of the same recipe

• Download as markdown

• Failed tasks show error message + stack trace in collapsible section

5.6 Creating Tasks

"New Task" button in Board header → opens existing create-recipe-modal.tsx:

• 1-step recipe = simple task

• N-step recipe = multi-step workflow

• No backend changes needed — recipe model already supports both

Task can also be created by:

• Agents spawning sub-tasks via platform_schedule_task tool (PRD-77)

• QA agent adding bug-fix tasks to Inbox

• Redis queue pushing new tasks (existing infrastructure)

5.7 Board Filters

Filter bar above the board:

[Agent ▼] [Priority ▼] [Type ▼] [Date Range ▼]  |  🔍 Search tasks...

• Agent: multi-select dropdown of workspace agents

• Priority: multi-select (Urgent/High/Medium/Low)

• Type: Routine / Task (single-step) / Workflow (multi-step)

• Date Range: Today / This Week / This Month / Custom

• Search: instant filter by task name/description

Filters persist in URL query params for shareability.

5.8 Real-Time Updates (SSE)

Board uses Server-Sent Events instead of polling:

GET /api/activity/board/stream?workspace_id=X

SSE Event Types:

Event	Payload	UI Effect
task_created	Full task object	Card appears in Inbox with log-slide-in animation
task_updated	{ id, status, agent_id, ... }	Card moves to new column with animation
task_progress	{ id, step_current, step_total, log_line }	Progress bar updates, log line appends
task_completed	{ id, status, duration, report_id }	Card moves to Review or Done
task_failed	{ id, error_message }	Card shows error badge, moves to Done

Fallback: If SSE connection drops, fall back to 60s polling via useActivityFeed() (existing hook). Reconnect SSE with exponential backoff (1s, 2s, 4s, max 30s).

Backend implementation:

• New SSE endpoint in orchestrator/api/activity.py

• Publishes events from: recipe executor, heartbeat service, manual status changes

• Uses Redis pub/sub as event bus (existing Redis infrastructure)

---

6. Calendar Tab

Full-page scheduler view showing all of Automatos's scheduled and recurring tasks. Lets users verify what's running proactively, spot unwanted schedules, and click through to task details.

6.1 Layout

┌──────────────────────────────────────────────────────────────┐
│  Scheduled Tasks                         [Week] [Today] [↻] │
│  {user}'s automated routines                                 │
├──────────────────────────────────────────────────────────────┤
│  ⚡ Always Running                                           │
│  ┌────────────────────────────────────────────┐              │
│  │ mission control check • Every 30 min       │              │
│  └────────────────────────────────────────────┘              │
├──────────────────────────────────────────────────────────────┤
│  Sun    │  Mon    │  Tue    │  Wed    │  Thu    │ *Fri* │ Sat│
│ ┌─────┐│┌─────┐  │┌─────┐  │┌─────┐  │┌─────┐  │┌─────┐│┌──┐│
│ │ai sc│││ai sc│  ││ai sc│  ││ai sc│  ││ai sc│  ││ai sc││    │
│ │5:00A│││5:00A│  ││5:00A│  ││5:00A│  ││5:00A│  ││5:00A││    │
│ ├─────┤│├─────┤  │├─────┤  │├─────┤  │├─────┤  │├─────┤│    │
│ │morn │││morn │  ││morn │  ││morn │  ││morn │  ││morn ││    │
│ │8:00A│││8:00A│  ││8:00A│  ││8:00A│  ││8:00A│  ││8:00A││    │
│ ├─────┤│├─────┤  │├─────┤  │├─────┤  │├─────┤  │├─────┤│    │
│ │comp │││comp │  ││newsl│  ││comp │  ││comp │  ││comp ││    │
│ │10:00│││10:00│  ││9:00A│  ││10:00│  ││10:00│  ││10:00││    │
│ └─────┘│└─────┘  │├─────┤  │└─────┘  │└─────┘  │└─────┘│    │
│        │         ││comp │  │         │         │       │    │
│        │         ││10:00│  │         │         │       │    │
│        │         │└─────┘  │         │         │       │    │
├──────────────────────────────────────────────────────────────┤
│  📅 Next Up                                                  │
│  mission control check .......................... In 30 min  │
│  competitor youtube scan ........................ In 1 hours  │
│  ai scarcity research .......................... In 20 hours │
│  morning brief .................................. In 23 hours │
│  newsletter reminder ............................ In 4 days  │
└──────────────────────────────────────────────────────────────┘

6.2 Three Sections

Always Running

• Shows high-frequency routines (interval < 1 hour)

• Pill badge: {name} • Every {interval}

• Click → opens task viewer

Week Grid

• 7-column day grid (Sun-Sat)

• Task pills colour-coded per agent (consistent colours from agent config)

• Shows task name (truncated) + scheduled time

• Parsed from cron expressions in schedule_config

• Current day column highlighted (today indicator)

• Click pill → opens task viewer slide-over

Next Up

• Sorted list of upcoming tasks by soonest

• Shows task name + relative time ("In 30 min", "In 1 hours")

• Colour-coded text matching agent colour

• Click row → opens task viewer

6.3 Data Source

Extend existing useActivitySchedule() hook:

• Already returns next_run_at, frequency, agent_name, agent_id

• Add: cron expression parsing to plot recurring items on week grid

• Add: "always running" filter for sub-hourly intervals

• Endpoint: GET /api/activity/schedule?range=7d (existing)

6.4 View Toggle

• Week (default): 7-day grid view

• Today: single-day view with hourly timeline

• Refresh button: re-fetches schedule data

6.5 Library Decision

Build custom (no external calendar library needed). The design is a simple week grid with pills — @fullcalendar is overkill. A CSS grid with mapped cron data is sufficient. If we later need month view or drag-to-reschedule, we can add @fullcalendar then.

---

7. Memory Tab

Two-panel layout inspired by daily journal apps. Browse and search through all of Automatos's memories, organized chronologically with rich content rendering.

7.1 Layout

┌─────────────────────────┬──────────────────────────────────────┐
│ 🔍 Search memory...     │  2026-02-26 — Thursday               │
│                         │  Thursday, Feb 26 · 4.8 KB · 772 w  │
│ 🧠 Long-Term Memory    │  ─────────────────────────────────── │
│    1,608 words          │                                      │
│    Updated 22 hours ago │  ⏰ 9:00 AM — Qwen 3.5 Medium       │
│                         │     Series Research                  │
│ 📖 DAILY JOURNAL       │                                      │
│    37 entries           │  What we discussed: Alex shared the  │
│                         │  Qwen 3.5 Medium announcement...     │
│ ▼ Yesterday (1)        │                                      │
│   📅 Mon, Mar 2        │  Key findings:                       │
│   8.7 KB • 1,253 words │  - 35B-A3B beats old 235B flagship   │
│                         │  - 122B-A10B matches 397B            │
│ ▼ This Week (1)        │  - 27B dense gets best SWE-bench     │
│   📅 Sun, Mar 1        │                                      │
│   6.8 KB • 1,068 words │  Recommendations given:              │
│                         │  1. Keep 397B on Studio 2            │
│ ▼ February 2026 (25)   │  2. Add 35B-A3B on Studio 1          │
│   📅 Sat, Feb 28       │  3. 122B-A10B as potential upgrade    │
│   12.6 KB • 2,001 w    │                                      │
│   📅 Fri, Feb 27       │  Decision: Pending                   │
│   3.9 KB • 614 words   │  ─────────────────────────────────── │
│   📅 Thu, Feb 26  ◀    │                                      │
│   4.8 KB • 772 words   │  Overnight — Reborn Factory Results  │
│   ...                   │  ...                                  │
└─────────────────────────┴──────────────────────────────────────┘

7.2 Left Sidebar

Search

• Full-text search across all memories

• Fix required: verify Mem0Client.search() POST fix merged from fix-memory branch

• Results replace the journal list with search results, ranked by relevance score

• Keyboard: Enter to search, Esc to clear

Long-Term Memory Card

• Shows total word count, last updated timestamp

• Click → displays consolidated long-term memory in right panel

• Badge: 🧠 emoji or Brain icon

Daily Journal

• Grouped by time period:

  • Yesterday (collapsible)

  • This Week (collapsible)

  • {Month Year} (collapsible) — e.g., "February 2026 (25)"

• Each entry row: 📅 {Day, Mon DD} + {size} • {word count}

• Click entry → loads full content in right panel

• Active entry highlighted with ◀ indicator and bg-primary/10 background

7.3 Right Content Panel

• Renders selected memory entry as rich markdown

• Header: {date} — {day name} + metadata (date, size, word count, modified time)

• Content sections with timestamp headers: ⏰ 9:00 AM — {Topic}

• Structured entries with: "What we discussed", "Key findings", "Recommendations", "Decision"

• Code blocks syntax-highlighted

• Images rendered inline

7.4 Organize By (Toggle)

Above the journal list, a segmented control:

[By Day] [By Agent] [By Type]

• By Day (default): grouped by date as shown above

• By Agent: grouped by agent name, each agent section shows their memories

• By Type: grouped by category (conversation, task, document, insight, research)

7.5 Enhanced Filters

• Agent dropdown (multi-select)

• Date range picker

• Type/Category filter (global / agent-specific)

• Sort: Newest (default) / Oldest / Relevance (when searching)

7.6 API Changes

Endpoint	Change
GET /api/memory/browse	Add group_by param: day / agent / type
GET /api/memory/browse	Add date_from, date_to params
POST /api/memory/search	Verify working (Mem0 POST fix)
GET /api/memory/health	Existing — no changes
GET /api/memory/journal/{date}	New — returns full day's memory content as structured markdown

---

8. Projects Tab

Evolved from the v1 "Missions" placeholder into a project management view. Projects are multi-task initiatives with progress tracking, assigned agents, and priority.

8.1 Layout

┌──────────────────────────────────────────────────────────────┐
│  📁 Projects                                     [+ New]     │
│  5 total • 2 active • 3 planning                             │
├──────────────────┬──────────────────┬───────────────────────┤
│ Agent Org Infra  │ Mission Control  │ Skool AI Extension    │
│ [Active]         │ [Active]         │                       │
│ Core infra for   │ Central dash-    │ Chrome ext for Vibe   │
│ the autonomous   │ board for agent  │ Code Academy. RAG     │
│ agent org...     │ activity, docs...│ pipeline over...      │
│                  │                  │                       │
│ ████████ 100%    │ ██████░░ 70%     │ ░░░░░░░░ 0%          │
│ 10/10            │ 0/0              │                       │
│                  │                  │                       │
│ 🟢 Charlie       │ 🟢 Henry         │ 🟢 Henry              │
│ [high]   8d ago  │ [high]   8d ago  │         8d ago        │
├──────────────────┼──────────────────┤                       │
│ Micro-SaaS       │ Even G2 Integr.  │                       │
│ [Planning]       │ [Planning]       │                       │
│ Violet's opp     │ Smart glasses    │                       │
│ engine — market  │ bridge app...    │                       │
│ gaps, validate...│                  │                       │
│                  │                  │                       │
│ ░░░░░░░░ 0%     │ ░░░░░░░░ 0%     │                       │
│ 0/0              │ 0/0              │                       │
│                  │                  │                       │
│ 🟣 Violet        │ ○ Unassigned     │                       │
│ [medium]  8d ago │ [medium]  8d ago │                       │
└──────────────────┴──────────────────┴───────────────────────┘

8.2 Project Card

┌─ glass-card ──────────────────────────────┐
│  Project Name                    [Active] │
│                                            │
│  Description text (line-clamp-3)...        │
│                                            │
│  ████████████░░░░░░░░  70%       7/10     │
│                                            │
│  🟢 Agent Name                             │
│  [priority]                    time ago    │
└────────────────────────────────────────────┘

Card fields:

• Project name

• Status badge: Active (green) / Planning (blue) / Complete (muted) / On Hold (amber)

• Description (line-clamp-3)

• Progress bar: completed tasks / total tasks, percentage

• Lead agent avatar + name

• Priority badge

• Time since creation

8.3 Project Detail (Click-Through)

Clicking a project card opens a detail view:

┌──────────────────────────────────────────────────────────────┐
│  ← Projects                                                  │
│                                                              │
│  Agent Org Infrastructure                         [Active]   │
│  Core infrastructure for the autonomous agent organization   │
│  Lead: 🟢 Charlie · Priority: High · Created 8 days ago     │
│                                                              │
│  ████████████████████ 100%    10/10 tasks                    │
│                                                              │
│  ┌─ Tasks ────────────────────────────────────────────────┐  │
│  │ ✓ Set up base agent framework                          │  │
│  │ ✓ Implement heartbeat service                          │  │
│  │ ✓ Configure inter-agent messaging                      │  │
│  │ ✓ Deploy monitoring stack                              │  │
│  │ ...                                                     │  │
│  └────────────────────────────────────────────────────────┘  │
│                                                              │
│  [+ Add Task] [Open in Board →]                              │
└──────────────────────────────────────────────────────────────┘

• Lists all tasks belonging to this project

• Each task row links to the Board task viewer

• "Add Task" creates a new task pre-linked to this project

• "Open in Board" navigates to Board tab filtered by this project's tasks

8.4 Data Model

Projects are a new concept — a group of recipes/tasks with shared metadata:

class Project(Base):
    __tablename__ = 'projects'

    id: int  # PK
    workspace_id: UUID
    name: str
    description: str
    status: str  # 'planning', 'active', 'complete', 'on_hold'
    lead_agent_id: int | None  # FK to agents
    priority: str  # 'low', 'medium', 'high', 'urgent'
    created_at: datetime
    updated_at: datetime

Link table for project ↔ task (recipe execution) relationship:

class ProjectTask(Base):
    __tablename__ = 'project_tasks'

    project_id: int  # FK to projects
    execution_id: str  # FK to recipe_executions
    order: int  # Display order within project

8.5 API Endpoints

Method	Path	Description
GET	/api/projects	List projects for workspace
POST	/api/projects	Create project
GET	/api/projects/{id}	Project detail with tasks
PATCH	/api/projects/{id}	Update project
POST	/api/projects/{id}/tasks	Link task to project
DELETE	/api/projects/{id}/tasks/{exec_id}	Unlink task

---

9. Global Search (Cmd+K)

Spotlight-style search overlay accessible from anywhere in the app.

9.1 Trigger

• Keyboard: Cmd+K (Mac) / Ctrl+K (Windows/Linux)

• Click: Search icon in top navigation bar

9.2 Layout

┌──────────────────────────────────────────────────┐
│  🔍 Mac Studio                          ✕   ESC │
├──────────────────────────────────────────────────┤
│  TASKS                                           │
│  ┌──────────────────────────────────────────┐    │
│  │ ✅ Flesh out $10K Mac Studio use cases  ●│    │
│  │    Clawdbot                            → │    │
│  ├──────────────────────────────────────────┤    │
│  │ ✅ Local model recommendations for Mac ● │    │
│  │    Infrastructure                      → │    │
│  ├──────────────────────────────────────────┤    │
│  │ ✅ Research Exo Labs dual-Studio clust ● │    │
│  │    Mac Studio Launch                   → │    │
│  └──────────────────────────────────────────┘    │
│                                                  │
│  MEMORIES                                        │
│  ┌──────────────────────────────────────────┐    │
│  │ 🧠 Mac Studio M2 Ultra performance...    │    │
│  │    Feb 26, 2026                         → │    │
│  └──────────────────────────────────────────┘    │
│                                                  │
│  ↑↓ navigate  ↵ select  esc close               │
├──────────────────────────────────────────────────┤

9.3 Search Categories

Results grouped by type:

Category	Icon	Source
Tasks	CheckSquare	Recipe executions (name, description)
Memories	Brain	Memory search (content)
Documents	FileText	Document metadata (name)
Agents	Bot	Agent names, roles
Projects	FolderKanban	Project names, descriptions

9.4 Result Item

Each result shows:

• Category icon

• Title (highlighted match)

• Subtitle (agent name, category, or date)

• Status dot (colour = current status for tasks)

• Arrow → navigates to item

9.5 Keyboard Navigation

• ↑↓ — move selection

• Enter — open selected item

• Esc — close search

• Type to filter in real-time (debounced 300ms)

9.6 API

GET /api/search?q={query}&workspace_id=X&types=tasks,memories,documents,agents,projects&limit=10

Returns results grouped by type, ordered by relevance. Backend searches across:

• recipe_executions (name, description via ILIKE)

• memories (Mem0 vector search)

• documents (name, metadata via ILIKE)

• agents (name, role via ILIKE)

• projects (name, description via ILIKE)

---

10. API Endpoints (Complete)

10.1 Existing Endpoints (No Changes)

Method	Path	Used For
GET	/api/recipes	Recipe list (used by Board internally)
GET	/api/recipes/{id}/executions	Recipe execution history
POST	/api/recipes/{id}/execute	Manual recipe run
GET	/api/activity/feed	Feed data (fallback for Board polling)
GET	/api/activity/stats	Stats for hero cards
GET	/api/activity/schedule	Calendar data source
GET	/api/agents	Agent list for sidebar
GET	/api/memory/health	Memory health stats
DELETE	/api/memory/{id}	Delete memory
POST	/api/memory/consolidate	Merge/summarize memories

10.2 Modified Endpoints

Method	Path	Change
GET	/api/activity/feed	Add status filter for board columns (inbox, assigned, in_progress, review, done)
GET	/api/memory/browse	Add group_by, date_from, date_to params

10.3 New Endpoints

Method	Path	Description
PATCH	/api/activity/tasks/{id}/status	Update task status (drag-and-drop)
PATCH	/api/activity/tasks/{id}	Update task fields (assignee, priority, tags, due_date, review_mode)
GET	/api/activity/board/stats	Board column counts + priority/type/agent breakdowns for Summary widgets
GET	/api/activity/board/stream	SSE — real-time task events stream
GET	/api/memory/journal/{date}	Full day's memory as structured markdown
GET	/api/projects	List projects
POST	/api/projects	Create project
GET	/api/projects/{id}	Project detail + linked tasks
PATCH	/api/projects/{id}	Update project
POST	/api/projects/{id}/tasks	Link task to project
DELETE	/api/projects/{id}/tasks/{exec_id}	Unlink task
GET	/api/search	Global search across all entity types

10.4 Database Migrations

Migration 1: Board statuses + task fields

-- Add new statuses to recipe_executions
ALTER TABLE recipe_executions
  ADD COLUMN priority VARCHAR(10) DEFAULT 'medium',
  ADD COLUMN tags JSONB DEFAULT '[]',
  ADD COLUMN assignee_agent_id INTEGER REFERENCES agents(id),
  ADD COLUMN review_mode VARCHAR(10) DEFAULT 'human',
  ADD COLUMN due_date TIMESTAMP,
  ADD COLUMN reviewed_at TIMESTAMP,
  ADD COLUMN reviewed_by VARCHAR(50);

-- Update status enum to include new values
-- (status is VARCHAR, so just document valid values)
-- Valid: 'inbox', 'assigned', 'pending', 'running', 'in_progress', 'review', 'completed', 'done', 'failed', 'cancelled'

Migration 2: Projects table

CREATE TABLE projects (
  id SERIAL PRIMARY KEY,
  workspace_id UUID NOT NULL REFERENCES workspaces(id),
  name VARCHAR(255) NOT NULL,
  description TEXT,
  status VARCHAR(20) DEFAULT 'planning',
  lead_agent_id INTEGER REFERENCES agents(id),
  priority VARCHAR(10) DEFAULT 'medium',
  created_at TIMESTAMP DEFAULT NOW(),
  updated_at TIMESTAMP DEFAULT NOW()
);

CREATE TABLE project_tasks (
  project_id INTEGER REFERENCES projects(id) ON DELETE CASCADE,
  execution_id VARCHAR(255) REFERENCES recipe_executions(execution_id),
  "order" INTEGER DEFAULT 0,
  PRIMARY KEY (project_id, execution_id)
);

CREATE INDEX idx_projects_workspace ON projects(workspace_id);
CREATE INDEX idx_project_tasks_project ON project_tasks(project_id);

10.5 Key Type Definitions

// Board task (extends ActivityFeedItem)
interface BoardTask {
  id: string
  type: 'routine' | 'recipe'
  name: string
  description?: string
  status: 'inbox' | 'assigned' | 'in_progress' | 'review' | 'done'
  priority: 'urgent' | 'high' | 'medium' | 'low'
  tags: string[]
  assignee?: {
    agent_id: number
    agent_name: string
    agent_icon?: string
  }
  creator?: string
  due_date?: string
  review_mode: 'human' | 'llm' | 'auto'
  started_at?: string
  completed_at?: string
  duration_ms?: number
  step_progress?: { current: number; total: number }
  error_message?: string
  report_id?: string
  source_id: string
  project_id?: number
}

// SSE event
interface BoardEvent {
  type: 'task_created' | 'task_updated' | 'task_progress' | 'task_completed' | 'task_failed'
  task: Partial<BoardTask> & { id: string }
  timestamp: string
}

// Project
interface Project {
  id: number
  name: string
  description: string
  status: 'planning' | 'active' | 'complete' | 'on_hold'
  lead_agent?: { id: number; name: string; icon?: string }
  priority: 'urgent' | 'high' | 'medium' | 'low'
  task_count: number
  completed_count: number
  progress_pct: number
  created_at: string
  updated_at: string
}

// Global search result
interface SearchResult {
  type: 'task' | 'memory' | 'document' | 'agent' | 'project'
  id: string
  title: string
  subtitle: string
  status?: string
  url: string
}

---

11. File Structure

frontend/
  app/
    activity/
      page.tsx                          # KEEP — route entry
  components/
    activity/
      activity-page.tsx                 # MODIFY — update tabs to Summary|Board|Calendar|Memory|Projects
      activity-feed.tsx                 # DELETE after Board is live (keep as fallback initially)
      activity-feed-item.tsx            # DELETE after Board is live
      activity-reports.tsx              # DELETE — merged into task viewer
      report-card.tsx                   # DELETE — replaced by board-card
      activity-missions.tsx             # MODIFY → activity-projects.tsx
      activity-memory.tsx               # REWRITE — two-panel layout
      memory-card.tsx                   # MODIFY → sidebar entry row
      execution-detail.tsx              # KEEP — internals reused in task viewer
      report-viewer.tsx                 # KEEP — internals reused in task viewer
      report-grade-form.tsx             # KEEP — embedded in Review view

      # NEW — Board
      board/
        board-view.tsx                  # Main kanban container with DragDropContext
        board-column.tsx                # Single droppable column
        board-card.tsx                  # Draggable task card
        board-filters.tsx              # Filter bar (agent, priority, type, date)
        board-agent-sidebar.tsx         # Agent roster sidebar
        board-task-viewer.tsx           # Unified slide-over (Inbox/InProgress/Review/Done views)

      # NEW — Calendar
      calendar/
        activity-calendar.tsx           # Main calendar container
        calendar-week-grid.tsx          # 7-day CSS grid with task pills
        calendar-always-running.tsx     # High-frequency routines section
        calendar-next-up.tsx            # Upcoming tasks list

      # NEW — Memory (rewrites)
      memory/
        memory-sidebar.tsx              # Left panel: search + journal list
        memory-viewer.tsx               # Right panel: content renderer
        memory-journal-entry.tsx        # Sidebar entry row (date + size)

      # NEW — Projects
      projects/
        activity-projects.tsx           # Project card grid
        project-card.tsx                # Individual project card
        project-detail.tsx              # Project detail with task list

      # NEW — Summary widgets
      widgets/
        command-centre-dashboard.tsx     # KEEP — drag grid container
        active-now-widget.tsx            # KEEP
        recent-activity-widget.tsx       # KEEP — add click-through to Board
        schedule-widget.tsx              # KEEP — add click-through to Calendar
        agent-reports-widget.tsx         # KEEP
        status-overview-widget.tsx       # NEW — donut chart
        priority-breakdown-widget.tsx    # NEW — bar chart
        types-of-work-widget.tsx         # NEW — horizontal bars
        team-workload-widget.tsx         # NEW — agent distribution bars

      # NEW — Global Search
    global-search/
      global-search.tsx                 # Cmd+K overlay
      search-result-item.tsx            # Individual result row

  hooks/
    use-activity-api.ts                 # MODIFY — add board stats, 60s polling
    use-reports-api.ts                  # KEEP
    use-memory-explorer-api.ts          # MODIFY — add group_by, date range
    use-board-sse.ts                    # NEW — SSE hook for real-time board
    use-board-tasks.ts                  # NEW — React Query + SSE hybrid
    use-projects-api.ts                 # NEW — project CRUD hooks
    use-global-search.ts               # NEW — debounced search hook
    use-memory-journal.ts              # NEW — journal day content hook

  types/
    board.ts                            # NEW — BoardTask, BoardEvent types
    project.ts                          # NEW — Project types
    search.ts                           # NEW — SearchResult types

orchestrator/
  api/
    activity.py                         # MODIFY — add board/stream SSE, tasks PATCH, board/stats
    projects.py                         # NEW — project CRUD endpoints
    search.py                           # NEW — global search endpoint
    memory.py                           # MODIFY — add journal/{date} endpoint, group_by param
  core/
    models/
      core.py                           # MODIFY — add Project, ProjectTask models + recipe_execution fields
  services/
    activity_service.py                 # MODIFY — add board stats aggregation
    project_service.py                  # NEW — project business logic
    search_service.py                   # NEW — cross-entity search

---

12. Implementation Phases

Phase 1: Tab Restructure + Summary Widgets (PR1 + PR2)

1. Rename tabs in activity-page.tsx: Summary | Board | Calendar | Memory | Projects

2. Change polling from 15s → 60s across all activity hooks

3. Update deep-link support for new tab names

4. Build 4 new Summary widgets: status donut, priority breakdown, types of work, team workload

5. Add GET /api/activity/board/stats endpoint

6. Wire "View All" click-throughs: Schedule widget → Calendar, Recent Activity → Board

Phase 2: Board — Backend (PR3)

1. Database migration: add priority, tags, assignee_agent_id, review_mode, due_date to recipe_executions

2. Add inbox, assigned, review as valid status values

3. Create PATCH /api/activity/tasks/{id}/status endpoint

4. Create PATCH /api/activity/tasks/{id} endpoint (update fields)

5. Extend GET /api/activity/feed with board status filters

6. Implement SSE endpoint GET /api/activity/board/stream using Redis pub/sub

Phase 3: Board — Frontend (PR4)

1. Install @hello-pangea/dnd

2. Build board-view.tsx — DragDropContext + 5 columns

3. Build board-column.tsx — Droppable column with header + count

4. Build board-card.tsx — Draggable card with task details

5. Build board-agent-sidebar.tsx — agent roster with filter-on-click

6. Build board-filters.tsx — filter bar

7. Implement optimistic drag-and-drop with revert on failure

8. Build use-board-sse.ts hook with reconnect + 60s polling fallback

Phase 4: Task Viewer (PR5)

1. Build board-task-viewer.tsx — slide-over with 4 context-aware views

2. Inbox/Assigned view: editable fields (assignee, priority, due date, tags, review mode)

3. In Progress view: live logs + step pipeline (reuse execution-detail.tsx internals)

4. Review view: report content + grade form + approve/reject (reuse report-viewer.tsx + report-grade-form.tsx)

5. Done view: results + execution summary + re-run/download actions

6. Wire task viewer to board card clicks

Phase 5: Calendar (PR6)

1. Build activity-calendar.tsx — container with view toggle

2. Build calendar-always-running.tsx — sub-hourly routines

3. Build calendar-week-grid.tsx — CSS grid with cron-parsed task pills

4. Build calendar-next-up.tsx — upcoming sorted list

5. Wire calendar events to task viewer slide-over

6. Wire Schedule widget "View All" → Calendar tab

Phase 6: Memory Enhancement (PR7)

1. Verify Mem0 search POST fix is merged

2. Rewrite activity-memory.tsx → two-panel layout

3. Build memory-sidebar.tsx — search + grouped journal list

4. Build memory-viewer.tsx — right panel markdown renderer

5. Add group_by param to GET /api/memory/browse

6. Create GET /api/memory/journal/{date} endpoint

7. Add organize-by toggle (Day / Agent / Type)

8. Enhanced filters: agent, date range, type, sort

Phase 7: Projects (PR8)

1. Database migration: projects + project_tasks tables

2. Create projects.py API with CRUD endpoints

3. Build activity-projects.tsx — card grid with stats header

4. Build project-card.tsx — card with progress bar + agent + priority

5. Build project-detail.tsx — detail view with linked task list

6. Wire "Open in Board" → Board tab filtered by project

Phase 8: Global Search (PR9)

1. Create GET /api/search endpoint — cross-entity search

2. Build global-search.tsx — Cmd+K overlay with keyboard navigation

3. Build search-result-item.tsx — categorized result rows

4. Wire results to navigation (Board task viewer, Memory entry, Agent page, Project detail)

5. Add Cmd+K listener to app layout

Phase 9: Cleanup (PR10)

1. Delete deprecated components: activity-feed.tsx, activity-feed-item.tsx, activity-reports.tsx, report-card.tsx

2. Update all internal links referencing old Feed/Reports tabs

3. Mobile responsive pass for all new components

4. Loading skeletons for Board, Calendar, Memory, Projects

5. Empty states for each tab

6. prefers-reduced-motion compliance

---

13. Dependencies

New npm Packages

Package	Purpose	Size
@hello-pangea/dnd	Drag-and-drop for Kanban board	~45KB gzipped

Existing Packages (Already Installed)

Package	Used For
react-grid-layout	Summary widget drag grid
recharts	Donut, bar charts on Summary
framer-motion	Animations
lucide-react	Icons

No @fullcalendar needed — Calendar is custom CSS grid.

---

14. Success Metrics

Metric	Target	How to Measure
Time to answer "what's running?"	< 3 seconds (one click from sidebar)	User testing
Task status visibility	100% of active tasks visible on Board	Compare DB count vs UI count
Drag-and-drop latency	< 200ms perceived (optimistic update)	Performance testing
SSE delivery latency	< 2s from event to UI update	Instrumentation
Calendar shows all schedules	100% of cron jobs visible	Compare scheduler DB vs Calendar
Memory search returns results	> 0 results for any known topic	Regression test
Global search response time	< 500ms	API monitoring
Page load time	< 1.5s initial, < 500ms tab switch	Lighthouse + RUM
Mobile usability	All tabs functional on 375px width	Manual test

---

15. Open Questions

1. SSE vs WebSocket: SSE is simpler and sufficient for server→client push. If we later need client→server streaming (e.g., collaborative editing), upgrade to WebSocket. For now, SSE.

2. Board pagination: If a column has 100+ cards, do we paginate or virtual-scroll? Recommendation: show latest 50 per column with "Load More" button. Virtual scrolling is a future optimization.

3. Project ↔ Mission relationship: Projects in this PRD are the UI evolution of Missions. The backend projects table is new. Do we migrate existing mission data or start fresh? Recommendation: start fresh — Missions tab was a placeholder with no data.

4. Review mode "LLM": Which agent performs LLM review? Options: (a) dedicated QA agent, (b) any agent with "reviewer" skill, (c) configurable per-task. Recommendation: configurable per-task with a workspace-level default.

5. Calendar — month view? v2 ships with Week and Today views only. Month view is a future enhancement if users request it. Keeps scope manageable.