diff --git a/docs/superpowers/specs/2026-03-29-dashboard-design.md b/docs/superpowers/specs/2026-03-29-dashboard-design.md
new file mode 100644
index 00000000..0146b80c
--- /dev/null
+++ b/docs/superpowers/specs/2026-03-29-dashboard-design.md
@@ -0,0 +1,197 @@
+# Dashboard Design Spec
+
+## Goal
+
+Redesign the Dashboard tab as a progressive drill-down dashboard for Apache Camel application monitoring. Follows the RED method (Rate, Errors, Duration) plus saturation (inflight exchanges). The sidebar drives scope: all applications → single application → single route. Each level answers a focused question with increasing detail.
+
+Business/support users are the primary audience — the dashboard focuses on exchange health, throughput, error rates, and SLA compliance. Ops/infrastructure monitoring stays on the Runtime tab (agent health, JVM metrics). Power users needing custom analysis will use Grafana; this dashboard targets the 80% sweet spot.
+
+## Architecture
+
+The Dashboard tab renders one of three views based on the current sidebar selection:
+
+| Sidebar state | Dashboard level | Question answered |
+|---|---|---|
+| No selection | Level 1: All Applications | Is my landscape healthy? Which app needs attention? |
+| Application selected | Level 2: Application | How is this app performing? Which route is the problem? |
+| Route selected | Level 3: Route | What's happening in this route? Where's the bottleneck? |
+
+All levels share the same time range (controlled by the top bar's time selector) and auto-refresh behavior (LIVE mode with 30s refresh).
+
+## Level 1: All Applications Overview
+
+### KPI Strip (4 metrics)
+
+| Metric | Source | Trend |
+|---|---|---|
+| Total throughput (msg/s) | `stats_1m_all` aggregate | vs previous period |
+| Error rate (%) | failed / total | vs previous period |
+| P99 latency (ms) | `approx_percentile(0.99)` | vs previous period |
+| Inflight exchanges | running count | current value |
+
+Uses the existing `useExecutionStats()` hook with no application/route filter.
+
+### Application Health Table
+
+Columns: App name, Status dot, Throughput (msg/s), Error Rate (%), P99 (ms), Active Routes, Sparkline (12-bucket trend).
+
+**Status dot derivation:**
+- Green: error rate < 1% AND P99 < SLA threshold (300ms)
+- Yellow: error rate 1-5% OR P99 between 200-300ms
+- Red: error rate > 5% OR P99 > SLA threshold
+
+Row click navigates to that application in the sidebar (transitions to Level 2).
+
+Data source: `useRouteMetrics()` aggregated by application. The route-level metrics are grouped by `appId` and aggregated to produce application-level rows.
+
+### Charts (2, side-by-side)
+
+1. **Throughput over time** — Stacked area chart, one series per application. Shows relative volume and total.
+2. **Error rate over time** — Line chart, one line per application. Highlights which app is misbehaving.
+
+Data source: `useStatsTimeseries()` with no application filter, extended to support per-app breakdown. This requires a new API parameter or a new endpoint that returns timeseries grouped by application.
+
+### New API endpoint needed
+
+`GET /api/v1/search/stats/timeseries/by-app` — Returns timeseries buckets grouped by application name. Query: `stats_1m_app` materialized view.
+
+Response shape:
+```json
+{
+  "apps": {
+    "order-service": [ { "time": "...", "totalCount": 42, "failedCount": 1, ... } ],
+    "payment-gateway": [ ... ]
+  }
+}
+```
+
+## Level 2: Single Application
+
+### KPI Strip (4 metrics)
+
+Same four metrics as Level 1 but scoped to the selected application. Uses `useExecutionStats({ application })`.
+
+### Route Performance Table
+
+Columns: Route ID, Throughput (msg/s), Success %, Avg Duration (ms), P99 Duration (ms), Error Rate (%), Sparkline.
+
+Sortable by any column. Row click navigates to that route in the sidebar (transitions to Level 3).
+
+Data source: `useRouteMetrics({ appId })` — already exists and returns per-route data filtered by application.
+
+### Charts (2, side-by-side)
+
+1. **Throughput over time** — Stacked area by route.
+2. **Latency percentiles over time** — P50, P95, P99 lines with SLA threshold (300ms horizontal dashed line).
+
+Data source: `useStatsTimeseries({ application })` for the aggregate latency chart. For per-route throughput breakdown, needs the same by-app pattern extended to by-route:
+
+### New API endpoint needed
+
+`GET /api/v1/search/stats/timeseries/by-route` — Returns timeseries buckets grouped by route ID, filtered by application. Query: `stats_1m_route` materialized view.
+
+Response shape:
+```json
+{
+  "routes": {
+    "process-order": [ { "time": "...", "totalCount": 42, ... } ],
+    "validate-payment": [ ... ]
+  }
+}
+```
+
+### Top 5 Errors
+
+Compact table: Error Type, Route, Count, Last Seen.
+
+Click navigates to the Exchanges tab with the error type pre-filled as a search filter. Section hidden when there are zero errors in the time window.
+
+Data source: PostgreSQL query on `executions` table, aggregating by `error_type` column for the selected application and time range.
+
+### New API endpoint needed
+
+`GET /api/v1/search/errors/top` — Returns top N errors grouped by error type.
+
+Parameters: `application` (optional), `routeId` (optional), `from`, `to`, `limit` (default 5).
+
+Response shape:
+```json
+[
+  { "errorType": "ConnectTimeoutException", "routeId": "validate-payment", "count": 47, "lastSeen": "2026-03-29T15:58:00Z" }
+]
+```
+
+Source: PostgreSQL query on `executions` table. Group by `error_type` column (added in V10 migration). Filter to `status = 'FAILED'` within the time range. Order by count descending, limit to N.
+
+## Level 3: Single Route
+
+### KPI Strip (4 metrics)
+
+Same four metrics scoped to the selected route. Uses `useExecutionStats({ application, routeId })`.
+
+### Charts (3, in a row)
+
+1. **Throughput over time** — Area chart, single series.
+2. **Latency percentiles over time** — P50, P95, P99 lines with SLA threshold.
+3. **Error rate over time** — Area chart, red-tinted.
+
+Data source: `useStatsTimeseries({ application, routeId })` — already exists.
+
+### Compact Process Diagram with Heatmap
+
+Reuses the `ProcessDiagram` component but with a **latency heatmap overlay** instead of the execution overlay. Processor nodes are colored by aggregate performance:
+
+- Color scale: green (fast) → yellow (moderate) → red (slow), computed relative to the route's own processors (not absolute thresholds). The slowest processor in the route gets the warmest color.
+- Data source: `useProcessorMetrics({ routeId, appId })` — already exists. Uses `avgDurationMs` or `p99DurationMs` for the color mapping.
+- No click interactions beyond visual identification. Clicking a node scrolls to its row in the processor table below.
+- Compact sizing: fixed height (~250px), the diagram fits-to-view within that space.
+
+Implementation: a new `heatmapOverlay` prop on `ProcessDiagram` (or a wrapper component) that takes a `Map<processorId, { avgMs, p99Ms }>` and colors nodes accordingly. Reuses the existing diagram layout and rendering — only the fill color logic changes.
+
+### Processor Metrics Table
+
+Columns: Processor ID, Type, Invocation Count, Avg Duration (ms), P99 Duration (ms), Error Rate (%).
+
+Default sort: P99 descending (slowest processor first — highlights bottlenecks).
+
+Data source: `useProcessorMetrics({ routeId, appId })` — already exists.
+
+### Top 5 Errors
+
+Same format as Level 2 but scoped to this route. Uses the same `top-errors` endpoint with `routeId` parameter.
+
+## Data Requirements Summary
+
+### Existing endpoints (no backend changes)
+
+| Endpoint | Used at |
+|---|---|
+| `GET /search/stats` | All levels (KPI strip) |
+| `GET /search/stats/timeseries` | Level 2, Level 3 (charts) |
+| `GET /routes/metrics` | Level 1 (app table, aggregated), Level 2 (route table) |
+| `GET /routes/metrics/processors` | Level 3 (processor table + heatmap) |
+
+### New endpoints needed
+
+| Endpoint | Used at | Source |
+|---|---|---|
+| `GET /search/stats/timeseries/by-app` | Level 1 (charts) | `stats_1m_app` view |
+| `GET /search/stats/timeseries/by-route` | Level 2 (throughput chart) | `stats_1m_route` view |
+| `GET /search/errors/top` | Level 2, Level 3 (top errors) | `executions` table |
+
+### Existing frontend components reused
+
+- `KpiStrip` / `TabKpis` — KPI display with trends
+- `DataTable` — sortable tables
+- `AreaChart`, `LineChart` — time-series charts
+- `Sparkline` — compact trend in table cells
+- `StatusDot` — health indicators
+- `ProcessDiagram` — route visualization (extended with heatmap)
+
+## Scope Exclusions
+
+- No user-customizable panels or drag-and-drop layout (power users use Grafana)
+- No JVM/infrastructure metrics on Dashboard tab (that's Runtime tab)
+- No alerting or threshold configuration (out of scope)
+- No comparison mode (e.g., "this week vs last week" side-by-side)
+- No export/PDF functionality