feat(tui): mouse click for settings sections, column sort, logs panel

- Click Alerts/Nodes/Users sub-tabs in settings to switch sections
- Click STATUS/NAME/LATENCY column headers to sort (click again
  to reverse direction)
- Logs panel show/hide: [l] toggles sidebar visibility, click
  focused panel to hide. Monitors expand to full width when hidden.
- Logs full-screen: Enter when focused on logs opens scrollable
  full-screen log view with severity tags, filter toggle [f],
  and live updates

.env.example

@@ -1,5 +1,5 @@
 # ─── uptop configuration ───────────────────────────────────
-# Copy to .env and edit. Only uncomment what you need.
+# Export in your environment or pass via docker run --env-file.\n# Only uncomment what you need.
 
 # ─── Core ──────────────────────────────────────────────────
 UPTOP_PORT=23234              # SSH server port
@@ -40,3 +40,5 @@ UPTOP_DB_DSN=/data/uptop.db   # File path (SQLite) or connection string (Postgre
 # UPTOP_ALLOW_PRIVATE_TARGETS=false    # Allow monitoring RFC1918/loopback addresses
 # UPTOP_METRICS_PUBLIC=false           # Expose /metrics without auth
 # UPTOP_CORS_ORIGIN=                   # Access-Control-Allow-Origin for /status/json
+# UPTOP_TRUSTED_PROXIES=              # Comma-separated CIDRs/IPs for X-Forwarded-For trust
+# UPTOP_MAINT_RETENTION=168h           # How long ended maintenance windows are kept

.gitea/issue_template/bug_report.yaml

@@ -16,7 +16,7 @@ body:
       label: What happened?
       description: Include what you expected to happen instead.
       placeholder: |
-        When I run `uptop serve`, the TUI crashes after 10 seconds.
+        When I run `uptop`, the TUI crashes after 10 seconds.
         I expected it to keep running and display monitor status.
     validations:
       required: true
@@ -25,7 +25,7 @@ body:
     attributes:
       label: Steps to reproduce
       placeholder: |
-        1. Run `uptop serve`
+        1. Run `uptop`
         2. Wait ~10 seconds
         3. TUI crashes with panic
     validations:
@@ -37,7 +37,7 @@ body:
       description: Output of `uptop version`, OS, terminal. Paste any errors below.
       render: shell
       placeholder: |
-        uptop version 2026.06.1
+        uptop 0.1.0 (abc1234, 2026-06-17)
         OS: Debian 13
         Terminal: Ghostty
 

.gitea/workflows/release-binaries.yml

@@ -49,9 +49,11 @@ jobs:
           version: "~> v2"
           args: release --clean --release-notes=/tmp/release-notes.md
         env:
+          GORELEASER_CURRENT_TAG: ${{ github.ref_name }}
           GORELEASER_FORCE_TOKEN: gitea
           GITEA_TOKEN: ${{ secrets.RELEASE_TOKEN }}
           GITEA_API_URL: http://gitea:3000/api/v1
+          HOMEBREW_TAP_GITHUB_TOKEN: ${{ secrets.HOMEBREW_TAP_GITHUB_TOKEN }}
 
       # GitHub release relaying is handled by .github/workflows/mirror-release.yml,
       # which runs on GitHub Actions when the push mirror delivers the tag and

.gitea/workflows/release-docker.yml

@@ -35,11 +35,16 @@ jobs:
 
           TAGS="lerkolabs/uptop:${TAG}"
           TAGS="${TAGS},lerkolabs/uptop:sha-${SHORT_SHA}"
+          TAGS="${TAGS},ghcr.io/lerkolabs/uptop:${TAG}"
+          TAGS="${TAGS},ghcr.io/lerkolabs/uptop:sha-${SHORT_SHA}"
           # :latest only for real releases — rc rehearsal tags must not move it
           if [ "${{ github.ref_type }}" = "tag" ]; then
             case "$TAG" in
               *-*) ;;
-              *) TAGS="${TAGS},lerkolabs/uptop:latest" ;;
+              *)
+                TAGS="${TAGS},lerkolabs/uptop:latest"
+                TAGS="${TAGS},ghcr.io/lerkolabs/uptop:latest"
+                ;;
             esac
           fi
           echo "tags=$TAGS" >> "$GITHUB_OUTPUT"
@@ -56,6 +61,13 @@ jobs:
           username: ${{ secrets.DOCKERHUB_USERNAME }}
           password: ${{ secrets.DOCKERHUB_TOKEN }}
 
+      - name: Log in to GHCR
+        uses: docker/login-action@v3
+        with:
+          registry: ghcr.io
+          username: ${{ secrets.GHCR_USERNAME }}
+          password: ${{ secrets.GHCR_TOKEN }}
+
       # Scan must gate the push: build amd64 locally, scan it, and only then run
       # the multi-arch push (amd64 layers come from the builder cache, so the
       # second build only adds the arm64 work).

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -0,0 +1,46 @@
+name: Bug Report
+description: Something isn't working as expected
+labels:
+  - bug
+body:
+  - type: checkboxes
+    id: search
+    attributes:
+      label: Before filing
+      options:
+        - label: I searched existing issues and didn't find a match
+          required: true
+  - type: textarea
+    id: description
+    attributes:
+      label: What happened?
+      description: Include what you expected to happen instead.
+      placeholder: |
+        When I run `uptop`, the TUI crashes after 10 seconds.
+        I expected it to keep running and display monitor status.
+    validations:
+      required: true
+  - type: textarea
+    id: reproduction
+    attributes:
+      label: Steps to reproduce
+      placeholder: |
+        1. Run `uptop`
+        2. Wait ~10 seconds
+        3. TUI crashes with panic
+    validations:
+      required: true
+  - type: textarea
+    id: environment
+    attributes:
+      label: Environment & logs
+      description: Output of `uptop version`, OS, terminal. Paste any errors below.
+      render: shell
+      placeholder: |
+        uptop 0.1.0 (abc1234, 2026-06-17)
+        OS: Debian 13
+        Terminal: Ghostty
+
+        [paste any error output here]
+    validations:
+      required: false

.github/ISSUE_TEMPLATE/config.yml

@@ -1,8 +1 @@
 blank_issues_enabled: false
-contact_links:
-  - name: Report a Bug
-    url: https://gitea.lerkolabs.com/lerkolabs/uptop/issues/new?template=bug_report.yaml
-    about: Report bugs on our Gitea instance
-  - name: Request a Feature
-    url: https://gitea.lerkolabs.com/lerkolabs/uptop/issues/new?template=feature_request.yaml
-    about: Suggest features on our Gitea instance

.github/ISSUE_TEMPLATE/feature_request.yml

@@ -0,0 +1,20 @@
+name: Feature Request
+description: Suggest a new feature or enhancement
+labels:
+  - feature
+body:
+  - type: textarea
+    id: problem
+    attributes:
+      label: Problem
+      description: What's frustrating or missing?
+      placeholder: I find myself always needing to ...
+    validations:
+      required: true
+  - type: textarea
+    id: solution
+    attributes:
+      label: Proposed solution
+      description: How would you like this to work?
+    validations:
+      required: false

.github/workflows/forward-issues.yml

@@ -0,0 +1,78 @@
+name: Forward Issues to Gitea
+
+on:
+  issues:
+    types: [opened]
+
+permissions:
+  issues: write
+
+jobs:
+  forward:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Build issue body
+        env:
+          ISSUE_BODY: ${{ github.event.issue.body }}
+          ISSUE_URL: ${{ github.event.issue.html_url }}
+          ISSUE_AUTHOR: ${{ github.event.issue.user.login }}
+        run: |
+          jq -n \
+            --arg url "$ISSUE_URL" \
+            --arg author "$ISSUE_AUTHOR" \
+            --arg body "$ISSUE_BODY" \
+            '">" + " Forwarded from GitHub: " + $url + "\n> Reported by: [@" + $author + "](https://github.com/" + $author + ")\n\n---\n\n" + $body' \
+            -r > /tmp/gitea-issue-body.md || exit 1
+
+      - name: Resolve label IDs
+        id: labels
+        env:
+          GITEA_TOKEN: ${{ secrets.GITEA_TOKEN }}
+          GH_LABELS: ${{ toJSON(github.event.issue.labels.*.name) }}
+        run: |
+          GITEA_LABELS=$(curl -f \
+            -H "Authorization: token ${GITEA_TOKEN}" \
+            "https://gitea.lerkolabs.com/api/v1/repos/lerkolabs/uptop/labels?limit=50") || exit 1
+
+          GH_LABELS_COMPACT=$(echo "$GH_LABELS" | jq -c '.')
+          LABEL_IDS=$(echo "$GITEA_LABELS" | jq -c --argjson gh "$GH_LABELS_COMPACT" '
+            [.[] | select(
+              .name == "github" or
+              (.name as $n | $gh | index($n) != null)
+            ) | .id]
+          ')
+
+          echo "ids=${LABEL_IDS}" >> "$GITHUB_OUTPUT"
+
+      - name: Create Gitea issue
+        id: create
+        env:
+          GITEA_TOKEN: ${{ secrets.GITEA_TOKEN }}
+          ISSUE_TITLE: ${{ github.event.issue.title }}
+          LABEL_IDS: ${{ steps.labels.outputs.ids }}
+        run: |
+          RESPONSE=$(curl -f -X POST \
+            -H "Authorization: token ${GITEA_TOKEN}" \
+            -H "Content-Type: application/json" \
+            "https://gitea.lerkolabs.com/api/v1/repos/lerkolabs/uptop/issues" \
+            -d "$(jq -n \
+              --arg title "$ISSUE_TITLE" \
+              --rawfile body /tmp/gitea-issue-body.md \
+              --argjson labels "$LABEL_IDS" \
+              '{title: $title, body: $body, labels: $labels}'
+            )") || exit 1
+
+          GITEA_URL=$(echo "$RESPONSE" | jq -re '.html_url') || exit 1
+          GITEA_NUM=$(echo "$RESPONSE" | jq -re '.number') || exit 1
+
+          echo "url=${GITEA_URL}" >> "$GITHUB_OUTPUT"
+          echo "number=${GITEA_NUM}" >> "$GITHUB_OUTPUT"
+
+      - name: Comment on GitHub issue
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          GITEA_URL: ${{ steps.create.outputs.url }}
+        run: |
+          gh issue comment "${{ github.event.issue.number }}" \
+            --repo "${{ github.repository }}" \
+            --body "Thanks for reporting! This issue has been forwarded to our [primary tracker](${GITEA_URL}). Discussion and updates will happen there."

.gitignore

@@ -25,4 +25,6 @@ authorized_keys
 tmp
 *.local.json
 *.local.md
+data/
 .env
+vhs

.goreleaser.yaml

@@ -59,6 +59,24 @@ nfpms:
         dst: /usr/share/doc/uptop/LICENSE
         type: doc
 
+brews:
+  - name: uptop
+    repository:
+      owner: lerkolabs
+      name: homebrew-tap
+      token: "{{ .Env.HOMEBREW_TAP_GITHUB_TOKEN }}"
+    url_template: "https://github.com/lerkolabs/uptop/releases/download/{{ .Tag }}/{{ .ArtifactName }}"
+    commit_author:
+      name: lerkolabs-bot
+      email: bot@lerkolabs.com
+    homepage: https://github.com/lerkolabs/uptop
+    description: Self-hosted uptime monitoring with a TUI over SSH
+    license: MIT
+    install: |
+      bin.install "uptop"
+    test: |
+      system bin/"uptop", "version"
+
 # Changelog generation must stay enabled: the --release-notes flag is consumed
 # by the changelog pipe, so disabling it silently drops the git-cliff notes
 # (empty release body on v0.1.0-rc.1). With --release-notes set, GoReleaser

Dockerfile

@@ -23,14 +23,13 @@ RUN mkdir -p /data/.ssh && chown -R uptop:uptop /data
 COPY --from=builder /app/uptop .
 COPY --chmod=755 docker-entrypoint.sh /usr/local/bin/
 
-ENV LIPGLOSS_RENDERER_HAS_DARK_BACKGROUND=true
 ENV UPTOP_DB_TYPE=sqlite
 ENV UPTOP_DB_DSN=/data/uptop.db
 ENV UPTOP_KEYS=/data/authorized_keys
 ENV UPTOP_SSH_HOST_KEY=/data/.ssh/id_ed25519
 ENV UPTOP_PORT=23234
 
-EXPOSE 23234
+EXPOSE 8080 23234
 HEALTHCHECK --interval=30s --timeout=5s --start-period=10s --retries=3 \
   CMD wget -qO- http://localhost:8080/api/health || exit 1
 USER uptop

README.md

@@ -1,16 +1,18 @@
 <div align="center">
-  <h1>uptop</h1>
+  <img src="assets/logo.svg" alt="uptop" width="320">
   <p>Self-hosted uptime monitoring with a TUI over SSH.</p>
   <p>No browser. No client install. Just <code>ssh -p 23234 your-server</code>.</p>
 
   <p>
-    <a href="https://gitea.lerkolabs.com/lerkolabs/uptop/actions/workflows/ci.yml"><img src="https://gitea.lerkolabs.com/lerkolabs/uptop/actions/workflows/ci.yml/badge.svg" alt="CI"></a>
-    <img src="https://img.shields.io/badge/license-MIT-blue" alt="MIT License">
+    <a href="https://gitea.lerkolabs.com/lerkolabs/uptop/actions"><img src="https://gitea.lerkolabs.com/lerkolabs/uptop/actions/workflows/ci.yml/badge.svg" alt="CI"></a>
+    <a href="https://github.com/lerkolabs/uptop/releases/latest"><img src="https://img.shields.io/github/v/release/lerkolabs/uptop" alt="Latest Release"></a>
+    <a href="https://goreportcard.com/report/github.com/lerkolabs/uptop"><img src="https://goreportcard.com/badge/github.com/lerkolabs/uptop" alt="Go Report Card"></a>
     <img src="https://img.shields.io/badge/go-1.26-00ADD8?logo=go&logoColor=white" alt="Go 1.26">
     <img src="https://img.shields.io/docker/pulls/lerkolabs/uptop" alt="Docker Pulls">
+    <img src="https://img.shields.io/badge/license-MIT-blue" alt="MIT License">
   </p>
 
-  <img src="assets/monitors.png" alt="uptop monitors view" width="800">
+  <img src="assets/demo.gif" alt="uptop demo" width="800">
 </div>
 
 ## What is this
@@ -27,7 +29,7 @@ Canonical repo: [gitea.lerkolabs.com/lerkolabs/uptop](https://gitea.lerkolabs.co
 - **10 alert providers** — Discord, Slack, Email, Ntfy, Webhook, Telegram, PagerDuty, Pushover, Gotify, Opsgenie
 - **Config as code** — define monitors in YAML, apply declaratively, version control your setup
 - **HA clustering** — leader/follower with automatic failover
-- **Prometheus metrics** — `/metrics` endpoint, wire it straight to Grafana
+- **Prometheus metrics** — `/metrics` endpoint (`UPTOP_METRICS_PUBLIC=true` to expose without auth)
 - **Public status page** — HTML + JSON, toggle with an env var
 - **SQLite or Postgres** — SQLite for single-node, Postgres for production
 - **Uptime Kuma import** — migrate from Kuma with one command
@@ -38,18 +40,27 @@ Canonical repo: [gitea.lerkolabs.com/lerkolabs/uptop](https://gitea.lerkolabs.co
 
 <table>
   <tr>
-    <td><img src="assets/detail.png" alt="detail panel" width="400"></td>
-    <td><img src="assets/alerts.png" alt="alerts view" width="400"></td>
+    <td><img src="assets/monitors.png" alt="monitors + log sidebar" width="400"></td>
+    <td><img src="assets/detail.png" alt="inline detail panel" width="400"></td>
   </tr>
   <tr>
-    <td><img src="assets/logs.png" alt="logs view" width="400"></td>
+    <td><img src="assets/fulldetail.png" alt="full-screen detail" width="400"></td>
+    <td><img src="assets/settings.png" alt="settings / alerts" width="400"></td>
+  </tr>
+  <tr>
+    <td><img src="assets/maint.png" alt="maintenance windows" width="400"></td>
     <td><img src="assets/nodes.png" alt="cluster nodes" width="400"></td>
   </tr>
-  <tr>
-    <td colspan="2" align="center"><img src="assets/theme.png" alt="theme selection" width="600"></td>
-  </tr>
 </table>
 
+### Themes
+
+Five built-in themes: Flexoki Dark, Tokyo Night, Catppuccin Mocha, Nord, Gruvbox. Press `T` to cycle.
+
+<p align="center">
+  <img src="assets/themes.png" alt="all five themes" width="800">
+</p>
+
 ## Quick start
 
 ```bash
@@ -146,6 +157,8 @@ Full reference in [docs/config-as-code.md](docs/config-as-code.md).
 | `UPTOP_INSECURE_SKIP_VERIFY` | `false` | Skip TLS verification for checks |
 | `UPTOP_ALLOW_PRIVATE_TARGETS` | `false` | Allow monitoring RFC1918/loopback addresses |
 | `UPTOP_ADMIN_KEY` | | SSH public key seeded as first admin on startup |
+| `UPTOP_METRICS_PUBLIC` | `false` | Expose `/metrics` without auth |
+| `UPTOP_MAINT_RETENTION` | `168h` | How long ended maintenance windows are kept |
 | `UPTOP_TRUSTED_PROXIES` | | Comma-separated CIDRs/IPs whose `X-Forwarded-For` is trusted ([details](#running-behind-a-reverse-proxy)) |
 
 See [`.env.example`](.env.example) for all options including TLS, probes, and advanced settings.
@@ -179,7 +192,7 @@ uptop prunes its own history in the background — no external cleanup jobs need
 | Check history | newest 1,000 checks per monitor |
 | State changes (UP/DOWN transitions) | newest 5,000 per monitor |
 | Logs | newest 200 entries |
-| Maintenance windows | 7 days after they end (configurable) |
+| Maintenance windows | 7 days after they end (`UPTOP_MAINT_RETENTION`) |
 
 Sparklines, uptime percentages, and SLA reports are computed from these windows, so very long-horizon stats aren't retained. Export to Prometheus via `/metrics` if you need unlimited history.
 

assets/logo.svg

@@ -0,0 +1,10 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 320 80" fill="none">
+  <!-- Terminal-inspired text mark for uptop -->
+  <rect width="320" height="80" rx="8" fill="#1C1B1A"/>
+  <!-- Prompt caret -->
+  <text x="16" y="52" font-family="'JetBrains Mono','Fira Code','SF Mono',monospace" font-size="36" font-weight="700" fill="#3AA99F">▲</text>
+  <!-- "uptop" in monospace -->
+  <text x="52" y="52" font-family="'JetBrains Mono','Fira Code','SF Mono',monospace" font-size="36" font-weight="700" fill="#CECDC3">uptop</text>
+  <!-- Status dot -->
+  <circle cx="296" cy="40" r="6" fill="#879A39"/>
+</svg>

cmd/uptop/keycache_test.go

@@ -43,10 +43,10 @@ func TestKeyCache_AllowsKnownDeniesUnknown(t *testing.T) {
 	_, unknown := testKey(t)
 	kc := newKeyCache(&kcMockStore{users: []models.User{{PublicKey: authorized}}})
 
-	if !kc.IsAllowed(known) {
+	if !kc.IsAllowed(context.Background(), known) {
 		t.Error("known key denied")
 	}
-	if kc.IsAllowed(unknown) {
+	if kc.IsAllowed(context.Background(), unknown) {
 		t.Error("unknown key allowed")
 	}
 }
@@ -56,7 +56,7 @@ func TestKeyCache_RetainsKeysOnRefreshError(t *testing.T) {
 	ms := &kcMockStore{users: []models.User{{PublicKey: authorized}}}
 	kc := newKeyCache(ms)
 
-	if !kc.IsAllowed(known) {
+	if !kc.IsAllowed(context.Background(), known) {
 		t.Fatal("known key denied on first refresh")
 	}
 
@@ -67,7 +67,7 @@ func TestKeyCache_RetainsKeysOnRefreshError(t *testing.T) {
 	kc.updated = time.Now().Add(-time.Hour)
 	kc.mu.Unlock()
 
-	if !kc.IsAllowed(known) {
+	if !kc.IsAllowed(context.Background(), known) {
 		t.Error("transient refresh error locked out a previously valid key")
 	}
 }
@@ -77,7 +77,7 @@ func TestKeyCache_FailsClosedAfterInvalidate(t *testing.T) {
 	ms := &kcMockStore{users: []models.User{{PublicKey: authorized}}}
 	kc := newKeyCache(ms)
 
-	if !kc.IsAllowed(known) {
+	if !kc.IsAllowed(context.Background(), known) {
 		t.Fatal("known key denied on first refresh")
 	}
 
@@ -86,7 +86,7 @@ func TestKeyCache_FailsClosedAfterInvalidate(t *testing.T) {
 	ms.err = errors.New("db down")
 	kc.Invalidate()
 
-	if kc.IsAllowed(known) {
+	if kc.IsAllowed(context.Background(), known) {
 		t.Error("revoked key still allowed while DB is down — fails open")
 	}
 }
@@ -97,7 +97,7 @@ func TestUserInvalidatingStore_DeleteDropsKeyCache(t *testing.T) {
 	kc := newKeyCache(ms)
 	s := &userInvalidatingStore{Store: ms, kc: kc}
 
-	if !kc.IsAllowed(known) {
+	if !kc.IsAllowed(context.Background(), known) {
 		t.Fatal("known key denied on first refresh")
 	}
 
@@ -109,7 +109,7 @@ func TestUserInvalidatingStore_DeleteDropsKeyCache(t *testing.T) {
 	ms.users = nil
 	ms.err = errors.New("db down")
 
-	if kc.IsAllowed(known) {
+	if kc.IsAllowed(context.Background(), known) {
 		t.Error("deleted user's key still allowed from stale cache")
 	}
 }

cmd/uptop/main.go

@@ -52,7 +52,14 @@ func init() {
 		return
 	}
 	if mv := info.Main.Version; mv != "" && mv != "(devel)" {
-		version = strings.TrimPrefix(mv, "v")
+		mv = strings.TrimPrefix(mv, "v")
+		// Pseudo-versions (e.g. "0.1.1-0.20260620165311-5ca534b0b100+dirty")
+		// are noisy in the TUI footer. Extract just the base semver.
+		if i := strings.Index(mv, "-0."); i > 0 {
+			mv = mv[:i]
+		}
+		mv = strings.TrimSuffix(mv, "+dirty")
+		version = mv
 	}
 	for _, s := range info.Settings {
 		switch s.Key {
@@ -77,17 +84,37 @@ func main() {
 		case "export":
 			runExport(os.Args[2:])
 			return
-		case "version", "--version", "-v":
+		case "version", "--version", "-v", "-version":
 			printVersion()
 			return
 		case "migrate-secrets":
 			runMigrateSecrets(os.Args[2:])
 			return
+		case "help", "--help", "-h":
+			printUsage()
+			return
+		case "serve":
+			runServe(os.Args[2:])
+			return
 		}
 	}
 	runServe(os.Args[1:])
 }
 
+func printUsage() {
+	fmt.Fprintf(os.Stderr, `Usage: uptop <command> [flags]
+
+Commands:
+  serve             Start the server (default if no command given)
+  apply             Apply monitors from a YAML file
+  export            Export monitors to YAML
+  migrate-secrets   Re-encrypt alert credentials with current key
+  version           Print version and exit
+
+Run 'uptop serve --help' for server flags.
+`)
+}
+
 func printVersion() {
 	out := "uptop " + version
 	var meta []string
@@ -152,7 +179,7 @@ func parseTrustedProxies(raw string) []*net.IPNet {
 	return cidrs
 }
 
-func openStore(dbType, dsn string) store.Store {
+func openStore(ctx context.Context, dbType, dsn string) store.Store {
 	var ss *store.SQLStore
 	var err error
 	if dbType == "postgres" {
@@ -174,7 +201,7 @@ func openStore(dbType, dsn string) store.Store {
 	} else {
 		slog.Warn("no UPTOP_ENCRYPTION_KEY set, alert credentials stored unencrypted")
 	}
-	if err := ss.Init(context.Background()); err != nil {
+	if err := ss.Init(ctx); err != nil {
 		slog.Error("database init failed", "err", err)
 		os.Exit(1)
 	}
@@ -186,7 +213,7 @@ func runApply(args []string) {
 	filePath := fs.String("f", "", "Path to YAML config file (required)")
 	dryRun := fs.Bool("dry-run", false, "Show planned changes without applying")
 	prune := fs.Bool("prune", false, "Delete monitors/alerts not in YAML")
-	dbType := fs.String("db-type", envOrDefault("UPTOP_DB_TYPE", "sqlite"), "Database type")
+	dbType := fs.String("db-type", envOrDefault("UPTOP_DB_TYPE", "sqlite"), "Database type (sqlite or postgres)")
 	dsn := fs.String("dsn", envOrDefault("UPTOP_DB_DSN", "uptop.db"), "Database DSN")
 	_ = fs.Parse(args) // ExitOnError: parse errors exit before returning
 
@@ -196,7 +223,8 @@ func runApply(args []string) {
 		os.Exit(1)
 	}
 
-	s := openStore(*dbType, *dsn)
+	ctx := context.Background()
+	s := openStore(ctx, *dbType, *dsn)
 
 	f, err := config.LoadFile(*filePath)
 	if err != nil {
@@ -204,7 +232,7 @@ func runApply(args []string) {
 		os.Exit(1)
 	}
 
-	changes, err := config.Apply(context.Background(), s, f, config.ApplyOpts{
+	changes, err := config.Apply(ctx, s, f, config.ApplyOpts{
 		DryRun: *dryRun,
 		Prune:  *prune,
 	})
@@ -219,13 +247,14 @@ func runApply(args []string) {
 func runExport(args []string) {
 	fs := flag.NewFlagSet("export", flag.ExitOnError)
 	outPath := fs.String("o", "-", "Output file path (- for stdout)")
-	dbType := fs.String("db-type", envOrDefault("UPTOP_DB_TYPE", "sqlite"), "Database type")
+	dbType := fs.String("db-type", envOrDefault("UPTOP_DB_TYPE", "sqlite"), "Database type (sqlite or postgres)")
 	dsn := fs.String("dsn", envOrDefault("UPTOP_DB_DSN", "uptop.db"), "Database DSN")
 	_ = fs.Parse(args) // ExitOnError: parse errors exit before returning
 
-	s := openStore(*dbType, *dsn)
+	ctx := context.Background()
+	s := openStore(ctx, *dbType, *dsn)
 
-	f, err := config.Export(context.Background(), s)
+	f, err := config.Export(ctx, s)
 	if err != nil {
 		slog.Error("export failed", "err", err)
 		os.Exit(1)
@@ -239,7 +268,7 @@ func runExport(args []string) {
 
 func runMigrateSecrets(args []string) {
 	fs := flag.NewFlagSet("migrate-secrets", flag.ExitOnError)
-	dbType := fs.String("db-type", envOrDefault("UPTOP_DB_TYPE", "sqlite"), "Database type")
+	dbType := fs.String("db-type", envOrDefault("UPTOP_DB_TYPE", "sqlite"), "Database type (sqlite or postgres)")
 	dsn := fs.String("dsn", envOrDefault("UPTOP_DB_DSN", "uptop.db"), "Database DSN")
 	_ = fs.Parse(args)
 
@@ -254,6 +283,8 @@ func runMigrateSecrets(args []string) {
 		os.Exit(1)
 	}
 
+	ctx := context.Background()
+
 	var ss *store.SQLStore
 	if *dbType == "postgres" {
 		ss, err = store.NewPostgresStore(*dsn)
@@ -264,21 +295,21 @@ func runMigrateSecrets(args []string) {
 		slog.Error("database connection failed", "err", err)
 		os.Exit(1)
 	}
-	if err := ss.Init(context.Background()); err != nil {
+	if err := ss.Init(ctx); err != nil {
 		slog.Error("database init failed", "err", err)
 		os.Exit(1)
 	}
 
 	ss.SetEncryptor(enc)
 
-	alerts, err := ss.GetAllAlerts(context.Background())
+	alerts, err := ss.GetAllAlerts(ctx)
 	if err != nil {
 		slog.Error("failed to load alerts", "err", err)
 		os.Exit(1)
 	}
 	migrated := 0
 	for _, a := range alerts {
-		if err := ss.UpdateAlert(context.Background(), a.ID, a.Name, a.Type, a.Settings); err != nil {
+		if err := ss.UpdateAlert(ctx, a.ID, a.Name, a.Type, a.Settings); err != nil {
 			slog.Error("alert migration failed", "alert", a.Name, "err", err)
 			os.Exit(1)
 		}
@@ -331,7 +362,7 @@ func runServe(args []string) {
 
 	fs := flag.NewFlagSet("serve", flag.ExitOnError)
 	port := fs.Int("port", cfg.Port, "SSH Port")
-	flagDBType := fs.String("db-type", cfg.DBType, "Database type")
+	flagDBType := fs.String("db-type", cfg.DBType, "Database type (sqlite or postgres)")
 	flagDSN := fs.String("dsn", cfg.DBDSN, "Database DSN")
 	demo := fs.Bool("demo", false, "Seed demo data")
 	importKuma := fs.String("import-kuma", "", "Import Uptime Kuma backup JSON file")
@@ -365,15 +396,19 @@ func runServe(args []string) {
 
 	kc := newKeyCache(ss)
 	var s store.Store = &userInvalidatingStore{Store: ss, kc: kc}
-	if err := s.Init(context.Background()); err != nil {
+
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	if err := s.Init(ctx); err != nil {
 		slog.Error("database init failed", "err", err)
 		os.Exit(1)
 	}
 	if *demo {
-		seedDemoData(s)
+		seedDemoData(ctx, s)
 	}
 
-	seedKeysFromEnv(s)
+	seedKeysFromEnv(ctx, s)
 
 	if *importKuma != "" {
 		kb, err := importer.LoadKumaFile(*importKuma)
@@ -382,7 +417,7 @@ func runServe(args []string) {
 			os.Exit(1)
 		}
 		backup := importer.ConvertKuma(kb)
-		if err := s.ImportData(context.Background(), backup); err != nil {
+		if err := s.ImportData(ctx, backup); err != nil {
 			slog.Error("import failed", "err", err)
 			os.Exit(1)
 		}
@@ -402,12 +437,9 @@ func runServe(args []string) {
 	}
 	eng.SetMaintRetention(cfg.MaintRetention)
 
-	ctx, cancel := context.WithCancel(context.Background())
-	defer cancel()
-
-	eng.InitHistory()
-	eng.InitLogs()
-	eng.InitAlertHealth()
+	eng.InitHistory(ctx)
+	eng.InitLogs(ctx)
+	eng.InitAlertHealth(ctx)
 	eng.Start(ctx)
 
 	localTUI := isatty.IsTerminal(os.Stdout.Fd()) || isatty.IsCygwinTerminal(os.Stdout.Fd())
@@ -423,7 +455,7 @@ func runServe(args []string) {
 	sshSrv := startSSHServer(*port, s, eng, kc)
 
 	if localTUI {
-		p := tea.NewProgram(tui.InitialModel(true, s, eng, version), tea.WithAltScreen(), tea.WithMouseCellMotion())
+		p := tea.NewProgram(tui.InitialModel(ctx, true, s, eng, version), tea.WithAltScreen(), tea.WithMouseCellMotion())
 		if _, err := p.Run(); err != nil {
 			slog.Error("TUI failed", "err", err)
 		}
@@ -457,11 +489,11 @@ func startSSHServer(port int, db store.Store, eng *monitor.Engine, kc *keyCache)
 		wish.WithAddress(fmt.Sprintf(":%d", port)),
 		wish.WithHostKeyPath(envOrDefault("UPTOP_SSH_HOST_KEY", ".ssh/id_ed25519")),
 		wish.WithPublicKeyAuth(func(ctx ssh.Context, key ssh.PublicKey) bool {
-			return kc.IsAllowed(key)
+			return kc.IsAllowed(ctx, key)
 		}),
 		wish.WithMiddleware(
 			bm.Middleware(func(s ssh.Session) (tea.Model, []tea.ProgramOption) {
-				return tui.InitialModel(false, db, eng, version), []tea.ProgramOption{tea.WithAltScreen(), tea.WithMouseCellMotion()}
+				return tui.InitialModel(s.Context(), false, db, eng, version), []tea.ProgramOption{tea.WithAltScreen(), tea.WithMouseCellMotion()}
 			}),
 		),
 	)
@@ -477,8 +509,7 @@ func startSSHServer(port int, db store.Store, eng *monitor.Engine, kc *keyCache)
 	return s
 }
 
-func seedDemoData(s store.Store) {
-	ctx := context.Background()
+func seedDemoData(ctx context.Context, s store.Store) {
 	existing, _ := s.GetSites(ctx)
 	if len(existing) > 0 {
 		return
@@ -539,8 +570,8 @@ func newKeyCache(db store.Store) *keyCache {
 	return &keyCache{db: db, ttl: 30 * time.Second}
 }
 
-func (c *keyCache) refresh() {
-	users, err := c.db.GetAllUsers(context.Background())
+func (c *keyCache) refresh(ctx context.Context) {
+	users, err := c.db.GetAllUsers(ctx)
 	if err != nil {
 		// Keep the previous key set: a transient DB error must not lock every
 		// admin out. Revocation still fails closed because Invalidate clears
@@ -573,13 +604,13 @@ func (c *keyCache) Invalidate() {
 	c.mu.Unlock()
 }
 
-func (c *keyCache) IsAllowed(incomingKey ssh.PublicKey) bool {
+func (c *keyCache) IsAllowed(ctx context.Context, incomingKey ssh.PublicKey) bool {
 	c.mu.RLock()
 	stale := time.Since(c.updated) > c.ttl
 	c.mu.RUnlock()
 
 	if stale {
-		c.refresh()
+		c.refresh(ctx)
 	}
 
 	c.mu.RLock()
@@ -625,8 +656,7 @@ func (s *userInvalidatingStore) ImportData(ctx context.Context, data models.Back
 	return err
 }
 
-func seedKeysFromEnv(s store.Store) {
-	ctx := context.Background()
+func seedKeysFromEnv(ctx context.Context, s store.Store) {
 	var keys []string
 
 	if v := os.Getenv("UPTOP_ADMIN_KEY"); v != "" {

deploy/docker-compose.cluster.yml

@@ -5,6 +5,8 @@ services:
   leader:
     image: lerkolabs/uptop:latest
     container_name: uptop-leader
+    sysctls:
+      - net.ipv4.ping_group_range=0 2147483647
     ports:
       - "23234:23234" # SSH
       - "8080:8080"   # HTTP
@@ -40,6 +42,8 @@ services:
   follower:
     image: lerkolabs/uptop:latest
     container_name: uptop-follower
+    sysctls:
+      - net.ipv4.ping_group_range=0 2147483647
     ports:
       - "23233:23234" # SSH (Mapped to different host port)
       - "8081:8080"   # HTTP (Mapped to different host port)

deploy/docker-compose.dev.yml

@@ -13,7 +13,7 @@ services:
       - UPTOP_DB_TYPE=postgres
       - UPTOP_DB_DSN=postgres://devuser:devpass@postgres:5432/uptop_dev?sslmode=disable
       
-      # --- Web Server Configuration (Phase 4) ---
+      # --- Web Server Configuration ---
       - UPTOP_HTTP_PORT=8080
       - UPTOP_STATUS_ENABLED=true
       - UPTOP_STATUS_TITLE=Dev Infrastructure Status

deploy/docker-compose.probe.yml

@@ -1,6 +1,8 @@
 services:
   leader:
     image: lerkolabs/uptop:latest
+    sysctls:
+      - net.ipv4.ping_group_range=0 2147483647
     environment:
       - UPTOP_CLUSTER_MODE=leader
       - UPTOP_CLUSTER_SECRET=changeme  # EXAMPLE ONLY — rotate before use
@@ -12,6 +14,8 @@ services:
 
   probe-us-east:
     image: lerkolabs/uptop:latest
+    healthcheck:
+      disable: true
     environment:
       - UPTOP_CLUSTER_MODE=probe
       - UPTOP_NODE_ID=us-east-1
@@ -24,6 +28,8 @@ services:
 
   probe-eu-west:
     image: lerkolabs/uptop:latest
+    healthcheck:
+      disable: true
     environment:
       - UPTOP_CLUSTER_MODE=probe
       - UPTOP_NODE_ID=eu-west-1

deploy/docker-compose.yml

@@ -8,6 +8,8 @@ services:
       - ALL
     security_opt:
       - no-new-privileges:true
+    sysctls:
+      - net.ipv4.ping_group_range=0 2147483647
     tmpfs:
       - /tmp
     ports:

docs/clustering.md

@@ -47,13 +47,11 @@ Probes are lightweight, stateless nodes that run checks from different locations
 | Node | Variable | Value |
 |------|----------|-------|
 | Both | `UPTOP_CLUSTER_SECRET` | Same shared secret |
-| Leader | `UPTOP_AGG_STRATEGY` | `any-down`, `majority-down`, or `all-down` |
 | Probe | `UPTOP_CLUSTER_MODE` | `probe` |
 | Probe | `UPTOP_PEER_URL` | Leader's HTTP URL |
 | Probe | `UPTOP_NODE_ID` | Unique identifier (e.g. `probe-us-east`) |
-| Probe | `UPTOP_NODE_REGION` | Region tag matching monitor assignments |
 
-Optional: `UPTOP_NODE_NAME` for a human-readable label in the TUI.
+Optional: `UPTOP_AGG_STRATEGY` (default `any-down`), `UPTOP_NODE_REGION` (omit to match all monitors), `UPTOP_NODE_NAME` (human-readable label in the TUI).
 
 See [`deploy/docker-compose.probe.yml`](../deploy/docker-compose.probe.yml) for a multi-region example.
 
@@ -80,6 +78,6 @@ Set via `UPTOP_AGG_STRATEGY` on the leader.
 
 ## Security
 
-- Set `UPTOP_CLUSTER_SECRET` on all nodes. Without it, cluster API endpoints are unauthenticated.
+- Set `UPTOP_CLUSTER_SECRET` on all nodes. Without it, cluster API endpoints reject all requests (fail closed); only `/api/health` stays open.
 - Secrets are sent in HTTP headers (`X-Uptop-Secret`). Use TLS or a reverse proxy for production.
 - uptop warns on startup if the cluster secret is missing or if cluster mode is active without TLS.

internal/alert/alert.go

@@ -18,7 +18,9 @@ import (
 	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
 )
 
-var alertClient = &http.Client{Timeout: 10 * time.Second}
+const alertHTTPTimeout = 10 * time.Second
+
+var alertClient = &http.Client{Timeout: alertHTTPTimeout}
 
 // sanitizeError strips the request URL from transport errors before they are
 // stored or displayed. *url.Error embeds the full URL, which for several

internal/cluster/cluster.go

@@ -38,14 +38,20 @@ func Start(ctx context.Context, cfg Config, eng *monitor.Engine) {
 	// "probe" mode is handled directly in main.go before cluster.Start is called
 }
 
+const (
+	followerTimeout        = 2 * time.Second
+	leaderFailureThreshold = 3
+	followerRetryInterval  = 5 * time.Second
+)
+
 func runFollowerLoop(ctx context.Context, cfg Config, eng *monitor.Engine) {
-	client := http.Client{Timeout: 2 * time.Second}
+	client := http.Client{Timeout: followerTimeout}
 	failures := 0
-	threshold := 3
+	threshold := leaderFailureThreshold
 
 	for {
 		select {
-		case <-time.After(5 * time.Second):
+		case <-time.After(followerRetryInterval):
 		case <-ctx.Done():
 			return
 		}
@@ -59,7 +65,7 @@ func runFollowerLoop(ctx context.Context, cfg Config, eng *monitor.Engine) {
 		isLeaderHealthy := false
 
 		if err == nil {
-			isLeaderHealthy = resp.StatusCode == 200
+			isLeaderHealthy = resp.StatusCode == http.StatusOK
 			_ = resp.Body.Close()
 		}
 

internal/cluster/probe.go

@@ -26,12 +26,18 @@ type ProbeConfig struct {
 	AllowPrivateTargets bool
 }
 
+const (
+	probeMinInterval     = 10
+	probeDefaultInterval = 30
+	probeAPITimeout      = 10 * time.Second
+)
+
 func RunProbe(ctx context.Context, cfg ProbeConfig) error {
-	if cfg.Interval < 10 {
-		cfg.Interval = 30
+	if cfg.Interval < probeMinInterval {
+		cfg.Interval = probeDefaultInterval
 	}
 
-	apiClient := &http.Client{Timeout: 10 * time.Second}
+	apiClient := &http.Client{Timeout: probeAPITimeout}
 	dial := monitor.SafeDialContext(cfg.AllowPrivateTargets)
 	strictClient := &http.Client{
 		Transport: &http.Transport{

internal/models/models.go

@@ -86,6 +86,11 @@ type ProbeNode struct {
 	Version  string
 }
 
+type LogEntry struct {
+	Message   string
+	CreatedAt time.Time
+}
+
 // AlertHealthRecord is the persisted send health of an alert channel. It lets the
 // "last sent" / health indicators survive restarts instead of resetting to "never".
 type AlertHealthRecord struct {

internal/monitor/alerts.go

@@ -0,0 +1,272 @@
+package monitor
+
+import (
+	"context"
+	"fmt"
+	"time"
+
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/alert"
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+)
+
+type AlertHealth struct {
+	LastSendAt time.Time
+	LastSendOK bool
+	LastError  string
+	SendCount  int
+	FailCount  int
+}
+
+// InitAlertHealth restores persisted alert send health so the dashboard shows real
+// "last sent" / health state on startup instead of resetting every channel to "never".
+func (e *Engine) InitAlertHealth(ctx context.Context) {
+	records, err := e.db.LoadAlertHealth(ctx)
+	if err != nil {
+		return
+	}
+	e.alertHealthMu.Lock()
+	defer e.alertHealthMu.Unlock()
+	for id, r := range records {
+		e.alertHealth[id] = AlertHealth{
+			LastSendAt: r.LastSendAt,
+			LastSendOK: r.LastSendOK,
+			LastError:  r.LastError,
+			SendCount:  r.SendCount,
+			FailCount:  r.FailCount,
+		}
+	}
+}
+
+// handleStatusChange folds a check result into the live state. snap is the
+// stale snapshot the check ran against; the actual mutation is applied onto the
+// CURRENT live entry via applyState, so a concurrent pause / config edit /
+// heartbeat is never reverted by this write. Logs and alerts are emitted after
+// the lock is released, off the critical section.
+func (e *Engine) handleStatusChange(snap models.Site, rawStatus string, code int, latency time.Duration, errorReason string) {
+	if !e.IsActive() {
+		return
+	}
+
+	inMaint := e.isInMaintenance(snap.ID)
+	status := models.Status(rawStatus)
+
+	var (
+		prev, next            models.Status
+		name, typ             string
+		alertID               int
+		failCount, maxRetries int
+		confirmedDown         bool
+		failedCheck           bool
+		downSince             time.Time
+		sslWarnFire           bool
+		sslDays               int
+		skipped               bool
+		changed               bool
+	)
+
+	_, exists := e.applyState(snap.ID, func(s *models.Site) {
+		// A non-UP result computed from a stale snapshot must not override a
+		// heartbeat (or newer check) that landed while we were evaluating.
+		if status != models.StatusUp && s.LastCheck.After(snap.LastCheck) {
+			skipped = true
+			return
+		}
+
+		prev = s.Status
+		name = s.Name
+		typ = s.Type
+		alertID = s.AlertID
+		maxRetries = s.MaxRetries
+		downSince = s.StatusChangedAt
+
+		// Fresh check results (measured by the run against snap).
+		s.StatusCode = code
+		s.Latency = snap.Latency
+		s.LastCheck = snap.LastCheck
+		s.HasSSL = snap.HasSSL
+		s.CertExpiry = snap.CertExpiry
+		s.LastError = errorReason
+		if status == models.StatusUp {
+			s.LastSuccessAt = time.Now()
+			s.LastError = ""
+		}
+
+		// Status + failure-count transition, based on the CURRENT live status.
+		if status == models.StatusUp {
+			s.FailureCount = 0
+			s.Status = models.StatusUp
+		} else {
+			if s.FailureCount <= s.MaxRetries {
+				s.FailureCount++
+			}
+			if s.FailureCount > s.MaxRetries {
+				if s.Status != status {
+					confirmedDown = true
+				}
+				s.Status = status
+				s.FailureCount = s.MaxRetries + 1
+			} else {
+				failedCheck = true
+			}
+		}
+		failCount = s.FailureCount
+
+		if s.Status != prev && prev != models.StatusPending {
+			s.StatusChangedAt = time.Now()
+		} else if s.StatusChangedAt.IsZero() && s.Status != models.StatusPending {
+			s.StatusChangedAt = time.Now()
+		}
+
+		// SSL expiry warning (fresh HasSSL/CertExpiry + config threshold).
+		if typ == "http" && s.CheckSSL && s.HasSSL {
+			days := int(time.Until(s.CertExpiry).Hours() / 24)
+			if days <= s.ExpiryThreshold && !s.SentSSLWarning && status != models.StatusSSLExp {
+				sslWarnFire = true
+				sslDays = days
+				s.SentSSLWarning = true
+			} else if days > s.ExpiryThreshold {
+				s.SentSSLWarning = false
+			}
+		}
+
+		next = s.Status
+		changed = next != prev
+	})
+
+	if !exists || skipped {
+		return
+	}
+
+	e.recordCheck(snap.ID, latency, status == models.StatusUp)
+
+	if confirmedDown {
+		if errorReason != "" {
+			e.AddLog(fmt.Sprintf("Monitor '%s' confirmed DOWN: %s", name, errorReason))
+		} else {
+			e.AddLog(fmt.Sprintf("Monitor '%s' confirmed DOWN", name))
+		}
+	} else if failedCheck {
+		e.AddLog(fmt.Sprintf("Monitor '%s' failed check %d/%d", name, failCount, maxRetries))
+	}
+
+	if changed && prev != models.StatusPending {
+		e.enqueueWrite(writeStateChange{siteID: snap.ID, fromStatus: string(prev), toStatus: string(next), reason: errorReason})
+	}
+
+	if sslWarnFire {
+		if !inMaint {
+			e.triggerAlert(alertID, "SSL WARNING", fmt.Sprintf("SSL for '%s' expires in %d days", name, sslDays))
+		} else {
+			e.AddLog(fmt.Sprintf("SSL warning for '%s' suppressed (maintenance)", name))
+		}
+	}
+
+	if prev == models.StatusUp && next == models.StatusLate {
+		e.AddLog(fmt.Sprintf("Monitor '%s' heartbeat overdue", name))
+	}
+
+	if !prev.IsBroken() && next.IsBroken() && next != models.StatusPending {
+		if inMaint {
+			e.AddLog(fmt.Sprintf("Monitor '%s' is DOWN (alerts suppressed — maintenance)", name))
+		} else {
+			msg := fmt.Sprintf("Monitor '%s' is DOWN (%s)", name, rawStatus)
+			if errorReason != "" {
+				msg = fmt.Sprintf("Monitor '%s' is DOWN: %s", name, errorReason)
+			}
+			if typ == "push" {
+				msg = fmt.Sprintf("Push Monitor '%s' missed heartbeat.", name)
+			}
+			e.triggerAlert(alertID, "🚨 ALERT", msg)
+		}
+	}
+	if prev.IsBroken() && next == models.StatusUp {
+		downDur := ""
+		if !downSince.IsZero() {
+			downDur = fmt.Sprintf(" (was down %s)", fmtDurationShort(time.Since(downSince)))
+		}
+		e.AddLog(fmt.Sprintf("Monitor '%s' recovered%s", name, downDur))
+		if !inMaint {
+			e.triggerAlert(alertID, "✅ RECOVERY", fmt.Sprintf("Monitor '%s' is UP%s", name, downDur))
+		}
+	}
+	if prev == models.StatusLate && next == models.StatusUp && !prev.IsBroken() {
+		e.AddLog(fmt.Sprintf("Monitor '%s' heartbeat arrived (was late)", name))
+	}
+}
+
+func (e *Engine) triggerAlert(alertID int, title, message string) {
+	if alertID <= 0 {
+		return
+	}
+	cfg, err := e.db.GetAlert(e.ctx, alertID)
+	if err != nil {
+		e.AddLog(fmt.Sprintf("Failed to load alert config %d: %v", alertID, err))
+		return
+	}
+	provider := alert.GetProvider(cfg)
+	if provider != nil {
+		go func() {
+			ctx, cancel := context.WithTimeout(e.ctx, alertSendTimeout)
+			defer cancel()
+			if err := provider.Send(ctx, title, message); err != nil {
+				e.AddLog(fmt.Sprintf("Alert send failed (%s): %v", cfg.Name, err))
+				e.recordAlertResult(alertID, false, err.Error())
+			} else {
+				e.recordAlertResult(alertID, true, "")
+			}
+		}()
+	}
+}
+
+func (e *Engine) recordAlertResult(alertID int, ok bool, errMsg string) {
+	e.alertHealthMu.Lock()
+	defer e.alertHealthMu.Unlock()
+	h := e.alertHealth[alertID]
+	h.LastSendAt = time.Now()
+	h.LastSendOK = ok
+	h.SendCount++
+	if ok {
+		h.LastError = ""
+	} else {
+		h.LastError = errMsg
+		h.FailCount++
+	}
+	e.alertHealth[alertID] = h
+
+	// Persist so health survives restarts; DB IO off the alert path.
+	e.enqueueWrite(writeAlertHealth{rec: models.AlertHealthRecord{
+		AlertID:    alertID,
+		LastSendAt: h.LastSendAt,
+		LastSendOK: h.LastSendOK,
+		LastError:  h.LastError,
+		SendCount:  h.SendCount,
+		FailCount:  h.FailCount,
+	}})
+}
+
+func (e *Engine) GetAlertHealth(alertID int) AlertHealth {
+	e.alertHealthMu.RLock()
+	defer e.alertHealthMu.RUnlock()
+	return e.alertHealth[alertID]
+}
+
+func (e *Engine) TestAlert(ctx context.Context, alertID int) error {
+	cfg, err := e.db.GetAlert(ctx, alertID)
+	if err != nil {
+		return fmt.Errorf("failed to load alert: %w", err)
+	}
+	provider := alert.GetProvider(cfg)
+	if provider == nil {
+		return fmt.Errorf("no provider for type %q", cfg.Type)
+	}
+	ctx, cancel := context.WithTimeout(ctx, alertSendTimeout)
+	defer cancel()
+	err = provider.Send(ctx, "🧪 Test Alert", fmt.Sprintf("Test notification from uptop for channel '%s'.", cfg.Name))
+	if err != nil {
+		e.recordAlertResult(alertID, false, err.Error())
+		return err
+	}
+	e.recordAlertResult(alertID, true, "")
+	e.AddLog(fmt.Sprintf("Test alert sent to '%s'", cfg.Name))
+	return nil
+}

internal/monitor/alerts_test.go

@@ -0,0 +1,487 @@
+package monitor
+
+import (
+	"context"
+	"testing"
+	"time"
+
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+)
+
+// --- Group 1: State Machine ---
+
+func TestHandleStatusChange_PendingToUp(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", MaxRetries: 3, AlertID: 1},
+		SiteState:  models.SiteState{Status: "PENDING"},
+	}
+	injectSite(e, site)
+
+	e.handleStatusChange(site, "UP", 200, 10*time.Millisecond, "")
+
+	s, _ := getSite(e, 1)
+	if s.Status != "UP" {
+		t.Errorf("expected UP, got %s", s.Status)
+	}
+	if s.FailureCount != 0 {
+		t.Errorf("expected FailureCount 0, got %d", s.FailureCount)
+	}
+	waitAsync()
+	if len(ms.getAlertCallsSnapshot()) != 0 {
+		t.Error("expected no alert for PENDING→UP")
+	}
+}
+
+func TestHandleStatusChange_UpIncrementFailure(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", MaxRetries: 3},
+		SiteState:  models.SiteState{Status: "UP", FailureCount: 0},
+	}
+	injectSite(e, site)
+
+	e.handleStatusChange(site, "DOWN", 500, 0, "test error")
+
+	s, _ := getSite(e, 1)
+	if s.Status != "UP" {
+		t.Errorf("expected UP (under retry threshold), got %s", s.Status)
+	}
+	if s.FailureCount != 1 {
+		t.Errorf("expected FailureCount 1, got %d", s.FailureCount)
+	}
+}
+
+func TestHandleStatusChange_UpToDown_ExceedsRetries(t *testing.T) {
+	ms := newMockStore()
+	ms.alerts[1] = models.AlertConfig{ID: 1, Name: "discord", Type: "webhook", Settings: map[string]string{"url": "http://example.com"}}
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", MaxRetries: 2, AlertID: 1},
+		SiteState:  models.SiteState{Status: "UP", FailureCount: 2},
+	}
+	injectSite(e, site)
+
+	e.handleStatusChange(site, "DOWN", 500, 0, "test error")
+
+	s, _ := getSite(e, 1)
+	if s.Status != "DOWN" {
+		t.Errorf("expected DOWN, got %s", s.Status)
+	}
+	if s.FailureCount != 3 {
+		t.Errorf("expected FailureCount 3, got %d", s.FailureCount)
+	}
+	waitAsync()
+	calls := ms.getAlertCallsSnapshot()
+	if len(calls) == 0 || calls[0] != 1 {
+		t.Errorf("expected alert call for alertID 1, got %v", calls)
+	}
+}
+
+func TestHandleStatusChange_UpToDown_ZeroRetries(t *testing.T) {
+	ms := newMockStore()
+	ms.alerts[1] = models.AlertConfig{ID: 1, Name: "test", Type: "webhook", Settings: map[string]string{"url": "http://example.com"}}
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", MaxRetries: 0, AlertID: 1},
+		SiteState:  models.SiteState{Status: "UP", FailureCount: 0},
+	}
+	injectSite(e, site)
+
+	e.handleStatusChange(site, "DOWN", 0, 0, "test error")
+
+	s, _ := getSite(e, 1)
+	if s.Status != "DOWN" {
+		t.Errorf("expected DOWN, got %s", s.Status)
+	}
+	waitAsync()
+	if len(ms.getAlertCallsSnapshot()) == 0 {
+		t.Error("expected alert on immediate DOWN")
+	}
+}
+
+func TestHandleStatusChange_DownToUp_Recovery(t *testing.T) {
+	ms := newMockStore()
+	ms.alerts[1] = models.AlertConfig{ID: 1, Name: "test", Type: "webhook", Settings: map[string]string{"url": "http://example.com"}}
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", AlertID: 1},
+		SiteState:  models.SiteState{Status: "DOWN", FailureCount: 4},
+	}
+	injectSite(e, site)
+
+	e.handleStatusChange(site, "UP", 200, 5*time.Millisecond, "")
+
+	s, _ := getSite(e, 1)
+	if s.Status != "UP" {
+		t.Errorf("expected UP, got %s", s.Status)
+	}
+	if s.FailureCount != 0 {
+		t.Errorf("expected FailureCount 0, got %d", s.FailureCount)
+	}
+	waitAsync()
+	if len(ms.getAlertCallsSnapshot()) == 0 {
+		t.Error("expected recovery alert")
+	}
+}
+
+func TestHandleStatusChange_DownStaysDown(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", MaxRetries: 2},
+		SiteState:  models.SiteState{Status: "DOWN", FailureCount: 3},
+	}
+	injectSite(e, site)
+
+	e.handleStatusChange(site, "DOWN", 0, 0, "test error")
+
+	s, _ := getSite(e, 1)
+	if s.Status != "DOWN" {
+		t.Errorf("expected DOWN, got %s", s.Status)
+	}
+	waitAsync()
+	if len(ms.getAlertCallsSnapshot()) != 0 {
+		t.Error("expected no re-alert for already DOWN")
+	}
+}
+
+func TestHandleStatusChange_SSLExpired(t *testing.T) {
+	ms := newMockStore()
+	ms.alerts[1] = models.AlertConfig{ID: 1, Name: "test", Type: "webhook", Settings: map[string]string{"url": "http://example.com"}}
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", MaxRetries: 0, AlertID: 1},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	injectSite(e, site)
+
+	e.handleStatusChange(site, "SSL EXP", 0, 0, "SSL certificate expired")
+
+	s, _ := getSite(e, 1)
+	if s.Status != "SSL EXP" {
+		t.Errorf("expected SSL EXP, got %s", s.Status)
+	}
+	waitAsync()
+	if len(ms.getAlertCallsSnapshot()) == 0 {
+		t.Error("expected alert on SSL EXP")
+	}
+}
+
+func TestHandleStatusChange_AlertSuppressedMaintenance(t *testing.T) {
+	ms := newMockStore()
+	ms.maintenance[1] = true
+	ms.alerts[1] = models.AlertConfig{ID: 1, Name: "test", Type: "webhook", Settings: map[string]string{"url": "http://example.com"}}
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", MaxRetries: 0, AlertID: 1},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	injectSite(e, site)
+	e.refreshMaintenanceCache(context.Background())
+
+	e.handleStatusChange(site, "DOWN", 0, 0, "test error")
+
+	s, _ := getSite(e, 1)
+	if s.Status != "DOWN" {
+		t.Errorf("expected DOWN, got %s", s.Status)
+	}
+	waitAsync()
+	if len(ms.getAlertCallsSnapshot()) != 0 {
+		t.Error("expected no alert during maintenance")
+	}
+	logs := e.GetLogs()
+	found := false
+	for _, l := range logs {
+		if containsStr(l.Message, "suppressed") {
+			found = true
+			break
+		}
+	}
+	if !found {
+		t.Error("expected log mentioning suppressed")
+	}
+}
+
+func TestHandleStatusChange_RecoverySuppressedMaintenance(t *testing.T) {
+	ms := newMockStore()
+	ms.maintenance[1] = true
+	ms.alerts[1] = models.AlertConfig{ID: 1, Name: "test", Type: "webhook", Settings: map[string]string{"url": "http://example.com"}}
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", AlertID: 1},
+		SiteState:  models.SiteState{Status: "DOWN"},
+	}
+	injectSite(e, site)
+	e.refreshMaintenanceCache(context.Background())
+
+	e.handleStatusChange(site, "UP", 200, 0, "")
+
+	s, _ := getSite(e, 1)
+	if s.Status != "UP" {
+		t.Errorf("expected UP, got %s", s.Status)
+	}
+	waitAsync()
+	if len(ms.getAlertCallsSnapshot()) != 0 {
+		t.Error("expected no alert during maintenance recovery")
+	}
+}
+
+func TestHandleStatusChange_SSLWarning(t *testing.T) {
+	ms := newMockStore()
+	ms.alerts[1] = models.AlertConfig{ID: 1, Name: "test", Type: "webhook", Settings: map[string]string{"url": "http://example.com"}}
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", Type: "http", CheckSSL: true, ExpiryThreshold: 30, AlertID: 1},
+		SiteState:  models.SiteState{Status: "UP", HasSSL: true, SentSSLWarning: false, CertExpiry: time.Now().Add(15 * 24 * time.Hour)},
+	}
+	injectSite(e, site)
+
+	e.handleStatusChange(site, "UP", 200, 0, "")
+
+	s, _ := getSite(e, 1)
+	if !s.SentSSLWarning {
+		t.Error("expected SentSSLWarning=true")
+	}
+	waitAsync()
+	if len(ms.getAlertCallsSnapshot()) == 0 {
+		t.Error("expected SSL warning alert")
+	}
+}
+
+func TestHandleStatusChange_SSLWarningNotRepeated(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", Type: "http", CheckSSL: true, ExpiryThreshold: 30, AlertID: 1},
+		SiteState:  models.SiteState{Status: "UP", HasSSL: true, SentSSLWarning: true, CertExpiry: time.Now().Add(15 * 24 * time.Hour)},
+	}
+	injectSite(e, site)
+
+	e.handleStatusChange(site, "UP", 200, 0, "")
+
+	waitAsync()
+	if len(ms.getAlertCallsSnapshot()) != 0 {
+		t.Error("expected no repeat SSL warning")
+	}
+}
+
+func TestHandleStatusChange_SSLWarningReset(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", Type: "http", CheckSSL: true, ExpiryThreshold: 30},
+		SiteState:  models.SiteState{Status: "UP", HasSSL: true, SentSSLWarning: true, CertExpiry: time.Now().Add(60 * 24 * time.Hour)},
+	}
+	injectSite(e, site)
+
+	e.handleStatusChange(site, "UP", 200, 0, "")
+
+	s, _ := getSite(e, 1)
+	if s.SentSSLWarning {
+		t.Error("expected SentSSLWarning reset to false")
+	}
+}
+
+func TestHandleStatusChange_SSLWarningSuppressedMaint(t *testing.T) {
+	ms := newMockStore()
+	ms.maintenance[1] = true
+	ms.alerts[1] = models.AlertConfig{ID: 1, Name: "test", Type: "webhook", Settings: map[string]string{"url": "http://example.com"}}
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", Type: "http", CheckSSL: true, ExpiryThreshold: 30, AlertID: 1},
+		SiteState:  models.SiteState{Status: "UP", HasSSL: true, SentSSLWarning: false, CertExpiry: time.Now().Add(15 * 24 * time.Hour)},
+	}
+	injectSite(e, site)
+	e.refreshMaintenanceCache(context.Background())
+
+	e.handleStatusChange(site, "UP", 200, 0, "")
+
+	s, _ := getSite(e, 1)
+	if !s.SentSSLWarning {
+		t.Error("expected SentSSLWarning=true even in maintenance")
+	}
+	waitAsync()
+	if len(ms.getAlertCallsSnapshot()) != 0 {
+		t.Error("expected no alert during maintenance")
+	}
+}
+
+func TestHandleStatusChange_InactiveEngine(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", MaxRetries: 0},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	injectSite(e, site)
+	e.SetActive(false)
+
+	e.handleStatusChange(site, "DOWN", 0, 0, "test error")
+
+	s, _ := getSite(e, 1)
+	if s.Status != "UP" {
+		t.Error("expected no state change when inactive")
+	}
+}
+
+// --- Group 10: liveState merge (lost-update race) ---
+
+// A pause that lands while a check is in flight must survive the check's
+// write-back. The old code snapshotted the site, ran the check, then wrote the
+// whole stale struct back — reverting the pause.
+func TestHandleStatusChange_PauseDuringCheckSurvives(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", MaxRetries: 0},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	injectSite(e, site)
+
+	// `site` is the stale snapshot the check ran against (Paused=false).
+	// Meanwhile the user pauses the monitor.
+	e.ToggleSitePause(1)
+
+	// Check completes and folds its result in using the stale snapshot.
+	e.handleStatusChange(site, "DOWN", 500, 0, "boom")
+
+	s, _ := getSite(e, 1)
+	if !s.Paused {
+		t.Error("pause was reverted by a stale check write-back")
+	}
+	if s.Status != "DOWN" {
+		t.Errorf("expected check result still applied (DOWN), got %s", s.Status)
+	}
+}
+
+// A config edit that lands while a check is in flight must survive; the check
+// must not resurrect the old config from its snapshot.
+func TestHandleStatusChange_ConfigEditDuringCheckSurvives(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", URL: "http://old.com", Type: "http", MaxRetries: 0, Interval: 30},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	injectSite(e, site)
+
+	// Config changes mid-check.
+	e.UpdateSiteConfig(models.SiteConfig{ID: 1, Name: "test", URL: "http://new.com", Type: "http", Interval: 60})
+
+	// Stale check (ran against http://old.com) folds its result in.
+	e.handleStatusChange(site, "UP", 200, 5*time.Millisecond, "")
+
+	s, _ := getSite(e, 1)
+	if s.URL != "http://new.com" {
+		t.Errorf("config edit reverted: URL=%s", s.URL)
+	}
+	if s.Interval != 60 {
+		t.Errorf("config edit reverted: Interval=%d", s.Interval)
+	}
+}
+
+// The classic push false-DOWN: a heartbeat marks the monitor UP while a
+// staleness evaluation (computed from the older LastCheck) is mid-flight.
+// The stale DOWN must not overwrite the fresh heartbeat.
+func TestHandleStatusChange_HeartbeatNotOverwrittenByStaleDown(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	// Snapshot the engine would have taken before evaluating staleness:
+	// LastCheck is old, so checkPush decided "DOWN".
+	snap := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "push", Type: "push", Token: "tok", Interval: 10},
+		SiteState:  models.SiteState{Status: "UP", LastCheck: time.Now().Add(-120 * time.Second)},
+	}
+	injectSite(e, snap)
+
+	// A heartbeat lands first, advancing LastCheck and confirming UP.
+	if !e.RecordHeartbeat("tok") {
+		t.Fatal("heartbeat rejected")
+	}
+
+	// Now the in-flight stale evaluation tries to write DOWN.
+	e.handleStatusChange(snap, "DOWN", 0, 0, "heartbeat missed")
+
+	s, _ := getSite(e, 1)
+	if s.Status != "UP" {
+		t.Errorf("stale DOWN overwrote a fresh heartbeat: status=%s", s.Status)
+	}
+}
+
+// A check result for a site removed mid-check must be dropped, not recreate it.
+func TestHandleStatusChange_RemovedSiteDropped(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", MaxRetries: 0},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	injectSite(e, site)
+
+	e.RemoveSite(1)
+	e.handleStatusChange(site, "DOWN", 500, 0, "boom")
+
+	if _, ok := getSite(e, 1); ok {
+		t.Error("removed site was recreated by a late check write-back")
+	}
+}
+
+// --- Group 12: Phase 3 engine correctness ---
+
+// PENDING→DOWN must honor MaxRetries instead of alerting on first failure.
+func TestHandleStatusChange_PendingRetriesBeforeDown(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "new-monitor", MaxRetries: 2},
+		SiteState:  models.SiteState{Status: "PENDING"},
+	}
+	injectSite(e, site)
+
+	e.handleStatusChange(site, "DOWN", 0, 0, "timeout")
+	s, _ := getSite(e, 1)
+	if s.Status != "PENDING" {
+		t.Errorf("expected PENDING during retry, got %s", s.Status)
+	}
+	if s.FailureCount != 1 {
+		t.Errorf("expected FailureCount 1, got %d", s.FailureCount)
+	}
+
+	e.handleStatusChange(s, "DOWN", 0, 0, "timeout")
+	s, _ = getSite(e, 1)
+	if s.Status != "PENDING" {
+		t.Errorf("expected PENDING during retry 2, got %s", s.Status)
+	}
+
+	e.handleStatusChange(s, "DOWN", 0, 0, "timeout")
+	s, _ = getSite(e, 1)
+	if s.Status != "DOWN" {
+		t.Errorf("expected DOWN after retries exhausted, got %s", s.Status)
+	}
+}
+
+// LATE→DOWN must also honor MaxRetries.
+func TestHandleStatusChange_LateRetriesBeforeDown(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "push-mon", MaxRetries: 1},
+		SiteState:  models.SiteState{Status: "LATE"},
+	}
+	injectSite(e, site)
+
+	e.handleStatusChange(site, "DOWN", 0, 0, "missed heartbeat")
+	s, _ := getSite(e, 1)
+	if s.Status != "LATE" {
+		t.Errorf("expected LATE during retry, got %s", s.Status)
+	}
+
+	e.handleStatusChange(s, "DOWN", 0, 0, "missed heartbeat")
+	s, _ = getSite(e, 1)
+	if s.Status != "DOWN" {
+		t.Errorf("expected DOWN after retries exhausted, got %s", s.Status)
+	}
+}

internal/monitor/checker_test.go

@@ -130,8 +130,14 @@ func TestRunCheck_Port_Open(t *testing.T) {
 	}
 	defer ln.Close()
 
-	_, portStr, _ := net.SplitHostPort(ln.Addr().String())
-	port, _ := strconv.Atoi(portStr)
+	_, portStr, err := net.SplitHostPort(ln.Addr().String())
+	if err != nil {
+		t.Fatalf("SplitHostPort: %v", err)
+	}
+	port, err := strconv.Atoi(portStr)
+	if err != nil {
+		t.Fatalf("Atoi: %v", err)
+	}
 
 	site := models.SiteConfig{ID: 1, Type: "port", Hostname: "127.0.0.1", Port: port, Timeout: 2}
 	result := RunCheck(context.Background(), site, nil, nil, false, true)
@@ -149,8 +155,14 @@ func TestRunCheck_Port_Closed(t *testing.T) {
 	if err != nil {
 		t.Fatal(err)
 	}
-	_, portStr, _ := net.SplitHostPort(ln.Addr().String())
-	port, _ := strconv.Atoi(portStr)
+	_, portStr, err := net.SplitHostPort(ln.Addr().String())
+	if err != nil {
+		t.Fatalf("SplitHostPort: %v", err)
+	}
+	port, err := strconv.Atoi(portStr)
+	if err != nil {
+		t.Fatalf("Atoi: %v", err)
+	}
 	ln.Close()
 
 	site := models.SiteConfig{ID: 1, Type: "port", Hostname: "127.0.0.1", Port: port, Timeout: 1}
@@ -168,8 +180,14 @@ func TestRunPortCheck_UsesPinnedIP(t *testing.T) {
 	}
 	defer ln.Close()
 
-	_, portStr, _ := net.SplitHostPort(ln.Addr().String())
-	port, _ := strconv.Atoi(portStr)
+	_, portStr, err := net.SplitHostPort(ln.Addr().String())
+	if err != nil {
+		t.Fatalf("SplitHostPort: %v", err)
+	}
+	port, err := strconv.Atoi(portStr)
+	if err != nil {
+		t.Fatalf("Atoi: %v", err)
+	}
 
 	// Pass a pinned IP — runPortCheck should dial it instead of resolving Hostname.
 	site := models.SiteConfig{ID: 1, Type: "port", Hostname: "will-not-resolve.invalid", Port: port, Timeout: 2}
@@ -187,8 +205,14 @@ func TestRunPortCheck_NilPinnedIP_UsesHostname(t *testing.T) {
 	}
 	defer ln.Close()
 
-	_, portStr, _ := net.SplitHostPort(ln.Addr().String())
-	port, _ := strconv.Atoi(portStr)
+	_, portStr, err := net.SplitHostPort(ln.Addr().String())
+	if err != nil {
+		t.Fatalf("SplitHostPort: %v", err)
+	}
+	port, err := strconv.Atoi(portStr)
+	if err != nil {
+		t.Fatalf("Atoi: %v", err)
+	}
 
 	site := models.SiteConfig{ID: 1, Type: "port", Hostname: "127.0.0.1", Port: port, Timeout: 2}
 	result := runPortCheck(context.Background(), site, nil)
@@ -205,8 +229,14 @@ func TestRunCheck_Port_BlocksPrivateByDefault(t *testing.T) {
 	}
 	defer ln.Close()
 
-	_, portStr, _ := net.SplitHostPort(ln.Addr().String())
-	port, _ := strconv.Atoi(portStr)
+	_, portStr, err := net.SplitHostPort(ln.Addr().String())
+	if err != nil {
+		t.Fatalf("SplitHostPort: %v", err)
+	}
+	port, err := strconv.Atoi(portStr)
+	if err != nil {
+		t.Fatalf("Atoi: %v", err)
+	}
 
 	site := models.SiteConfig{ID: 1, Type: "port", Hostname: "127.0.0.1", Port: port, Timeout: 2}
 	result := RunCheck(context.Background(), site, nil, nil, false, false)

internal/monitor/checks.go

@@ -0,0 +1,329 @@
+package monitor
+
+import (
+	"context"
+	"fmt"
+	"math/rand/v2"
+	"time"
+
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+)
+
+func (e *Engine) RecordHeartbeat(token string) bool {
+	if !e.IsActive() {
+		return false
+	}
+
+	e.mu.RLock()
+	targetID, ok := e.tokenIndex[token]
+	e.mu.RUnlock()
+	if !ok {
+		return false
+	}
+
+	var (
+		prevStatus models.Status
+		name       string
+		alertID    int
+		downSince  time.Time
+	)
+	_, exists := e.applyState(targetID, func(s *models.Site) {
+		prevStatus = s.Status
+		name = s.Name
+		alertID = s.AlertID
+		downSince = s.StatusChangedAt // captured before mutation = when it went down
+
+		s.LastCheck = time.Now()
+		s.Status = models.StatusUp
+		s.FailureCount = 0
+		s.Latency = 0
+		s.LastError = ""
+		s.LastSuccessAt = time.Now()
+		if prevStatus != models.StatusUp {
+			s.StatusChangedAt = time.Now()
+		}
+	})
+	if !exists {
+		return false
+	}
+
+	switch prevStatus {
+	case models.StatusPending:
+		e.AddLog(fmt.Sprintf("Push Monitor '%s' received first heartbeat", name))
+	case models.StatusLate:
+		e.AddLog(fmt.Sprintf("Push Monitor '%s' heartbeat arrived (was late)", name))
+	case models.StatusStale:
+		e.AddLog(fmt.Sprintf("Push Monitor '%s' heartbeat arrived (was stale)", name))
+	case models.StatusDown:
+		downDur := ""
+		if !downSince.IsZero() {
+			downDur = fmt.Sprintf(" (was down %s)", fmtDurationShort(time.Since(downSince)))
+		}
+		e.AddLog(fmt.Sprintf("Push Monitor '%s' recovered%s", name, downDur))
+		go e.triggerAlert(alertID, "✅ RECOVERY", fmt.Sprintf("Push Monitor '%s' is receiving heartbeats.%s", name, downDur))
+	}
+
+	e.recordCheck(targetID, 0, true)
+
+	if prevStatus != models.StatusUp && prevStatus != models.StatusPending {
+		e.enqueueWrite(writeStateChange{siteID: targetID, fromStatus: string(prevStatus), toStatus: string(models.StatusUp)})
+	}
+
+	return true
+}
+
+func (e *Engine) getRecheckChan(id int) chan struct{} {
+	e.recheckMu.Lock()
+	defer e.recheckMu.Unlock()
+	ch, ok := e.recheck[id]
+	if !ok {
+		ch = make(chan struct{}, 1)
+		e.recheck[id] = ch
+	}
+	return ch
+}
+
+func (e *Engine) signalRecheck(id int) {
+	ch := e.getRecheckChan(id)
+	select {
+	case ch <- struct{}{}:
+	default:
+	}
+}
+
+func (e *Engine) monitorRoutine(ctx context.Context, id int) {
+	recheckCh := e.getRecheckChan(id)
+
+	// Stagger initial check to avoid thundering herd on startup
+	stagger := time.Duration(rand.IntN(3000)) * time.Millisecond //nolint:gosec // non-security jitter
+	select {
+	case <-time.After(stagger):
+	case <-ctx.Done():
+		return
+	}
+
+	e.checkByID(ctx, id)
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		default:
+		}
+
+		if !e.IsActive() {
+			select {
+			case <-time.After(pollInterval):
+			case <-ctx.Done():
+				return
+			case <-recheckCh:
+			}
+			continue
+		}
+
+		e.mu.RLock()
+		site, exists := e.liveState[id]
+		e.mu.RUnlock()
+		if !exists {
+			return
+		}
+
+		if site.Paused {
+			select {
+			case <-time.After(pollInterval):
+			case <-ctx.Done():
+				return
+			case <-recheckCh:
+			}
+			continue
+		}
+
+		interval := site.Interval
+		if interval < minCheckInterval {
+			interval = minCheckInterval
+		}
+		jitter := time.Duration(rand.IntN(interval*100)) * time.Millisecond //nolint:gosec // non-security jitter
+		select {
+		case <-time.After(time.Duration(interval)*time.Second + jitter):
+		case <-ctx.Done():
+			return
+		case <-recheckCh:
+		}
+		e.checkByID(ctx, id)
+	}
+}
+
+func (e *Engine) checkByID(ctx context.Context, id int) {
+	if !e.IsActive() {
+		return
+	}
+
+	e.mu.RLock()
+	site, exists := e.liveState[id]
+	e.mu.RUnlock()
+	if !exists || site.Paused {
+		return
+	}
+
+	switch site.Type {
+	case "push":
+		e.checkPush(ctx, site)
+	case "group":
+		e.checkGroup(ctx, site)
+	default:
+		result := RunCheck(ctx, site.SiteConfig, e.strictClient, e.insecureClient, e.insecureSkipVerify, e.allowPrivateTargets)
+		updatedSite := site
+		updatedSite.HasSSL = result.HasSSL
+		updatedSite.CertExpiry = result.CertExpiry
+		updatedSite.Latency = time.Duration(result.LatencyNs)
+		updatedSite.LastCheck = time.Now()
+		e.handleStatusChange(updatedSite, result.Status, result.StatusCode, time.Duration(result.LatencyNs), result.ErrorReason)
+	}
+}
+
+func (e *Engine) checkPush(_ context.Context, site models.Site) {
+	if site.Status == models.StatusPending {
+		return
+	}
+
+	interval := time.Duration(site.Interval) * time.Second
+	grace := interval / 2
+	if grace < minPushGrace {
+		grace = minPushGrace
+	}
+
+	overdue := site.LastCheck.Add(interval)
+	staleMark := overdue.Add(grace / 2)
+	graceEnd := overdue.Add(grace)
+	now := time.Now()
+
+	if now.After(graceEnd) {
+		if site.Status != models.StatusDown {
+			e.handleStatusChange(site, string(models.StatusDown), 0, 0, "heartbeat missed")
+		}
+	} else if now.After(staleMark) {
+		if site.Status != models.StatusStale {
+			e.handleStatusChange(site, string(models.StatusStale), 0, 0, "heartbeat stale")
+		}
+	} else if now.After(overdue) {
+		if site.Status != models.StatusLate {
+			e.handleStatusChange(site, string(models.StatusLate), 0, 0, "heartbeat overdue")
+		}
+	}
+}
+
+func (e *Engine) checkGroup(_ context.Context, site models.Site) {
+	e.mu.RLock()
+	status := models.StatusUp
+	hasChildren := false
+	for _, child := range e.liveState {
+		if child.ParentID != site.ID || child.Type == "group" {
+			continue
+		}
+		hasChildren = true
+		if child.Paused || e.isInMaintenance(child.ID) {
+			continue
+		}
+		if child.Status == models.StatusDown || child.Status == models.StatusSSLExp {
+			status = models.StatusDown
+		} else if child.Status == models.StatusStale && status != models.StatusDown {
+			status = models.StatusStale
+		} else if child.Status == models.StatusLate && status != models.StatusDown && status != models.StatusStale {
+			status = models.StatusLate
+		} else if child.Status == models.StatusPending && status != models.StatusDown && status != models.StatusStale && status != models.StatusLate {
+			status = models.StatusPending
+		}
+	}
+	e.mu.RUnlock()
+
+	if !hasChildren {
+		status = models.StatusPending
+	}
+
+	var prev models.Status
+	e.applyState(site.ID, func(s *models.Site) {
+		prev = s.Status
+		s.Status = status
+		if status != prev && prev != models.StatusPending {
+			s.StatusChangedAt = time.Now()
+		}
+	})
+	e.recordCheck(site.ID, 0, !status.IsBroken())
+
+	if status != prev && prev != models.StatusPending {
+		e.enqueueWrite(writeStateChange{siteID: site.ID, fromStatus: string(prev), toStatus: string(status)})
+		if status.IsBroken() {
+			e.AddLog(fmt.Sprintf("Group '%s' is %s", site.Name, status))
+		} else if prev.IsBroken() {
+			e.AddLog(fmt.Sprintf("Group '%s' recovered", site.Name))
+		}
+	}
+}
+
+func (e *Engine) EnqueueProbeCheck(siteID int, nodeID string, latencyNs int64, isUp bool) {
+	e.enqueueWrite(writeProbeCheck{siteID: siteID, nodeID: nodeID, latencyNs: latencyNs, isUp: isUp})
+}
+
+// SetAggStrategy must be called before Start: the field is read by the probe
+// aggregation path without synchronization.
+func (e *Engine) SetAggStrategy(strategy AggregationStrategy) {
+	e.aggStrategy = strategy
+}
+
+func (e *Engine) IngestProbeResult(nodeID string, siteID int, latencyNs int64, isUp bool, errorReason string) {
+	e.mu.RLock()
+	site, exists := e.liveState[siteID]
+	e.mu.RUnlock()
+	if !exists {
+		return
+	}
+
+	staleAfter := time.Duration(site.Interval) * time.Second * 3
+	if staleAfter < time.Minute {
+		staleAfter = time.Minute
+	}
+
+	now := time.Now()
+	e.probeResultsMu.Lock()
+	if e.probeResults[siteID] == nil {
+		e.probeResults[siteID] = make(map[string]NodeResult)
+	}
+	e.probeResults[siteID][nodeID] = NodeResult{
+		NodeID:      nodeID,
+		IsUp:        isUp,
+		LatencyNs:   latencyNs,
+		CheckedAt:   now,
+		ErrorReason: errorReason,
+	}
+	results := make([]NodeResult, 0, len(e.probeResults[siteID]))
+	for id, r := range e.probeResults[siteID] {
+		if now.Sub(r.CheckedAt) > staleAfter {
+			delete(e.probeResults[siteID], id)
+			continue
+		}
+		results = append(results, r)
+	}
+	e.probeResultsMu.Unlock()
+
+	aggUp, avgLatency := AggregateStatus(results, e.aggStrategy)
+
+	probeStatus := models.StatusUp
+	if !aggUp {
+		probeStatus = models.StatusDown
+	}
+
+	updatedSite := site
+	updatedSite.Latency = time.Duration(avgLatency)
+	updatedSite.LastCheck = time.Now()
+	e.handleStatusChange(updatedSite, string(probeStatus), 0, time.Duration(avgLatency), errorReason)
+}
+
+func (e *Engine) GetProbeResults(siteID int) map[string]NodeResult {
+	e.probeResultsMu.RLock()
+	defer e.probeResultsMu.RUnlock()
+	src := e.probeResults[siteID]
+	cp := make(map[string]NodeResult, len(src))
+	for k, v := range src {
+		cp[k] = v
+	}
+	return cp
+}

internal/monitor/checks_test.go

@@ -0,0 +1,389 @@
+package monitor
+
+import (
+	"context"
+	"testing"
+	"time"
+
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+)
+
+// --- Group 2: Heartbeat ---
+
+func TestRecordHeartbeat_ValidToken(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "push-test", Type: "push", Token: "abc123"},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	injectSite(e, site)
+
+	if !e.RecordHeartbeat("abc123") {
+		t.Error("expected true for valid token")
+	}
+
+	s, _ := getSite(e, 1)
+	if s.Status != "UP" {
+		t.Errorf("expected UP, got %s", s.Status)
+	}
+	if time.Since(s.LastCheck) > time.Second {
+		t.Error("expected LastCheck to be recent")
+	}
+}
+
+func TestRecordHeartbeat_RecoveryFromDown(t *testing.T) {
+	ms := newMockStore()
+	ms.alerts[1] = models.AlertConfig{ID: 1, Name: "test", Type: "webhook", Settings: map[string]string{"url": "http://example.com"}}
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "push-test", Type: "push", Token: "abc123", AlertID: 1},
+		SiteState:  models.SiteState{Status: "DOWN", FailureCount: 3},
+	}
+	injectSite(e, site)
+
+	if !e.RecordHeartbeat("abc123") {
+		t.Error("expected true")
+	}
+
+	s, _ := getSite(e, 1)
+	if s.Status != "UP" {
+		t.Errorf("expected UP, got %s", s.Status)
+	}
+	if s.FailureCount != 0 {
+		t.Errorf("expected FailureCount 0, got %d", s.FailureCount)
+	}
+	waitAsync()
+	if len(ms.getAlertCallsSnapshot()) == 0 {
+		t.Error("expected recovery alert")
+	}
+}
+
+func TestRecordHeartbeat_UnknownToken(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+
+	if e.RecordHeartbeat("unknown") {
+		t.Error("expected false for unknown token")
+	}
+}
+
+func TestRecordHeartbeat_InactiveEngine(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Type: "push", Token: "abc123"},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	injectSite(e, site)
+	e.SetActive(false)
+
+	if e.RecordHeartbeat("abc123") {
+		t.Error("expected false when inactive")
+	}
+}
+
+// --- Group 3: Push Deadline ---
+
+func TestCheckPush_DeadlineMissed(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "push", Type: "push", Interval: 10, MaxRetries: 0},
+		SiteState:  models.SiteState{Status: "UP", LastCheck: time.Now().Add(-120 * time.Second)},
+	}
+	injectSite(e, site)
+
+	e.checkPush(context.Background(), site)
+
+	s, _ := getSite(e, 1)
+	if s.Status != "DOWN" {
+		t.Errorf("expected DOWN after missed deadline, got %s", s.Status)
+	}
+}
+
+func TestCheckPush_OverdueBecomesLate(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "push", Type: "push", Interval: 300},
+		SiteState:  models.SiteState{Status: "UP", LastCheck: time.Now().Add(-310 * time.Second)},
+	}
+	injectSite(e, site)
+
+	e.checkPush(context.Background(), site)
+
+	s, _ := getSite(e, 1)
+	if s.Status != "LATE" {
+		t.Errorf("expected LATE when overdue but within grace, got %s", s.Status)
+	}
+}
+
+func TestCheckPush_OverdueBecomesStale(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	// interval=300, grace=150 (300/2), staleMark=overdue+75
+	// at 380s: past staleMark(375) but before graceEnd(450)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "push", Type: "push", Interval: 300},
+		SiteState:  models.SiteState{Status: "UP", LastCheck: time.Now().Add(-380 * time.Second)},
+	}
+	injectSite(e, site)
+
+	e.checkPush(context.Background(), site)
+
+	s, _ := getSite(e, 1)
+	if s.Status != "STALE" {
+		t.Errorf("expected STALE when past midpoint of grace, got %s", s.Status)
+	}
+}
+
+func TestCheckPush_WithinDeadline(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "push", Type: "push", Interval: 60},
+		SiteState:  models.SiteState{Status: "UP", LastCheck: time.Now()},
+	}
+	injectSite(e, site)
+
+	e.checkPush(context.Background(), site)
+
+	s, _ := getSite(e, 1)
+	if s.Status != "UP" {
+		t.Errorf("expected UP, got %s", s.Status)
+	}
+}
+
+func TestCheckPush_PendingStaysPending(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "push", Type: "push", Interval: 60},
+		SiteState:  models.SiteState{Status: "PENDING"},
+	}
+	injectSite(e, site)
+
+	e.checkPush(context.Background(), site)
+
+	s, _ := getSite(e, 1)
+	if s.Status != "PENDING" {
+		t.Errorf("expected PENDING to stay until first heartbeat, got %s", s.Status)
+	}
+}
+
+// --- Group 4: Group Checks ---
+
+func TestCheckGroup_AllChildrenUp(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	group := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "group", Type: "group"},
+		SiteState:  models.SiteState{Status: "PENDING"},
+	}
+	child1 := models.Site{
+		SiteConfig: models.SiteConfig{ID: 2, Name: "child1", Type: "http", ParentID: 1},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	child2 := models.Site{
+		SiteConfig: models.SiteConfig{ID: 3, Name: "child2", Type: "http", ParentID: 1},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	injectSite(e, group)
+	injectSite(e, child1)
+	injectSite(e, child2)
+
+	e.checkGroup(context.Background(), group)
+
+	s, _ := getSite(e, 1)
+	if s.Status != "UP" {
+		t.Errorf("expected group UP, got %s", s.Status)
+	}
+}
+
+func TestCheckGroup_OneChildDown(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	group := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "group", Type: "group"},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	child1 := models.Site{
+		SiteConfig: models.SiteConfig{ID: 2, Name: "child1", Type: "http", ParentID: 1},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	child2 := models.Site{
+		SiteConfig: models.SiteConfig{ID: 3, Name: "child2", Type: "http", ParentID: 1},
+		SiteState:  models.SiteState{Status: "DOWN"},
+	}
+	injectSite(e, group)
+	injectSite(e, child1)
+	injectSite(e, child2)
+
+	e.checkGroup(context.Background(), group)
+
+	s, _ := getSite(e, 1)
+	if s.Status != "DOWN" {
+		t.Errorf("expected group DOWN, got %s", s.Status)
+	}
+}
+
+func TestCheckGroup_PausedChildIgnored(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	group := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "group", Type: "group"},
+	}
+	child1 := models.Site{
+		SiteConfig: models.SiteConfig{ID: 2, Name: "child1", Type: "http", ParentID: 1},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	child2 := models.Site{
+		SiteConfig: models.SiteConfig{ID: 3, Name: "child2", Type: "http", ParentID: 1, Paused: true},
+		SiteState:  models.SiteState{Status: "DOWN"},
+	}
+	injectSite(e, group)
+	injectSite(e, child1)
+	injectSite(e, child2)
+
+	e.checkGroup(context.Background(), group)
+
+	s, _ := getSite(e, 1)
+	if s.Status != "UP" {
+		t.Errorf("expected UP (paused child ignored), got %s", s.Status)
+	}
+}
+
+func TestCheckGroup_MaintenanceChildIgnored(t *testing.T) {
+	ms := newMockStore()
+	ms.maintenance[3] = true
+	e := newTestEngine(ms)
+	group := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "group", Type: "group"},
+	}
+	child1 := models.Site{
+		SiteConfig: models.SiteConfig{ID: 2, Name: "child1", Type: "http", ParentID: 1},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	child2 := models.Site{
+		SiteConfig: models.SiteConfig{ID: 3, Name: "child2", Type: "http", ParentID: 1},
+		SiteState:  models.SiteState{Status: "DOWN"},
+	}
+	injectSite(e, group)
+	injectSite(e, child1)
+	injectSite(e, child2)
+	e.refreshMaintenanceCache(context.Background())
+
+	e.checkGroup(context.Background(), group)
+
+	s, _ := getSite(e, 1)
+	if s.Status != "UP" {
+		t.Errorf("expected UP (maint child ignored), got %s", s.Status)
+	}
+}
+
+func TestCheckGroup_NoChildren(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	group := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "group", Type: "group"},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	injectSite(e, group)
+
+	e.checkGroup(context.Background(), group)
+
+	s, _ := getSite(e, 1)
+	if s.Status != "PENDING" {
+		t.Errorf("expected PENDING for no children, got %s", s.Status)
+	}
+}
+
+// Groups must not auto-pause when all children are paused — that creates a
+// one-way trap because monitorRoutine skips paused sites.
+func TestCheckGroup_AllPausedNoAutoFreeze(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	group := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "group", Type: "group"},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	child1 := models.Site{
+		SiteConfig: models.SiteConfig{ID: 2, Name: "child1", Type: "http", ParentID: 1, Paused: true},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	child2 := models.Site{
+		SiteConfig: models.SiteConfig{ID: 3, Name: "child2", Type: "http", ParentID: 1, Paused: true},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	injectSite(e, group)
+	injectSite(e, child1)
+	injectSite(e, child2)
+
+	e.checkGroup(context.Background(), group)
+
+	s, _ := getSite(e, 1)
+	if s.Paused {
+		t.Error("group must not auto-pause when all children are paused")
+	}
+}
+
+// Dead probe results must be expired so they don't poison aggregation.
+func TestIngestProbeResult_ExpiresStaleProbes(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", Type: "http", Interval: 30},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	injectSite(e, site)
+
+	e.probeResultsMu.Lock()
+	e.probeResults[1] = map[string]NodeResult{
+		"dead-probe": {
+			NodeID:    "dead-probe",
+			IsUp:      false,
+			CheckedAt: time.Now().Add(-10 * time.Minute),
+		},
+	}
+	e.probeResultsMu.Unlock()
+
+	e.IngestProbeResult("live-probe", 1, 5000, true, "")
+
+	e.probeResultsMu.RLock()
+	_, deadExists := e.probeResults[1]["dead-probe"]
+	_, liveExists := e.probeResults[1]["live-probe"]
+	e.probeResultsMu.RUnlock()
+
+	if deadExists {
+		t.Error("stale probe result should have been expired")
+	}
+	if !liveExists {
+		t.Error("live probe result should still exist")
+	}
+}
+
+// RemoveSite must clean up probeResults.
+func TestRemoveSite_CleansProbeResults(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", Type: "http"},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	injectSite(e, site)
+
+	e.probeResultsMu.Lock()
+	e.probeResults[1] = map[string]NodeResult{
+		"node-a": {NodeID: "node-a", IsUp: true, CheckedAt: time.Now()},
+	}
+	e.probeResultsMu.Unlock()
+
+	e.RemoveSite(1)
+
+	e.probeResultsMu.RLock()
+	defer e.probeResultsMu.RUnlock()
+	if _, exists := e.probeResults[1]; exists {
+		t.Error("probe results should be cleaned up after RemoveSite")
+	}
+}

internal/monitor/history.go

@@ -14,8 +14,8 @@ type SiteHistory struct {
 	UpChecks    int
 }
 
-func (e *Engine) InitHistory() {
-	all, err := e.db.LoadAllHistory(context.Background(), maxHistoryLen)
+func (e *Engine) InitHistory(ctx context.Context) {
+	all, err := e.db.LoadAllHistory(ctx, maxHistoryLen)
 	if err != nil {
 		e.AddLog("Failed to load check history: " + err.Error())
 		return

internal/monitor/maintenance.go

@@ -0,0 +1,70 @@
+package monitor
+
+import (
+	"context"
+	"fmt"
+	"time"
+)
+
+func (e *Engine) maintenancePruner(ctx context.Context) {
+	ticker := time.NewTicker(maintPruneInterval)
+	defer ticker.Stop()
+
+	e.pruneMaintenanceWindows(ctx)
+
+	for {
+		select {
+		case <-ticker.C:
+			e.pruneMaintenanceWindows(ctx)
+		case <-ctx.Done():
+			return
+		}
+	}
+}
+
+func (e *Engine) pruneMaintenanceWindows(ctx context.Context) {
+	pruned, err := e.db.PruneExpiredMaintenanceWindows(ctx, e.maintRetention)
+	if err != nil {
+		e.AddLog(fmt.Sprintf("Maintenance prune error: %v", err))
+		return
+	}
+	if pruned > 0 {
+		e.AddLog(fmt.Sprintf("Pruned %d expired maintenance window(s)", pruned))
+	}
+}
+
+func (e *Engine) isInMaintenance(monitorID int) bool {
+	e.maintCacheMu.RLock()
+	defer e.maintCacheMu.RUnlock()
+	return e.maintCache[monitorID]
+}
+
+func (e *Engine) refreshMaintenanceCache(ctx context.Context) {
+	windows, err := e.db.GetActiveMaintenanceWindows(ctx)
+	if err != nil {
+		return
+	}
+
+	directMaint := make(map[int]bool)
+	var globalMaint bool
+	for _, w := range windows {
+		if w.MonitorID == 0 {
+			globalMaint = true
+		} else {
+			directMaint[w.MonitorID] = true
+		}
+	}
+
+	resolved := make(map[int]bool)
+	e.mu.RLock()
+	for id, site := range e.liveState {
+		if globalMaint || directMaint[id] || (site.ParentID > 0 && directMaint[site.ParentID]) {
+			resolved[id] = true
+		}
+	}
+	e.mu.RUnlock()
+
+	e.maintCacheMu.Lock()
+	e.maintCache = resolved
+	e.maintCacheMu.Unlock()
+}

internal/monitor/maintenance_test.go

@@ -0,0 +1,53 @@
+package monitor
+
+import (
+	"context"
+	"testing"
+
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+)
+
+// Maintenance cache resolves parent relationships correctly.
+func TestIsInMaintenance_UsesCache(t *testing.T) {
+	ms := newMockStore()
+	ms.maintenance[10] = true // direct maintenance on group
+	e := newTestEngine(ms)
+	group := models.Site{
+		SiteConfig: models.SiteConfig{ID: 10, Name: "group", Type: "group"},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	child := models.Site{
+		SiteConfig: models.SiteConfig{ID: 20, Name: "child", Type: "http", ParentID: 10},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	injectSite(e, group)
+	injectSite(e, child)
+	e.refreshMaintenanceCache(context.Background())
+
+	if !e.isInMaintenance(10) {
+		t.Error("group should be in maintenance (direct)")
+	}
+	if !e.isInMaintenance(20) {
+		t.Error("child should be in maintenance (parent)")
+	}
+	if e.isInMaintenance(99) {
+		t.Error("unknown monitor should not be in maintenance")
+	}
+}
+
+// Global maintenance (monitor_id=0) applies to all monitors.
+func TestIsInMaintenance_GlobalMaintenance(t *testing.T) {
+	ms := newMockStore()
+	ms.maintenance[0] = true
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", Type: "http"},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	injectSite(e, site)
+	e.refreshMaintenanceCache(context.Background())
+
+	if !e.isInMaintenance(1) {
+		t.Error("all monitors should be in maintenance during global window")
+	}
+}

internal/monitor/monitor.go

@@ -4,14 +4,12 @@ import (
 	"context"
 	"crypto/tls"
 	"fmt"
-	"math/rand/v2"
 	"net/http"
 	"regexp"
 	"strings"
 	"sync"
 	"time"
 
-	"gitea.lerkolabs.com/lerkolabs/uptop/internal/alert"
 	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
 	"gitea.lerkolabs.com/lerkolabs/uptop/internal/store"
 )
@@ -24,23 +22,16 @@ const (
 	maintPruneInterval    = 15 * time.Minute
 	defaultMaintRetention = 7 * 24 * time.Hour
 	dbWriteBuffer         = 4096
+	alertSendTimeout      = 30 * time.Second
 	dbPruneInterval       = 10 * time.Minute
 )
 
-type AlertHealth struct {
-	LastSendAt time.Time
-	LastSendOK bool
-	LastError  string
-	SendCount  int
-	FailCount  int
-}
-
 type Engine struct {
 	mu        sync.RWMutex
 	liveState map[int]models.Site
 
 	logMu    sync.RWMutex
-	logStore []string
+	logStore []models.LogEntry
 
 	activeMu sync.RWMutex
 	isActive bool
@@ -73,6 +64,7 @@ type Engine struct {
 	dbWrites  chan dbWrite
 	writerWG  sync.WaitGroup
 	checkerWG sync.WaitGroup
+	ctx       context.Context
 	cancel    context.CancelFunc
 	stopOnce  sync.Once
 }
@@ -99,6 +91,7 @@ func newEngine(s store.Store, allowPrivateTargets bool) *Engine {
 		allowPrivateTargets: allowPrivateTargets,
 		maintRetention:      defaultMaintRetention,
 		dbWrites:            make(chan dbWrite, dbWriteBuffer),
+		ctx:                 context.Background(),
 		db:                  s,
 		strictClient: &http.Client{
 			Transport: &http.Transport{
@@ -152,11 +145,13 @@ func fmtDurationShort(d time.Duration) string {
 // appendLog adds a timestamped entry to the in-memory ring buffer and returns
 // it. It never touches the database, so it is safe to call from the db-write
 // drop/error path without recursing back through the write queue.
-func (e *Engine) appendLog(msg string) string {
-	ts := time.Now().Format("15:04:05")
-	entry := fmt.Sprintf("[%s] %s", ts, sanitizeLog(msg))
+func (e *Engine) appendLog(msg string) models.LogEntry {
+	entry := models.LogEntry{
+		Message:   sanitizeLog(msg),
+		CreatedAt: time.Now(),
+	}
 	e.logMu.Lock()
-	e.logStore = append([]string{entry}, e.logStore...)
+	e.logStore = append([]models.LogEntry{entry}, e.logStore...)
 	if len(e.logStore) > maxLogEntries {
 		e.logStore = e.logStore[:maxLogEntries]
 	}
@@ -166,7 +161,7 @@ func (e *Engine) appendLog(msg string) string {
 
 func (e *Engine) AddLog(msg string) {
 	entry := e.appendLog(msg)
-	e.enqueueWrite(writeLog{message: entry})
+	e.enqueueWrite(writeLog{message: entry.Message})
 }
 
 // enqueueWrite hands a persistence task to the writer goroutine without
@@ -207,6 +202,8 @@ func (e *Engine) dbWriter(ctx context.Context) {
 }
 
 // drainWrites flushes everything still buffered, best-effort, at shutdown.
+// Uses context.Background because the engine ctx is already cancelled when
+// this runs — writes still need to reach the DB.
 func (e *Engine) drainWrites() {
 	for {
 		select {
@@ -245,169 +242,33 @@ func (e *Engine) Stop() {
 	})
 }
 
-func (e *Engine) InitLogs() {
-	logs, err := e.db.LoadLogs(context.Background(), maxLogEntries)
+func (e *Engine) InitLogs(ctx context.Context) {
+	entries, err := e.db.LoadLogs(ctx, maxLogEntries)
 	if err != nil {
 		return
 	}
-	if len(logs) == 0 {
+	if len(entries) == 0 {
 		return
 	}
 	e.logMu.Lock()
 	defer e.logMu.Unlock()
-	e.logStore = logs
+	e.logStore = entries
 }
 
-// InitAlertHealth restores persisted alert send health so the dashboard shows real
-// "last sent" / health state on startup instead of resetting every channel to "never".
-func (e *Engine) InitAlertHealth() {
-	records, err := e.db.LoadAlertHealth(context.Background())
-	if err != nil {
-		return
-	}
-	e.alertHealthMu.Lock()
-	defer e.alertHealthMu.Unlock()
-	for id, r := range records {
-		e.alertHealth[id] = AlertHealth{
-			LastSendAt: r.LastSendAt,
-			LastSendOK: r.LastSendOK,
-			LastError:  r.LastError,
-			SendCount:  r.SendCount,
-			FailCount:  r.FailCount,
-		}
-	}
-}
-
-func (e *Engine) GetLogs() []string {
+func (e *Engine) GetLogs() []models.LogEntry {
 	e.logMu.RLock()
 	defer e.logMu.RUnlock()
-	logs := make([]string, len(e.logStore))
+	logs := make([]models.LogEntry, len(e.logStore))
 	copy(logs, e.logStore)
 	return logs
 }
 
-func (e *Engine) SetActive(active bool) {
-	e.activeMu.Lock()
-	defer e.activeMu.Unlock()
-	if e.isActive != active {
-		e.isActive = active
-		status := "RESUMED (Active)"
-		if !active {
-			status = "PAUSED (Passive)"
-		}
-		e.AddLog(fmt.Sprintf("Engine %s", status))
-	}
-}
-
-func (e *Engine) IsActive() bool {
-	e.activeMu.RLock()
-	defer e.activeMu.RUnlock()
-	return e.isActive
-}
-
-func (e *Engine) GetAllSites() []models.Site {
-	e.mu.RLock()
-	defer e.mu.RUnlock()
-	sites := make([]models.Site, 0, len(e.liveState))
-	for _, s := range e.liveState {
-		sites = append(sites, s)
-	}
-	return sites
-}
-
-func (e *Engine) GetLiveState() map[int]models.Site {
-	e.mu.RLock()
-	defer e.mu.RUnlock()
-	cp := make(map[int]models.Site, len(e.liveState))
-	for k, v := range e.liveState {
-		cp[k] = v
-	}
-	return cp
-}
-
-func (e *Engine) RecordHeartbeat(token string) bool {
-	if !e.IsActive() {
-		return false
-	}
-
-	e.mu.RLock()
-	targetID, ok := e.tokenIndex[token]
-	e.mu.RUnlock()
-	if !ok {
-		return false
-	}
-
-	var (
-		prevStatus models.Status
-		name       string
-		alertID    int
-		downSince  time.Time
-	)
-	_, exists := e.applyState(targetID, func(s *models.Site) {
-		prevStatus = s.Status
-		name = s.Name
-		alertID = s.AlertID
-		downSince = s.StatusChangedAt // captured before mutation = when it went down
-
-		s.LastCheck = time.Now()
-		s.Status = models.StatusUp
-		s.FailureCount = 0
-		s.Latency = 0
-		s.LastError = ""
-		s.LastSuccessAt = time.Now()
-		if prevStatus != models.StatusUp {
-			s.StatusChangedAt = time.Now()
-		}
-	})
-	if !exists {
-		return false
-	}
-
-	switch prevStatus {
-	case models.StatusPending:
-		e.AddLog(fmt.Sprintf("Push Monitor '%s' received first heartbeat", name))
-	case models.StatusLate:
-		e.AddLog(fmt.Sprintf("Push Monitor '%s' heartbeat arrived (was late)", name))
-	case models.StatusStale:
-		e.AddLog(fmt.Sprintf("Push Monitor '%s' heartbeat arrived (was stale)", name))
-	case models.StatusDown:
-		downDur := ""
-		if !downSince.IsZero() {
-			downDur = fmt.Sprintf(" (was down %s)", fmtDurationShort(time.Since(downSince)))
-		}
-		e.AddLog(fmt.Sprintf("Push Monitor '%s' recovered%s", name, downDur))
-		go e.triggerAlert(alertID, "✅ RECOVERY", fmt.Sprintf("Push Monitor '%s' is receiving heartbeats.%s", name, downDur))
-	}
-
-	e.recordCheck(targetID, 0, true)
-
-	if prevStatus != models.StatusUp && prevStatus != models.StatusPending {
-		e.enqueueWrite(writeStateChange{siteID: targetID, fromStatus: string(prevStatus), toStatus: string(models.StatusUp)})
-	}
-
-	return true
-}
-
-func (e *Engine) addToTokenIndex(site models.Site) {
-	if site.Type == "push" && site.Token != "" {
-		e.tokenIndex[site.Token] = site.ID
-	}
-}
-
-func (e *Engine) removeFromTokenIndex(id int) {
-	for token, sid := range e.tokenIndex {
-		if sid == id {
-			delete(e.tokenIndex, token)
-			return
-		}
-	}
-}
-
 func (e *Engine) Start(ctx context.Context) {
 	// e.cancel is invoked by Stop() to drain and halt the writer; gosec can't
 	// trace the cross-method call, and cancelling the parent reaps this child
 	// regardless, so the leak it warns about can't occur.
 	ctx, e.cancel = context.WithCancel(ctx) //nolint:gosec // cancel is called in Stop()
+	e.ctx = ctx
 
 	e.writerWG.Add(1)
 	go e.dbWriter(ctx)
@@ -422,8 +283,6 @@ func (e *Engine) Start(ctx context.Context) {
 			default:
 			}
 
-			e.refreshMaintenanceCache(ctx)
-
 			configs, err := e.db.GetSites(ctx)
 			if err != nil {
 				e.AddLog(fmt.Sprintf("Failed to load sites: %v", err))
@@ -466,6 +325,10 @@ func (e *Engine) Start(ctx context.Context) {
 				}
 			}
 
+			// Refresh after sites load so the cache covers newly added sites.
+			// On first iteration liveState was empty before the loop above.
+			e.refreshMaintenanceCache(ctx)
+
 			e.mu.RLock()
 			var vanished []int
 			for id := range e.liveState {
@@ -493,637 +356,3 @@ func (e *Engine) Start(ctx context.Context) {
 		e.maintenancePruner(ctx)
 	}()
 }
-
-func (e *Engine) maintenancePruner(ctx context.Context) {
-	ticker := time.NewTicker(maintPruneInterval)
-	defer ticker.Stop()
-
-	e.pruneMaintenanceWindows(ctx)
-
-	for {
-		select {
-		case <-ticker.C:
-			e.pruneMaintenanceWindows(ctx)
-		case <-ctx.Done():
-			return
-		}
-	}
-}
-
-func (e *Engine) pruneMaintenanceWindows(ctx context.Context) {
-	pruned, err := e.db.PruneExpiredMaintenanceWindows(ctx, e.maintRetention)
-	if err != nil {
-		e.AddLog(fmt.Sprintf("Maintenance prune error: %v", err))
-		return
-	}
-	if pruned > 0 {
-		e.AddLog(fmt.Sprintf("Pruned %d expired maintenance window(s)", pruned))
-	}
-}
-
-func (e *Engine) UpdateSiteConfig(cfg models.SiteConfig) {
-	e.mu.Lock()
-	if existing, ok := e.liveState[cfg.ID]; ok {
-		e.removeFromTokenIndex(cfg.ID)
-		existing.SiteConfig = cfg
-		e.liveState[cfg.ID] = existing
-		e.addToTokenIndex(existing)
-	}
-	e.mu.Unlock()
-
-	e.signalRecheck(cfg.ID)
-}
-
-func (e *Engine) getRecheckChan(id int) chan struct{} {
-	e.recheckMu.Lock()
-	defer e.recheckMu.Unlock()
-	ch, ok := e.recheck[id]
-	if !ok {
-		ch = make(chan struct{}, 1)
-		e.recheck[id] = ch
-	}
-	return ch
-}
-
-func (e *Engine) signalRecheck(id int) {
-	ch := e.getRecheckChan(id)
-	select {
-	case ch <- struct{}{}:
-	default:
-	}
-}
-
-func (e *Engine) RemoveSite(id int) {
-	e.mu.Lock()
-	e.removeFromTokenIndex(id)
-	delete(e.liveState, id)
-	e.mu.Unlock()
-	e.removeHistory(id)
-
-	e.probeResultsMu.Lock()
-	delete(e.probeResults, id)
-	e.probeResultsMu.Unlock()
-
-	e.recheckMu.Lock()
-	delete(e.recheck, id)
-	e.recheckMu.Unlock()
-}
-
-func (e *Engine) ToggleSitePause(id int) bool {
-	var (
-		paused bool
-		name   string
-	)
-	_, ok := e.applyState(id, func(s *models.Site) {
-		s.Paused = !s.Paused
-		paused = s.Paused
-		name = s.Name
-	})
-	if !ok {
-		return false
-	}
-	if paused {
-		e.AddLog(fmt.Sprintf("Monitor '%s' paused", name))
-	} else {
-		e.AddLog(fmt.Sprintf("Monitor '%s' resumed", name))
-	}
-	return paused
-}
-
-func (e *Engine) monitorRoutine(ctx context.Context, id int) {
-	recheckCh := e.getRecheckChan(id)
-
-	// Stagger initial check to avoid thundering herd on startup
-	stagger := time.Duration(rand.IntN(3000)) * time.Millisecond //nolint:gosec // non-security jitter
-	select {
-	case <-time.After(stagger):
-	case <-ctx.Done():
-		return
-	}
-
-	e.checkByID(ctx, id)
-	for {
-		select {
-		case <-ctx.Done():
-			return
-		default:
-		}
-
-		if !e.IsActive() {
-			select {
-			case <-time.After(pollInterval):
-			case <-ctx.Done():
-				return
-			case <-recheckCh:
-			}
-			continue
-		}
-
-		e.mu.RLock()
-		site, exists := e.liveState[id]
-		e.mu.RUnlock()
-		if !exists {
-			return
-		}
-
-		if site.Paused {
-			select {
-			case <-time.After(pollInterval):
-			case <-ctx.Done():
-				return
-			case <-recheckCh:
-			}
-			continue
-		}
-
-		interval := site.Interval
-		if interval < minCheckInterval {
-			interval = minCheckInterval
-		}
-		jitter := time.Duration(rand.IntN(interval*100)) * time.Millisecond //nolint:gosec // non-security jitter
-		select {
-		case <-time.After(time.Duration(interval)*time.Second + jitter):
-		case <-ctx.Done():
-			return
-		case <-recheckCh:
-		}
-		e.checkByID(ctx, id)
-	}
-}
-
-// applyState atomically reads, mutates, and writes back the live entry for id.
-// The mutator runs under the engine write lock and receives a pointer to the
-// CURRENT live state, so concurrent config edits, pauses, and heartbeats are
-// never clobbered by a stale snapshot. The mutator must only touch runtime /
-// check-result fields — config fields (Name/URL/Type/Token/Interval/AlertID/…)
-// are owned by UpdateSiteConfig and must not be written here. Returns the
-// post-mutation copy and whether the site still exists.
-func (e *Engine) applyState(id int, mutate func(s *models.Site)) (models.Site, bool) {
-	e.mu.Lock()
-	defer e.mu.Unlock()
-	cur, ok := e.liveState[id]
-	if !ok {
-		return models.Site{}, false
-	}
-	mutate(&cur)
-	e.liveState[id] = cur
-	return cur, true
-}
-
-func (e *Engine) checkByID(ctx context.Context, id int) {
-	if !e.IsActive() {
-		return
-	}
-
-	e.mu.RLock()
-	site, exists := e.liveState[id]
-	e.mu.RUnlock()
-	if !exists || site.Paused {
-		return
-	}
-
-	switch site.Type {
-	case "push":
-		e.checkPush(ctx, site)
-	case "group":
-		e.checkGroup(ctx, site)
-	default:
-		result := RunCheck(ctx, site.SiteConfig, e.strictClient, e.insecureClient, e.insecureSkipVerify, e.allowPrivateTargets)
-		updatedSite := site
-		updatedSite.HasSSL = result.HasSSL
-		updatedSite.CertExpiry = result.CertExpiry
-		updatedSite.Latency = time.Duration(result.LatencyNs)
-		updatedSite.LastCheck = time.Now()
-		e.handleStatusChange(updatedSite, result.Status, result.StatusCode, time.Duration(result.LatencyNs), result.ErrorReason)
-	}
-}
-
-func (e *Engine) checkPush(_ context.Context, site models.Site) {
-	if site.Status == models.StatusPending {
-		return
-	}
-
-	interval := time.Duration(site.Interval) * time.Second
-	grace := interval / 2
-	if grace < minPushGrace {
-		grace = minPushGrace
-	}
-
-	overdue := site.LastCheck.Add(interval)
-	staleMark := overdue.Add(grace / 2)
-	graceEnd := overdue.Add(grace)
-	now := time.Now()
-
-	if now.After(graceEnd) {
-		if site.Status != models.StatusDown {
-			e.handleStatusChange(site, string(models.StatusDown), 0, 0, "heartbeat missed")
-		}
-	} else if now.After(staleMark) {
-		if site.Status != models.StatusStale {
-			e.handleStatusChange(site, string(models.StatusStale), 0, 0, "heartbeat stale")
-		}
-	} else if now.After(overdue) {
-		if site.Status != models.StatusLate {
-			e.handleStatusChange(site, string(models.StatusLate), 0, 0, "heartbeat overdue")
-		}
-	}
-}
-
-// handleStatusChange folds a check result into the live state. snap is the
-// stale snapshot the check ran against; the actual mutation is applied onto the
-// CURRENT live entry via applyState, so a concurrent pause / config edit /
-// heartbeat is never reverted by this write. Logs and alerts are emitted after
-// the lock is released, off the critical section.
-func (e *Engine) handleStatusChange(snap models.Site, rawStatus string, code int, latency time.Duration, errorReason string) {
-	if !e.IsActive() {
-		return
-	}
-
-	inMaint := e.isInMaintenance(snap.ID)
-	status := models.Status(rawStatus)
-
-	var (
-		prev, next            models.Status
-		name, typ             string
-		alertID               int
-		failCount, maxRetries int
-		confirmedDown         bool
-		failedCheck           bool
-		downSince             time.Time
-		sslWarnFire           bool
-		sslDays               int
-		skipped               bool
-		changed               bool
-	)
-
-	_, exists := e.applyState(snap.ID, func(s *models.Site) {
-		// A non-UP result computed from a stale snapshot must not override a
-		// heartbeat (or newer check) that landed while we were evaluating.
-		if status != models.StatusUp && s.LastCheck.After(snap.LastCheck) {
-			skipped = true
-			return
-		}
-
-		prev = s.Status
-		name = s.Name
-		typ = s.Type
-		alertID = s.AlertID
-		maxRetries = s.MaxRetries
-		downSince = s.StatusChangedAt
-
-		// Fresh check results (measured by the run against snap).
-		s.StatusCode = code
-		s.Latency = snap.Latency
-		s.LastCheck = snap.LastCheck
-		s.HasSSL = snap.HasSSL
-		s.CertExpiry = snap.CertExpiry
-		s.LastError = errorReason
-		if status == models.StatusUp {
-			s.LastSuccessAt = time.Now()
-			s.LastError = ""
-		}
-
-		// Status + failure-count transition, based on the CURRENT live status.
-		if status == models.StatusUp {
-			s.FailureCount = 0
-			s.Status = models.StatusUp
-		} else {
-			if s.FailureCount <= s.MaxRetries {
-				s.FailureCount++
-			}
-			if s.FailureCount > s.MaxRetries {
-				if s.Status != status {
-					confirmedDown = true
-				}
-				s.Status = status
-				s.FailureCount = s.MaxRetries + 1
-			} else {
-				failedCheck = true
-			}
-		}
-		failCount = s.FailureCount
-
-		if s.Status != prev && prev != models.StatusPending {
-			s.StatusChangedAt = time.Now()
-		} else if s.StatusChangedAt.IsZero() && s.Status != models.StatusPending {
-			s.StatusChangedAt = time.Now()
-		}
-
-		// SSL expiry warning (fresh HasSSL/CertExpiry + config threshold).
-		if typ == "http" && s.CheckSSL && s.HasSSL {
-			days := int(time.Until(s.CertExpiry).Hours() / 24)
-			if days <= s.ExpiryThreshold && !s.SentSSLWarning && status != models.StatusSSLExp {
-				sslWarnFire = true
-				sslDays = days
-				s.SentSSLWarning = true
-			} else if days > s.ExpiryThreshold {
-				s.SentSSLWarning = false
-			}
-		}
-
-		next = s.Status
-		changed = next != prev
-	})
-
-	if !exists || skipped {
-		return
-	}
-
-	e.recordCheck(snap.ID, latency, status == models.StatusUp)
-
-	if confirmedDown {
-		if errorReason != "" {
-			e.AddLog(fmt.Sprintf("Monitor '%s' confirmed DOWN: %s", name, errorReason))
-		} else {
-			e.AddLog(fmt.Sprintf("Monitor '%s' confirmed DOWN", name))
-		}
-	} else if failedCheck {
-		e.AddLog(fmt.Sprintf("Monitor '%s' failed check %d/%d", name, failCount, maxRetries))
-	}
-
-	if changed && prev != models.StatusPending {
-		e.enqueueWrite(writeStateChange{siteID: snap.ID, fromStatus: string(prev), toStatus: string(next), reason: errorReason})
-	}
-
-	if sslWarnFire {
-		if !inMaint {
-			e.triggerAlert(alertID, "SSL WARNING", fmt.Sprintf("SSL for '%s' expires in %d days", name, sslDays))
-		} else {
-			e.AddLog(fmt.Sprintf("SSL warning for '%s' suppressed (maintenance)", name))
-		}
-	}
-
-	if prev == models.StatusUp && next == models.StatusLate {
-		e.AddLog(fmt.Sprintf("Monitor '%s' heartbeat overdue", name))
-	}
-
-	if !prev.IsBroken() && next.IsBroken() && next != models.StatusPending {
-		if inMaint {
-			e.AddLog(fmt.Sprintf("Monitor '%s' is DOWN (alerts suppressed — maintenance)", name))
-		} else {
-			msg := fmt.Sprintf("Monitor '%s' is DOWN (%s)", name, rawStatus)
-			if errorReason != "" {
-				msg = fmt.Sprintf("Monitor '%s' is DOWN: %s", name, errorReason)
-			}
-			if typ == "push" {
-				msg = fmt.Sprintf("Push Monitor '%s' missed heartbeat.", name)
-			}
-			e.triggerAlert(alertID, "🚨 ALERT", msg)
-		}
-	}
-	if prev.IsBroken() && next == models.StatusUp {
-		downDur := ""
-		if !downSince.IsZero() {
-			downDur = fmt.Sprintf(" (was down %s)", fmtDurationShort(time.Since(downSince)))
-		}
-		e.AddLog(fmt.Sprintf("Monitor '%s' recovered%s", name, downDur))
-		if !inMaint {
-			e.triggerAlert(alertID, "✅ RECOVERY", fmt.Sprintf("Monitor '%s' is UP%s", name, downDur))
-		}
-	}
-	if prev == models.StatusLate && next == models.StatusUp && !prev.IsBroken() {
-		e.AddLog(fmt.Sprintf("Monitor '%s' heartbeat arrived (was late)", name))
-	}
-}
-
-func (e *Engine) triggerAlert(alertID int, title, message string) {
-	if alertID <= 0 {
-		return
-	}
-	cfg, err := e.db.GetAlert(context.Background(), alertID)
-	if err != nil {
-		e.AddLog(fmt.Sprintf("Failed to load alert config %d: %v", alertID, err))
-		return
-	}
-	provider := alert.GetProvider(cfg)
-	if provider != nil {
-		go func() {
-			ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
-			defer cancel()
-			if err := provider.Send(ctx, title, message); err != nil {
-				e.AddLog(fmt.Sprintf("Alert send failed (%s): %v", cfg.Name, err))
-				e.recordAlertResult(alertID, false, err.Error())
-			} else {
-				e.recordAlertResult(alertID, true, "")
-			}
-		}()
-	}
-}
-
-func (e *Engine) recordAlertResult(alertID int, ok bool, errMsg string) {
-	e.alertHealthMu.Lock()
-	defer e.alertHealthMu.Unlock()
-	h := e.alertHealth[alertID]
-	h.LastSendAt = time.Now()
-	h.LastSendOK = ok
-	h.SendCount++
-	if ok {
-		h.LastError = ""
-	} else {
-		h.LastError = errMsg
-		h.FailCount++
-	}
-	e.alertHealth[alertID] = h
-
-	// Persist so health survives restarts; DB IO off the alert path.
-	e.enqueueWrite(writeAlertHealth{rec: models.AlertHealthRecord{
-		AlertID:    alertID,
-		LastSendAt: h.LastSendAt,
-		LastSendOK: h.LastSendOK,
-		LastError:  h.LastError,
-		SendCount:  h.SendCount,
-		FailCount:  h.FailCount,
-	}})
-}
-
-func (e *Engine) GetAlertHealth(alertID int) AlertHealth {
-	e.alertHealthMu.RLock()
-	defer e.alertHealthMu.RUnlock()
-	return e.alertHealth[alertID]
-}
-
-func (e *Engine) TestAlert(alertID int) error {
-	cfg, err := e.db.GetAlert(context.Background(), alertID)
-	if err != nil {
-		return fmt.Errorf("failed to load alert: %w", err)
-	}
-	provider := alert.GetProvider(cfg)
-	if provider == nil {
-		return fmt.Errorf("no provider for type %q", cfg.Type)
-	}
-	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
-	defer cancel()
-	err = provider.Send(ctx, "🧪 Test Alert", fmt.Sprintf("Test notification from uptop for channel '%s'.", cfg.Name))
-	if err != nil {
-		e.recordAlertResult(alertID, false, err.Error())
-		return err
-	}
-	e.recordAlertResult(alertID, true, "")
-	e.AddLog(fmt.Sprintf("Test alert sent to '%s'", cfg.Name))
-	return nil
-}
-
-func (e *Engine) isInMaintenance(monitorID int) bool {
-	e.maintCacheMu.RLock()
-	defer e.maintCacheMu.RUnlock()
-	return e.maintCache[monitorID]
-}
-
-func (e *Engine) refreshMaintenanceCache(ctx context.Context) {
-	windows, err := e.db.GetActiveMaintenanceWindows(ctx)
-	if err != nil {
-		return
-	}
-
-	directMaint := make(map[int]bool)
-	var globalMaint bool
-	for _, w := range windows {
-		if w.MonitorID == 0 {
-			globalMaint = true
-		} else {
-			directMaint[w.MonitorID] = true
-		}
-	}
-
-	resolved := make(map[int]bool)
-	e.mu.RLock()
-	for id, site := range e.liveState {
-		if globalMaint || directMaint[id] || (site.ParentID > 0 && directMaint[site.ParentID]) {
-			resolved[id] = true
-		}
-	}
-	e.mu.RUnlock()
-
-	e.maintCacheMu.Lock()
-	e.maintCache = resolved
-	e.maintCacheMu.Unlock()
-}
-
-func (e *Engine) GetDisplayStatus(site models.Site) string {
-	if site.Paused {
-		return "PAUSED"
-	}
-	if e.isInMaintenance(site.ID) {
-		return "MAINT"
-	}
-	return string(site.Status)
-}
-
-func (e *Engine) checkGroup(_ context.Context, site models.Site) {
-	e.mu.RLock()
-	status := models.StatusUp
-	hasChildren := false
-	for _, child := range e.liveState {
-		if child.ParentID != site.ID || child.Type == "group" {
-			continue
-		}
-		hasChildren = true
-		if child.Paused || e.isInMaintenance(child.ID) {
-			continue
-		}
-		if child.Status == models.StatusDown || child.Status == models.StatusSSLExp {
-			status = models.StatusDown
-		} else if child.Status == models.StatusStale && status != models.StatusDown {
-			status = models.StatusStale
-		} else if child.Status == models.StatusLate && status != models.StatusDown && status != models.StatusStale {
-			status = models.StatusLate
-		} else if child.Status == models.StatusPending && status != models.StatusDown && status != models.StatusStale && status != models.StatusLate {
-			status = models.StatusPending
-		}
-	}
-	e.mu.RUnlock()
-
-	if !hasChildren {
-		status = models.StatusPending
-	}
-
-	e.applyState(site.ID, func(s *models.Site) {
-		s.Status = status
-	})
-	e.recordCheck(site.ID, 0, !status.IsBroken())
-}
-
-func (e *Engine) EnqueueProbeCheck(siteID int, nodeID string, latencyNs int64, isUp bool) {
-	e.enqueueWrite(writeProbeCheck{siteID: siteID, nodeID: nodeID, latencyNs: latencyNs, isUp: isUp})
-}
-
-// SetAggStrategy must be called before Start: the field is read by the probe
-// aggregation path without synchronization.
-func (e *Engine) SetAggStrategy(strategy AggregationStrategy) {
-	e.aggStrategy = strategy
-}
-
-func (e *Engine) IngestProbeResult(nodeID string, siteID int, latencyNs int64, isUp bool, errorReason string) {
-	e.mu.RLock()
-	site, exists := e.liveState[siteID]
-	e.mu.RUnlock()
-	if !exists {
-		return
-	}
-
-	staleAfter := time.Duration(site.Interval) * time.Second * 3
-	if staleAfter < time.Minute {
-		staleAfter = time.Minute
-	}
-
-	now := time.Now()
-	e.probeResultsMu.Lock()
-	if e.probeResults[siteID] == nil {
-		e.probeResults[siteID] = make(map[string]NodeResult)
-	}
-	e.probeResults[siteID][nodeID] = NodeResult{
-		NodeID:      nodeID,
-		IsUp:        isUp,
-		LatencyNs:   latencyNs,
-		CheckedAt:   now,
-		ErrorReason: errorReason,
-	}
-	results := make([]NodeResult, 0, len(e.probeResults[siteID]))
-	for id, r := range e.probeResults[siteID] {
-		if now.Sub(r.CheckedAt) > staleAfter {
-			delete(e.probeResults[siteID], id)
-			continue
-		}
-		results = append(results, r)
-	}
-	e.probeResultsMu.Unlock()
-
-	aggUp, avgLatency := AggregateStatus(results, e.aggStrategy)
-
-	probeStatus := models.StatusUp
-	if !aggUp {
-		probeStatus = models.StatusDown
-	}
-
-	updatedSite := site
-	updatedSite.Latency = time.Duration(avgLatency)
-	updatedSite.LastCheck = time.Now()
-	e.handleStatusChange(updatedSite, string(probeStatus), 0, time.Duration(avgLatency), errorReason)
-}
-
-func (e *Engine) GetProbeResults(siteID int) map[string]NodeResult {
-	e.probeResultsMu.RLock()
-	defer e.probeResultsMu.RUnlock()
-	src := e.probeResults[siteID]
-	cp := make(map[string]NodeResult, len(src))
-	for k, v := range src {
-		cp[k] = v
-	}
-	return cp
-}
-
-func (e *Engine) GetStateChanges(siteID int, limit int) []models.StateChange {
-	changes, err := e.db.GetStateChanges(context.Background(), siteID, limit)
-	if err != nil {
-		return nil
-	}
-	return changes
-}
-
-func (e *Engine) GetStateChangesSince(siteID int, since time.Time) []models.StateChange {
-	changes, err := e.db.GetStateChangesSince(context.Background(), siteID, since)
-	if err != nil {
-		return nil
-	}
-	return changes
-}

internal/monitor/safedial.go

@@ -7,6 +7,8 @@ import (
 	"time"
 )
 
+const dialTimeout = 10 * time.Second
+
 var privateRanges []*net.IPNet
 
 func init() {
@@ -60,7 +62,7 @@ func SafeDialContext(allowPrivate bool) func(ctx context.Context, network, addr
 			}
 		}
 
-		dialer := &net.Dialer{Timeout: 10 * time.Second}
+		dialer := &net.Dialer{Timeout: dialTimeout}
 		for _, ip := range ips {
 			target := net.JoinHostPort(ip.IP.String(), port)
 			conn, err := dialer.DialContext(ctx, network, target)

internal/monitor/sites.go

@@ -0,0 +1,158 @@
+package monitor
+
+import (
+	"context"
+	"fmt"
+	"time"
+
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+)
+
+func (e *Engine) SetActive(active bool) {
+	e.activeMu.Lock()
+	defer e.activeMu.Unlock()
+	if e.isActive != active {
+		e.isActive = active
+		status := "RESUMED (Active)"
+		if !active {
+			status = "PAUSED (Passive)"
+		}
+		e.AddLog(fmt.Sprintf("Engine %s", status))
+	}
+}
+
+func (e *Engine) IsActive() bool {
+	e.activeMu.RLock()
+	defer e.activeMu.RUnlock()
+	return e.isActive
+}
+
+func (e *Engine) GetAllSites() []models.Site {
+	e.mu.RLock()
+	defer e.mu.RUnlock()
+	sites := make([]models.Site, 0, len(e.liveState))
+	for _, s := range e.liveState {
+		sites = append(sites, s)
+	}
+	return sites
+}
+
+func (e *Engine) GetLiveState() map[int]models.Site {
+	e.mu.RLock()
+	defer e.mu.RUnlock()
+	cp := make(map[int]models.Site, len(e.liveState))
+	for k, v := range e.liveState {
+		cp[k] = v
+	}
+	return cp
+}
+
+func (e *Engine) addToTokenIndex(site models.Site) {
+	if site.Type == "push" && site.Token != "" {
+		e.tokenIndex[site.Token] = site.ID
+	}
+}
+
+func (e *Engine) removeFromTokenIndex(id int) {
+	for token, sid := range e.tokenIndex {
+		if sid == id {
+			delete(e.tokenIndex, token)
+			return
+		}
+	}
+}
+
+func (e *Engine) UpdateSiteConfig(cfg models.SiteConfig) {
+	e.mu.Lock()
+	if existing, ok := e.liveState[cfg.ID]; ok {
+		e.removeFromTokenIndex(cfg.ID)
+		existing.SiteConfig = cfg
+		e.liveState[cfg.ID] = existing
+		e.addToTokenIndex(existing)
+	}
+	e.mu.Unlock()
+
+	e.signalRecheck(cfg.ID)
+}
+
+func (e *Engine) RemoveSite(id int) {
+	e.mu.Lock()
+	e.removeFromTokenIndex(id)
+	delete(e.liveState, id)
+	e.mu.Unlock()
+	e.removeHistory(id)
+
+	e.probeResultsMu.Lock()
+	delete(e.probeResults, id)
+	e.probeResultsMu.Unlock()
+
+	e.recheckMu.Lock()
+	delete(e.recheck, id)
+	e.recheckMu.Unlock()
+}
+
+func (e *Engine) ToggleSitePause(id int) bool {
+	var (
+		paused bool
+		name   string
+	)
+	_, ok := e.applyState(id, func(s *models.Site) {
+		s.Paused = !s.Paused
+		paused = s.Paused
+		name = s.Name
+	})
+	if !ok {
+		return false
+	}
+	if paused {
+		e.AddLog(fmt.Sprintf("Monitor '%s' paused", name))
+	} else {
+		e.AddLog(fmt.Sprintf("Monitor '%s' resumed", name))
+	}
+	return paused
+}
+
+// applyState atomically reads, mutates, and writes back the live entry for id.
+// The mutator runs under the engine write lock and receives a pointer to the
+// CURRENT live state, so concurrent config edits, pauses, and heartbeats are
+// never clobbered by a stale snapshot. The mutator must only touch runtime /
+// check-result fields — config fields (Name/URL/Type/Token/Interval/AlertID/…)
+// are owned by UpdateSiteConfig and must not be written here. Returns the
+// post-mutation copy and whether the site still exists.
+func (e *Engine) applyState(id int, mutate func(s *models.Site)) (models.Site, bool) {
+	e.mu.Lock()
+	defer e.mu.Unlock()
+	cur, ok := e.liveState[id]
+	if !ok {
+		return models.Site{}, false
+	}
+	mutate(&cur)
+	e.liveState[id] = cur
+	return cur, true
+}
+
+func (e *Engine) GetDisplayStatus(site models.Site) string {
+	if site.Paused {
+		return "PAUSED"
+	}
+	if e.isInMaintenance(site.ID) {
+		return "MAINT"
+	}
+	return string(site.Status)
+}
+
+func (e *Engine) GetStateChanges(ctx context.Context, siteID int, limit int) []models.StateChange {
+	changes, err := e.db.GetStateChanges(ctx, siteID, limit)
+	if err != nil {
+		return nil
+	}
+	return changes
+}
+
+func (e *Engine) GetStateChangesSince(ctx context.Context, siteID int, since time.Time) []models.StateChange {
+	changes, err := e.db.GetStateChangesSince(ctx, siteID, since)
+	if err != nil {
+		return nil
+	}
+	return changes
+}

internal/monitor/sites_test.go

@@ -0,0 +1,133 @@
+package monitor
+
+import (
+	"testing"
+	"time"
+
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+)
+
+func TestUpdateSiteConfig_PreservesRuntime(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", URL: "http://old.com"},
+		SiteState:  models.SiteState{Status: "DOWN", FailureCount: 3, Latency: 100 * time.Millisecond},
+	}
+	injectSite(e, site)
+
+	updated := models.SiteConfig{ID: 1, Name: "test", URL: "http://new.com", Interval: 60}
+	e.UpdateSiteConfig(updated)
+
+	s, _ := getSite(e, 1)
+	if s.URL != "http://new.com" {
+		t.Errorf("expected URL updated, got %s", s.URL)
+	}
+	if s.Status != "DOWN" {
+		t.Errorf("expected Status preserved, got %s", s.Status)
+	}
+	if s.FailureCount != 3 {
+		t.Errorf("expected FailureCount preserved, got %d", s.FailureCount)
+	}
+	if s.Latency != 100*time.Millisecond {
+		t.Errorf("expected Latency preserved, got %v", s.Latency)
+	}
+}
+
+func TestRemoveSite_CleansUp(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test", Type: "push", Token: "tok1"},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	injectSite(e, site)
+	e.recordCheck(1, 5*time.Millisecond, true)
+
+	e.RemoveSite(1)
+
+	if _, ok := getSite(e, 1); ok {
+		t.Error("expected site removed from liveState")
+	}
+	if e.RecordHeartbeat("tok1") {
+		t.Error("expected token removed from index")
+	}
+	if _, ok := e.GetHistory(1); ok {
+		t.Error("expected history removed")
+	}
+}
+
+func TestToggleSitePause(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	site := models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "test"},
+		SiteState:  models.SiteState{Status: "UP"},
+	}
+	injectSite(e, site)
+
+	paused := e.ToggleSitePause(1)
+	if !paused {
+		t.Error("expected paused=true after first toggle")
+	}
+	s, _ := getSite(e, 1)
+	if !s.Paused {
+		t.Error("expected Paused=true in state")
+	}
+
+	paused = e.ToggleSitePause(1)
+	if paused {
+		t.Error("expected paused=false after second toggle")
+	}
+}
+
+func TestToggleSitePause_NonexistentSite(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	if e.ToggleSitePause(999) {
+		t.Error("expected false for nonexistent site")
+	}
+}
+
+func TestGetAllSites_ReturnsCopy(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	injectSite(e, models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "s1"},
+		SiteState:  models.SiteState{Status: "UP"},
+	})
+	injectSite(e, models.Site{
+		SiteConfig: models.SiteConfig{ID: 2, Name: "s2"},
+		SiteState:  models.SiteState{Status: "DOWN"},
+	})
+
+	sites := e.GetAllSites()
+	if len(sites) != 2 {
+		t.Fatalf("expected 2 sites, got %d", len(sites))
+	}
+	sites[0].Name = "mutated"
+
+	fresh := e.GetAllSites()
+	for _, s := range fresh {
+		if s.Name == "mutated" {
+			t.Error("GetAllSites returned reference, not copy")
+		}
+	}
+}
+
+func TestGetLiveState_ReturnsCopy(t *testing.T) {
+	ms := newMockStore()
+	e := newTestEngine(ms)
+	injectSite(e, models.Site{
+		SiteConfig: models.SiteConfig{ID: 1, Name: "s1"},
+		SiteState:  models.SiteState{Status: "UP"},
+	})
+
+	state := e.GetLiveState()
+	state[1] = models.Site{SiteConfig: models.SiteConfig{Name: "mutated"}}
+
+	fresh := e.GetLiveState()
+	if fresh[1].Name == "mutated" {
+		t.Error("GetLiveState returned reference, not copy")
+	}
+}

internal/server/ratelimit.go

@@ -14,6 +14,11 @@ import (
 // guard.
 const maxVisitors = 10000
 
+const (
+	visitorCleanupInterval = 5 * time.Minute
+	visitorIdleCutoff      = 10 * time.Minute
+)
+
 type visitor struct {
 	tokens   float64
 	lastSeen time.Time
@@ -90,13 +95,13 @@ func (rl *RateLimiter) evictOldest() {
 }
 
 func (rl *RateLimiter) cleanup() {
-	ticker := time.NewTicker(5 * time.Minute)
+	ticker := time.NewTicker(visitorCleanupInterval)
 	defer ticker.Stop()
 	for {
 		select {
 		case <-ticker.C:
 			rl.mu.Lock()
-			cutoff := time.Now().Add(-10 * time.Minute)
+			cutoff := time.Now().Add(-visitorIdleCutoff)
 			for ip, v := range rl.visitors {
 				if v.lastSeen.Before(cutoff) {
 					delete(rl.visitors, ip)

internal/server/server.go

@@ -45,15 +45,27 @@ type Server struct {
 	statusRL *RateLimiter
 }
 
+const (
+	pushRateLimit   = 60
+	probeRateLimit  = 30
+	backupRateLimit = 10
+	statusRateLimit = 120
+
+	httpReadHeaderTimeout = 10 * time.Second
+	httpReadTimeout       = 30 * time.Second
+	httpWriteTimeout      = 60 * time.Second
+	httpIdleTimeout       = 120 * time.Second
+)
+
 func NewServer(cfg ServerConfig, s store.Store, eng *monitor.Engine) *Server {
 	return &Server{
 		cfg:      cfg,
 		store:    s,
 		eng:      eng,
-		pushRL:   NewRateLimiter(60, cfg.TrustedProxies),
-		probeRL:  NewRateLimiter(30, cfg.TrustedProxies),
-		backupRL: NewRateLimiter(10, cfg.TrustedProxies),
-		statusRL: NewRateLimiter(120, cfg.TrustedProxies),
+		pushRL:   NewRateLimiter(pushRateLimit, cfg.TrustedProxies),
+		probeRL:  NewRateLimiter(probeRateLimit, cfg.TrustedProxies),
+		backupRL: NewRateLimiter(backupRateLimit, cfg.TrustedProxies),
+		statusRL: NewRateLimiter(statusRateLimit, cfg.TrustedProxies),
 	}
 }
 
@@ -77,10 +89,10 @@ func (s *Server) Start() *http.Server {
 	httpSrv := &http.Server{
 		Addr:              addr,
 		Handler:           handler,
-		ReadHeaderTimeout: 10 * time.Second,
-		ReadTimeout:       30 * time.Second,
-		WriteTimeout:      60 * time.Second,
-		IdleTimeout:       120 * time.Second,
+		ReadHeaderTimeout: httpReadHeaderTimeout,
+		ReadTimeout:       httpReadTimeout,
+		WriteTimeout:      httpWriteTimeout,
+		IdleTimeout:       httpIdleTimeout,
 	}
 	go func() {
 		if s.cfg.TLSCert != "" && s.cfg.TLSKey != "" {
@@ -159,10 +171,6 @@ func (s *Server) handleHealth(w http.ResponseWriter, r *http.Request) {
 		http.Error(w, "Method not allowed", http.StatusMethodNotAllowed)
 		return
 	}
-	if s.cfg.ClusterKey != "" && !checkSecret(r.Header.Get("X-Uptop-Secret"), s.cfg.ClusterKey) {
-		http.Error(w, "Unauthorized", http.StatusUnauthorized)
-		return
-	}
 	w.WriteHeader(http.StatusOK)
 	_, _ = w.Write([]byte("OK"))
 }

internal/server/server_test.go

@@ -77,8 +77,12 @@ func (m *mockStore) GetActiveMaintenanceWindows(_ context.Context) ([]models.Mai
 
 // --- Helpers ---
 
-func freePort() int {
-	ln, _ := net.Listen("tcp", "127.0.0.1:0")
+func freePort(t *testing.T) int {
+	t.Helper()
+	ln, err := net.Listen("tcp", "127.0.0.1:0")
+	if err != nil {
+		t.Fatalf("freePort: %v", err)
+	}
 	port := ln.Addr().(*net.TCPAddr).Port
 	ln.Close()
 	return port
@@ -95,7 +99,7 @@ func newTestServer(t *testing.T, clusterKey string, enableStatus bool) *testServ
 	t.Helper()
 	ms := newMockStore()
 	eng := monitor.NewEngine(ms)
-	port := freePort()
+	port := freePort(t)
 
 	srv := Start(ServerConfig{
 		Port:         port,
@@ -219,8 +223,8 @@ func TestHealth_WrongSecret(t *testing.T) {
 		t.Fatal(err)
 	}
 	defer resp.Body.Close()
-	if resp.StatusCode != 401 {
-		t.Errorf("expected 401, got %d", resp.StatusCode)
+	if resp.StatusCode != 200 {
+		t.Errorf("health is unauthenticated, expected 200, got %d", resp.StatusCode)
 	}
 }
 
@@ -285,7 +289,10 @@ func TestImport_MethodNotAllowed(t *testing.T) {
 
 func TestImport_Unauthorized(t *testing.T) {
 	ts := newTestServer(t, "secret", false)
-	body, _ := json.Marshal(models.Backup{})
+	body, err := json.Marshal(models.Backup{})
+	if err != nil {
+		t.Fatalf("json.Marshal: %v", err)
+	}
 	resp, err := authReq("POST", ts.baseURL+"/api/backup/import", "wrong", body)
 	if err != nil {
 		t.Fatal(err)
@@ -301,7 +308,10 @@ func TestImport_Success(t *testing.T) {
 	backup := models.Backup{
 		Sites: []models.SiteConfig{{Name: "imported", URL: "http://example.com"}},
 	}
-	body, _ := json.Marshal(backup)
+	body, err := json.Marshal(backup)
+	if err != nil {
+		t.Fatalf("json.Marshal: %v", err)
+	}
 	resp, err := authReq("POST", ts.baseURL+"/api/backup/import", "secret", body)
 	if err != nil {
 		t.Fatal(err)
@@ -333,9 +343,12 @@ func TestImport_InvalidJSON(t *testing.T) {
 
 func TestProbeRegister_Success(t *testing.T) {
 	ts := newTestServer(t, "secret", false)
-	body, _ := json.Marshal(map[string]string{
+	body, err := json.Marshal(map[string]string{
 		"id": "node-1", "name": "US East", "region": "us-east",
 	})
+	if err != nil {
+		t.Fatalf("json.Marshal: %v", err)
+	}
 	resp, err := authReq("POST", ts.baseURL+"/api/probe/register", "secret", body)
 	if err != nil {
 		t.Fatal(err)
@@ -356,7 +369,10 @@ func TestProbeRegister_Success(t *testing.T) {
 
 func TestProbeRegister_MissingID(t *testing.T) {
 	ts := newTestServer(t, "secret", false)
-	body, _ := json.Marshal(map[string]string{"name": "test"})
+	body, err := json.Marshal(map[string]string{"name": "test"})
+	if err != nil {
+		t.Fatalf("json.Marshal: %v", err)
+	}
 	resp, err := authReq("POST", ts.baseURL+"/api/probe/register", "secret", body)
 	if err != nil {
 		t.Fatal(err)
@@ -369,7 +385,10 @@ func TestProbeRegister_MissingID(t *testing.T) {
 
 func TestProbeRegister_Unauthorized(t *testing.T) {
 	ts := newTestServer(t, "secret", false)
-	body, _ := json.Marshal(map[string]string{"id": "node-1"})
+	body, err := json.Marshal(map[string]string{"id": "node-1"})
+	if err != nil {
+		t.Fatalf("json.Marshal: %v", err)
+	}
 	resp, err := authReq("POST", ts.baseURL+"/api/probe/register", "wrong", body)
 	if err != nil {
 		t.Fatal(err)
@@ -384,12 +403,15 @@ func TestProbeRegister_Unauthorized(t *testing.T) {
 
 func TestProbeResults_Success(t *testing.T) {
 	ts := newTestServer(t, "secret", false)
-	body, _ := json.Marshal(map[string]any{
+	body, err := json.Marshal(map[string]any{
 		"node_id": "node-1",
 		"results": []map[string]any{
 			{"site_id": 1, "latency_ns": 5000000, "is_up": true},
 		},
 	})
+	if err != nil {
+		t.Fatalf("json.Marshal: %v", err)
+	}
 	resp, err := authReq("POST", ts.baseURL+"/api/probe/results", "secret", body)
 	if err != nil {
 		t.Fatal(err)
@@ -402,9 +424,12 @@ func TestProbeResults_Success(t *testing.T) {
 
 func TestProbeResults_MissingNodeID(t *testing.T) {
 	ts := newTestServer(t, "secret", false)
-	body, _ := json.Marshal(map[string]any{
+	body, err := json.Marshal(map[string]any{
 		"results": []map[string]any{},
 	})
+	if err != nil {
+		t.Fatalf("json.Marshal: %v", err)
+	}
 	resp, err := authReq("POST", ts.baseURL+"/api/probe/results", "secret", body)
 	if err != nil {
 		t.Fatal(err)
@@ -573,7 +598,10 @@ func TestClientIP_TrustedProxyHandling(t *testing.T) {
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			r, _ := http.NewRequest(http.MethodGet, "/", nil)
+			r, err := http.NewRequest(http.MethodGet, "/", nil)
+			if err != nil {
+				t.Fatalf("http.NewRequest: %v", err)
+			}
 			r.RemoteAddr = tt.remoteAddr
 			if tt.xff != "" {
 				r.Header.Set("X-Forwarded-For", tt.xff)
@@ -591,7 +619,10 @@ func TestRateLimit_SpoofedXFFCannotBypass(t *testing.T) {
 	rl := NewRateLimiter(60, nil) // no trusted proxies
 	allowed := 0
 	for i := 0; i < 200; i++ {
-		r, _ := http.NewRequest(http.MethodGet, "/", nil)
+		r, err := http.NewRequest(http.MethodGet, "/", nil)
+		if err != nil {
+			t.Fatalf("http.NewRequest: %v", err)
+		}
 		r.RemoteAddr = "203.0.113.9:5000"
 		r.Header.Set("X-Forwarded-For", fmt.Sprintf("9.9.9.%d", i%256))
 		if rl.Allow(clientIP(r, rl.trusted)) {

internal/store/crypto_test.go

@@ -75,8 +75,14 @@ func TestEncryptorUniqueCiphertexts(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	a, _ := enc.Encrypt("same")
-	b, _ := enc.Encrypt("same")
+	a, err := enc.Encrypt("same")
+	if err != nil {
+		t.Fatalf("first encrypt: %v", err)
+	}
+	b, err := enc.Encrypt("same")
+	if err != nil {
+		t.Fatalf("second encrypt: %v", err)
+	}
 	if a == b {
 		t.Error("two encryptions of same plaintext should produce different ciphertexts")
 	}

internal/store/sqlstore.go

@@ -5,8 +5,8 @@ import (
 	"crypto/rand"
 	"database/sql"
 	"encoding/hex"
-	"encoding/json"
 	"fmt"
+	"strings"
 	"time"
 
 	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
@@ -17,6 +17,10 @@ const (
 	maxLogRows             = 200
 	maxStateChangesPerSite = 5000
 	maxMaintenanceExport   = 1000
+	maxOpenConns           = 25
+	maxIdleConns           = 5
+	connMaxLifetime        = 5 * time.Minute
+	tokenByteLen           = 16
 )
 
 type SQLStore struct {
@@ -31,9 +35,9 @@ func NewSQLStore(driverName, dsn string, dialect Dialect) (*SQLStore, error) {
 	if err != nil {
 		return nil, err
 	}
-	db.SetMaxOpenConns(25)
-	db.SetMaxIdleConns(5)
-	db.SetConnMaxLifetime(5 * time.Minute)
+	db.SetMaxOpenConns(maxOpenConns)
+	db.SetMaxIdleConns(maxIdleConns)
+	db.SetConnMaxLifetime(connMaxLifetime)
 	_, isDollar := dialect.(*PostgresDialect)
 	return &SQLStore{db: db, dialect: dialect, dollar: isDollar}, nil
 }
@@ -61,7 +65,7 @@ func (s *SQLStore) q(query string) string {
 }
 
 func generateToken() (string, error) {
-	b := make([]byte, 16)
+	b := make([]byte, tokenByteLen)
 	if _, err := rand.Read(b); err != nil {
 		return "", fmt.Errorf("crypto/rand failed: %w", err)
 	}
@@ -213,40 +217,6 @@ func (s *SQLStore) GetSiteByName(ctx context.Context, name string) (models.SiteC
 	return st, err
 }
 
-func (s *SQLStore) unmarshalSettings(raw string) (map[string]string, error) {
-	decrypted, err := s.decryptSettings(raw)
-	if err != nil {
-		return nil, fmt.Errorf("decrypt settings: %w", err)
-	}
-	var m map[string]string
-	if err := json.Unmarshal([]byte(decrypted), &m); err != nil {
-		return nil, fmt.Errorf("unmarshal settings: %w", err)
-	}
-	return m, nil
-}
-
-func (s *SQLStore) marshalSettings(settings map[string]string) (string, error) {
-	jsonBytes, err := json.Marshal(settings)
-	if err != nil {
-		return "", err
-	}
-	return s.encryptSettings(string(jsonBytes))
-}
-
-func (s *SQLStore) GetAlertByName(ctx context.Context, name string) (models.AlertConfig, error) {
-	var a models.AlertConfig
-	var settingsRaw string
-	err := s.db.QueryRowContext(ctx, s.q("SELECT id, name, type, settings FROM alerts WHERE name = ?"), name).Scan(&a.ID, &a.Name, &a.Type, &settingsRaw)
-	if err != nil {
-		return a, err
-	}
-	a.Settings, err = s.unmarshalSettings(settingsRaw)
-	if err != nil {
-		return a, fmt.Errorf("alert %q: %w", name, err)
-	}
-	return a, nil
-}
-
 func (s *SQLStore) AddSiteReturningID(ctx context.Context, site models.SiteConfig) (int, error) {
 	token := ""
 	if site.Type == "push" {
@@ -273,89 +243,6 @@ func (s *SQLStore) AddSiteReturningID(ctx context.Context, site models.SiteConfi
 	return int(id), err
 }
 
-func (s *SQLStore) AddAlertReturningID(ctx context.Context, name, aType string, settings map[string]string) (int, error) {
-	stored, err := s.marshalSettings(settings)
-	if err != nil {
-		return 0, err
-	}
-	if s.dollar {
-		var id int
-		err := s.db.QueryRowContext(ctx, s.q("INSERT INTO alerts (name, type, settings) VALUES (?, ?, ?) RETURNING id"), name, aType, stored).Scan(&id)
-		return id, err
-	}
-	result, err := s.db.ExecContext(ctx, s.q("INSERT INTO alerts (name, type, settings) VALUES (?, ?, ?)"), name, aType, stored)
-	if err != nil {
-		return 0, err
-	}
-	id, err := result.LastInsertId()
-	return int(id), err
-}
-
-func (s *SQLStore) GetAllAlerts(ctx context.Context) ([]models.AlertConfig, error) {
-	rows, err := s.db.QueryContext(ctx, "SELECT id, name, type, settings FROM alerts")
-	if err != nil {
-		return nil, err
-	}
-	defer rows.Close()
-	var alerts []models.AlertConfig
-	for rows.Next() {
-		var a models.AlertConfig
-		var settingsRaw string
-		if err := rows.Scan(&a.ID, &a.Name, &a.Type, &settingsRaw); err != nil {
-			return alerts, err
-		}
-		a.Settings, err = s.unmarshalSettings(settingsRaw)
-		if err != nil {
-			return alerts, fmt.Errorf("alert %q: %w", a.Name, err)
-		}
-		alerts = append(alerts, a)
-	}
-	return alerts, rows.Err()
-}
-
-func (s *SQLStore) GetAlert(ctx context.Context, id int) (models.AlertConfig, error) {
-	var a models.AlertConfig
-	var settingsRaw string
-	err := s.db.QueryRowContext(ctx, s.q("SELECT id, name, type, settings FROM alerts WHERE id = ?"), id).Scan(&a.ID, &a.Name, &a.Type, &settingsRaw)
-	if err != nil {
-		return a, err
-	}
-	a.Settings, err = s.unmarshalSettings(settingsRaw)
-	if err != nil {
-		return a, fmt.Errorf("alert %d: %w", id, err)
-	}
-	return a, nil
-}
-
-func (s *SQLStore) AddAlert(ctx context.Context, name, aType string, settings map[string]string) error {
-	stored, err := s.marshalSettings(settings)
-	if err != nil {
-		return err
-	}
-	_, err = s.db.ExecContext(ctx, s.q("INSERT INTO alerts (name, type, settings) VALUES (?, ?, ?)"), name, aType, stored)
-	return err
-}
-
-func (s *SQLStore) UpdateAlert(ctx context.Context, id int, name, aType string, settings map[string]string) error {
-	stored, err := s.marshalSettings(settings)
-	if err != nil {
-		return err
-	}
-	_, err = s.db.ExecContext(ctx, s.q("UPDATE alerts SET name=?, type=?, settings=? WHERE id=?"), name, aType, stored, id)
-	return err
-}
-
-func (s *SQLStore) DeleteAlert(ctx context.Context, id int) error {
-	if _, err := s.db.ExecContext(ctx, s.q("UPDATE sites SET alert_id = 0 WHERE alert_id = ?"), id); err != nil {
-		return err
-	}
-	if _, err := s.db.ExecContext(ctx, s.q("DELETE FROM alerts WHERE id=?"), id); err != nil {
-		return err
-	}
-	s.dialect.ResetSequenceOnEmpty(s.db, "alerts")
-	return nil
-}
-
 func (s *SQLStore) GetAllUsers(ctx context.Context) ([]models.User, error) {
 	rows, err := s.db.QueryContext(ctx, "SELECT id, username, public_key, role FROM users")
 	if err != nil {
@@ -388,85 +275,6 @@ func (s *SQLStore) DeleteUser(ctx context.Context, id int) error {
 	return err
 }
 
-func (s *SQLStore) SaveStateChange(ctx context.Context, siteID int, fromStatus, toStatus, errorReason string) error {
-	_, err := s.db.ExecContext(ctx, s.q("INSERT INTO state_changes (site_id, from_status, to_status, error_reason) VALUES (?, ?, ?, ?)"),
-		siteID, fromStatus, toStatus, errorReason)
-	return err
-}
-
-func (s *SQLStore) GetStateChanges(ctx context.Context, siteID int, limit int) ([]models.StateChange, error) {
-	rows, err := s.db.QueryContext(ctx, s.q("SELECT id, site_id, from_status, to_status, error_reason, changed_at FROM state_changes WHERE site_id = ? ORDER BY changed_at DESC LIMIT ?"), siteID, limit)
-	if err != nil {
-		return nil, err
-	}
-	defer rows.Close()
-	var changes []models.StateChange
-	for rows.Next() {
-		var sc models.StateChange
-		if err := rows.Scan(&sc.ID, &sc.SiteID, &sc.FromStatus, &sc.ToStatus, &sc.ErrorReason, &sc.ChangedAt); err != nil {
-			return changes, err
-		}
-		changes = append(changes, sc)
-	}
-	return changes, rows.Err()
-}
-
-func (s *SQLStore) GetStateChangesSince(ctx context.Context, siteID int, since time.Time) ([]models.StateChange, error) {
-	rows, err := s.db.QueryContext(ctx, s.q("SELECT id, site_id, from_status, to_status, error_reason, changed_at FROM state_changes WHERE site_id = ? AND changed_at >= ? ORDER BY changed_at DESC"), siteID, since)
-	if err != nil {
-		return nil, err
-	}
-	defer rows.Close()
-	var changes []models.StateChange
-	for rows.Next() {
-		var sc models.StateChange
-		if err := rows.Scan(&sc.ID, &sc.SiteID, &sc.FromStatus, &sc.ToStatus, &sc.ErrorReason, &sc.ChangedAt); err != nil {
-			return changes, err
-		}
-		changes = append(changes, sc)
-	}
-	return changes, rows.Err()
-}
-
-func (s *SQLStore) SaveCheck(ctx context.Context, siteID int, latencyNs int64, isUp bool) error {
-	return s.SaveCheckFromNode(ctx, siteID, "", latencyNs, isUp)
-}
-
-// SaveCheckFromNode inserts a single check row. Retention is handled out of
-// band by PruneCheckHistory on a timer, not per-insert, to keep the write hot
-// path a plain INSERT.
-func (s *SQLStore) SaveCheckFromNode(ctx context.Context, siteID int, nodeID string, latencyNs int64, isUp bool) error {
-	_, err := s.db.ExecContext(ctx, s.q("INSERT INTO check_history (site_id, node_id, latency_ns, is_up) VALUES (?, ?, ?, ?)"), siteID, nodeID, latencyNs, isUp)
-	return err
-}
-
-// PruneCheckHistory trims check_history to the newest maxCheckHistory rows per
-// site, across all sites, in one pass. Intended to run periodically.
-func (s *SQLStore) PruneCheckHistory(ctx context.Context) error {
-	q := fmt.Sprintf(`DELETE FROM check_history WHERE id IN (
-		SELECT id FROM (
-			SELECT id, ROW_NUMBER() OVER (PARTITION BY site_id ORDER BY checked_at DESC, id DESC) AS rn
-			FROM check_history
-		) ranked WHERE rn > %d
-	)`, maxCheckHistory)
-	_, err := s.db.ExecContext(ctx, s.q(q))
-	return err
-}
-
-// PruneStateChanges trims state_changes to the newest maxStateChangesPerSite
-// rows per site. Generous so realistic SLA windows are unaffected; bounds the
-// otherwise unbounded growth of a flapping monitor's history.
-func (s *SQLStore) PruneStateChanges(ctx context.Context) error {
-	q := fmt.Sprintf(`DELETE FROM state_changes WHERE id IN (
-		SELECT id FROM (
-			SELECT id, ROW_NUMBER() OVER (PARTITION BY site_id ORDER BY changed_at DESC, id DESC) AS rn
-			FROM state_changes
-		) ranked WHERE rn > %d
-	)`, maxStateChangesPerSite)
-	_, err := s.db.ExecContext(ctx, s.q(q))
-	return err
-}
-
 func (s *SQLStore) RegisterNode(ctx context.Context, node models.ProbeNode) error {
 	_, err := s.db.ExecContext(ctx, s.dialect.UpsertNodeSQL(), node.ID, node.Name, node.Region, node.Version)
 	return err
@@ -554,145 +362,27 @@ func (s *SQLStore) PruneLogs(ctx context.Context) error {
 	return err
 }
 
-func (s *SQLStore) LoadLogs(ctx context.Context, limit int) ([]string, error) {
-	rows, err := s.db.QueryContext(ctx, s.q("SELECT message FROM logs ORDER BY created_at DESC LIMIT ?"), limit)
+func (s *SQLStore) LoadLogs(ctx context.Context, limit int) ([]models.LogEntry, error) {
+	rows, err := s.db.QueryContext(ctx, s.q("SELECT message, created_at FROM logs ORDER BY created_at DESC, id DESC LIMIT ?"), limit)
 	if err != nil {
 		return nil, err
 	}
 	defer rows.Close()
-	var logs []string
+	var entries []models.LogEntry
 	for rows.Next() {
-		var msg string
-		if err := rows.Scan(&msg); err != nil {
-			return logs, err
+		var e models.LogEntry
+		if err := rows.Scan(&e.Message, &e.CreatedAt); err != nil {
+			return entries, err
 		}
-		logs = append(logs, msg)
-	}
-	return logs, rows.Err()
-}
-
-func (s *SQLStore) LoadAllHistory(ctx context.Context, limit int) (map[int][]models.CheckRecord, error) {
-	result := make(map[int][]models.CheckRecord)
-	rows, err := s.db.QueryContext(ctx, s.q(`
-		SELECT site_id, latency_ns, is_up FROM (
-			SELECT site_id, latency_ns, is_up,
-				ROW_NUMBER() OVER (PARTITION BY site_id ORDER BY checked_at DESC) AS rn
-			FROM check_history
-		) sub WHERE rn <= ?`), limit)
-	if err != nil {
-		return result, err
-	}
-	defer rows.Close()
-	for rows.Next() {
-		var r models.CheckRecord
-		if err := rows.Scan(&r.SiteID, &r.LatencyNs, &r.IsUp); err != nil {
-			return result, err
+		// Strip legacy [HH:MM] or [HH:MM:SS] prefix from pre-refactor entries.
+		if len(e.Message) > 3 && e.Message[0] == '[' {
+			if idx := strings.Index(e.Message, "]"); idx > 0 && idx < 12 {
+				e.Message = strings.TrimSpace(e.Message[idx+1:])
+			}
 		}
-		result[r.SiteID] = append(result[r.SiteID], r)
+		entries = append(entries, e)
 	}
-	for id, records := range result {
-		for i, j := 0, len(records)-1; i < j; i, j = i+1, j-1 {
-			records[i], records[j] = records[j], records[i]
-		}
-		result[id] = records
-	}
-	return result, rows.Err()
-}
-
-func (s *SQLStore) scanMaintenanceWindow(rows *sql.Rows) (models.MaintenanceWindow, error) {
-	var mw models.MaintenanceWindow
-	var endTime sql.NullTime
-	if err := rows.Scan(&mw.ID, &mw.MonitorID, &mw.Title, &mw.Description, &mw.Type, &mw.StartTime, &endTime, &mw.CreatedBy, &mw.CreatedAt); err != nil {
-		return mw, err
-	}
-	if endTime.Valid {
-		mw.EndTime = endTime.Time
-	}
-	return mw, nil
-}
-
-func (s *SQLStore) GetActiveMaintenanceWindows(ctx context.Context) ([]models.MaintenanceWindow, error) {
-	rows, err := s.db.QueryContext(ctx, s.q("SELECT id, monitor_id, title, description, type, start_time, end_time, created_by, created_at FROM maintenance_windows WHERE start_time <= CURRENT_TIMESTAMP AND (end_time IS NULL OR end_time > CURRENT_TIMESTAMP) ORDER BY start_time DESC"))
-	if err != nil {
-		return nil, err
-	}
-	defer rows.Close()
-	var windows []models.MaintenanceWindow
-	for rows.Next() {
-		mw, err := s.scanMaintenanceWindow(rows)
-		if err != nil {
-			return windows, err
-		}
-		windows = append(windows, mw)
-	}
-	return windows, rows.Err()
-}
-
-func (s *SQLStore) GetAllMaintenanceWindows(ctx context.Context, limit int) ([]models.MaintenanceWindow, error) {
-	rows, err := s.db.QueryContext(ctx, s.q("SELECT id, monitor_id, title, description, type, start_time, end_time, created_by, created_at FROM maintenance_windows ORDER BY created_at DESC LIMIT ?"), limit)
-	if err != nil {
-		return nil, err
-	}
-	defer rows.Close()
-	var windows []models.MaintenanceWindow
-	for rows.Next() {
-		mw, err := s.scanMaintenanceWindow(rows)
-		if err != nil {
-			return windows, err
-		}
-		windows = append(windows, mw)
-	}
-	return windows, rows.Err()
-}
-
-func (s *SQLStore) AddMaintenanceWindow(ctx context.Context, mw models.MaintenanceWindow) error {
-	if mw.StartTime.IsZero() {
-		mw.StartTime = time.Now()
-	}
-	_, err := s.db.ExecContext(ctx, s.q("INSERT INTO maintenance_windows (monitor_id, title, description, type, start_time, end_time, created_by) VALUES (?, ?, ?, ?, ?, ?, ?)"),
-		mw.MonitorID, mw.Title, mw.Description, mw.Type, mw.StartTime, sql.NullTime{Time: mw.EndTime, Valid: !mw.EndTime.IsZero()}, mw.CreatedBy)
-	return err
-}
-
-func (s *SQLStore) EndMaintenanceWindow(ctx context.Context, id int) error {
-	_, err := s.db.ExecContext(ctx, s.q("UPDATE maintenance_windows SET end_time = CURRENT_TIMESTAMP WHERE id = ?"), id)
-	return err
-}
-
-func (s *SQLStore) DeleteMaintenanceWindow(ctx context.Context, id int) error {
-	_, err := s.db.ExecContext(ctx, s.q("DELETE FROM maintenance_windows WHERE id = ?"), id)
-	if err != nil {
-		return err
-	}
-	s.dialect.ResetSequenceOnEmpty(s.db, "maintenance_windows")
-	return nil
-}
-
-func (s *SQLStore) PruneExpiredMaintenanceWindows(ctx context.Context, retention time.Duration) (int64, error) {
-	cutoff := time.Now().Add(-retention)
-	result, err := s.db.ExecContext(ctx,
-		s.q("DELETE FROM maintenance_windows WHERE end_time IS NOT NULL AND end_time < ?"),
-		cutoff,
-	)
-	if err != nil {
-		return 0, err
-	}
-	return result.RowsAffected()
-}
-
-func (s *SQLStore) IsMonitorInMaintenance(ctx context.Context, monitorID int) (bool, error) {
-	var count int
-	err := s.db.QueryRowContext(ctx, s.q(`SELECT COUNT(*) FROM maintenance_windows
-		WHERE type = 'maintenance'
-		AND start_time <= CURRENT_TIMESTAMP
-		AND (end_time IS NULL OR end_time > CURRENT_TIMESTAMP)
-		AND (monitor_id = 0 OR monitor_id = ?
-			OR monitor_id IN (SELECT parent_id FROM sites WHERE id = ? AND parent_id > 0))`),
-		monitorID, monitorID).Scan(&count)
-	if err != nil {
-		return false, err
-	}
-	return count > 0, nil
+	return entries, rows.Err()
 }
 
 func (s *SQLStore) GetPreference(ctx context.Context, key string) (string, error) {

internal/store/sqlstore_alerts.go

@@ -0,0 +1,126 @@
+package store
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+)
+
+func (s *SQLStore) unmarshalSettings(raw string) (map[string]string, error) {
+	decrypted, err := s.decryptSettings(raw)
+	if err != nil {
+		return nil, fmt.Errorf("decrypt settings: %w", err)
+	}
+	var m map[string]string
+	if err := json.Unmarshal([]byte(decrypted), &m); err != nil {
+		return nil, fmt.Errorf("unmarshal settings: %w", err)
+	}
+	return m, nil
+}
+
+func (s *SQLStore) marshalSettings(settings map[string]string) (string, error) {
+	jsonBytes, err := json.Marshal(settings)
+	if err != nil {
+		return "", err
+	}
+	return s.encryptSettings(string(jsonBytes))
+}
+
+func (s *SQLStore) GetAlertByName(ctx context.Context, name string) (models.AlertConfig, error) {
+	var a models.AlertConfig
+	var settingsRaw string
+	err := s.db.QueryRowContext(ctx, s.q("SELECT id, name, type, settings FROM alerts WHERE name = ?"), name).Scan(&a.ID, &a.Name, &a.Type, &settingsRaw)
+	if err != nil {
+		return a, err
+	}
+	a.Settings, err = s.unmarshalSettings(settingsRaw)
+	if err != nil {
+		return a, fmt.Errorf("alert %q: %w", name, err)
+	}
+	return a, nil
+}
+
+func (s *SQLStore) AddAlertReturningID(ctx context.Context, name, aType string, settings map[string]string) (int, error) {
+	stored, err := s.marshalSettings(settings)
+	if err != nil {
+		return 0, err
+	}
+	if s.dollar {
+		var id int
+		err := s.db.QueryRowContext(ctx, s.q("INSERT INTO alerts (name, type, settings) VALUES (?, ?, ?) RETURNING id"), name, aType, stored).Scan(&id)
+		return id, err
+	}
+	result, err := s.db.ExecContext(ctx, s.q("INSERT INTO alerts (name, type, settings) VALUES (?, ?, ?)"), name, aType, stored)
+	if err != nil {
+		return 0, err
+	}
+	id, err := result.LastInsertId()
+	return int(id), err
+}
+
+func (s *SQLStore) GetAllAlerts(ctx context.Context) ([]models.AlertConfig, error) {
+	rows, err := s.db.QueryContext(ctx, "SELECT id, name, type, settings FROM alerts")
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+	var alerts []models.AlertConfig
+	for rows.Next() {
+		var a models.AlertConfig
+		var settingsRaw string
+		if err := rows.Scan(&a.ID, &a.Name, &a.Type, &settingsRaw); err != nil {
+			return alerts, err
+		}
+		a.Settings, err = s.unmarshalSettings(settingsRaw)
+		if err != nil {
+			return alerts, fmt.Errorf("alert %q: %w", a.Name, err)
+		}
+		alerts = append(alerts, a)
+	}
+	return alerts, rows.Err()
+}
+
+func (s *SQLStore) GetAlert(ctx context.Context, id int) (models.AlertConfig, error) {
+	var a models.AlertConfig
+	var settingsRaw string
+	err := s.db.QueryRowContext(ctx, s.q("SELECT id, name, type, settings FROM alerts WHERE id = ?"), id).Scan(&a.ID, &a.Name, &a.Type, &settingsRaw)
+	if err != nil {
+		return a, err
+	}
+	a.Settings, err = s.unmarshalSettings(settingsRaw)
+	if err != nil {
+		return a, fmt.Errorf("alert %d: %w", id, err)
+	}
+	return a, nil
+}
+
+func (s *SQLStore) AddAlert(ctx context.Context, name, aType string, settings map[string]string) error {
+	stored, err := s.marshalSettings(settings)
+	if err != nil {
+		return err
+	}
+	_, err = s.db.ExecContext(ctx, s.q("INSERT INTO alerts (name, type, settings) VALUES (?, ?, ?)"), name, aType, stored)
+	return err
+}
+
+func (s *SQLStore) UpdateAlert(ctx context.Context, id int, name, aType string, settings map[string]string) error {
+	stored, err := s.marshalSettings(settings)
+	if err != nil {
+		return err
+	}
+	_, err = s.db.ExecContext(ctx, s.q("UPDATE alerts SET name=?, type=?, settings=? WHERE id=?"), name, aType, stored, id)
+	return err
+}
+
+func (s *SQLStore) DeleteAlert(ctx context.Context, id int) error {
+	if _, err := s.db.ExecContext(ctx, s.q("UPDATE sites SET alert_id = 0 WHERE alert_id = ?"), id); err != nil {
+		return err
+	}
+	if _, err := s.db.ExecContext(ctx, s.q("DELETE FROM alerts WHERE id=?"), id); err != nil {
+		return err
+	}
+	s.dialect.ResetSequenceOnEmpty(s.db, "alerts")
+	return nil
+}

internal/store/sqlstore_alerts_test.go

@@ -0,0 +1,102 @@
+package store
+
+import (
+	"context"
+	"strings"
+	"testing"
+
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+)
+
+func TestAlertCRUD(t *testing.T) {
+	s := newTestStore(t)
+
+	if err := s.AddAlert(context.Background(), "Discord", "discord", map[string]string{"url": "https://example.com/hook"}); err != nil {
+		t.Fatalf("AddAlert: %v", err)
+	}
+
+	alerts, err := s.GetAllAlerts(context.Background())
+	if err != nil {
+		t.Fatalf("GetAllAlerts: %v", err)
+	}
+	if len(alerts) != 1 {
+		t.Fatalf("expected 1 alert, got %d", len(alerts))
+	}
+	if alerts[0].Type != "discord" {
+		t.Errorf("expected type 'discord', got '%s'", alerts[0].Type)
+	}
+	if alerts[0].Settings["url"] != "https://example.com/hook" {
+		t.Errorf("settings url mismatch")
+	}
+
+	a, err := s.GetAlert(context.Background(), alerts[0].ID)
+	if err != nil {
+		t.Fatalf("GetAlert: %v", err)
+	}
+	if a.Name != "Discord" {
+		t.Errorf("expected name 'Discord', got '%s'", a.Name)
+	}
+
+	if err := s.UpdateAlert(context.Background(), a.ID, "Slack", "slack", map[string]string{"url": "https://slack.com/hook"}); err != nil {
+		t.Fatalf("UpdateAlert: %v", err)
+	}
+
+	a, err = s.GetAlert(context.Background(), a.ID)
+	if err != nil {
+		t.Fatalf("GetAlert: %v", err)
+	}
+	if a.Type != "slack" {
+		t.Errorf("expected type 'slack', got '%s'", a.Type)
+	}
+
+	if err := s.DeleteAlert(context.Background(), a.ID); err != nil {
+		t.Fatalf("DeleteAlert: %v", err)
+	}
+
+	alerts, err = s.GetAllAlerts(context.Background())
+	if err != nil {
+		t.Fatalf("GetAllAlerts: %v", err)
+	}
+	if len(alerts) != 0 {
+		t.Fatalf("expected 0 alerts after delete, got %d", len(alerts))
+	}
+}
+
+// ImportData must encrypt alert settings (like AddAlert/UpdateAlert) so a
+// restore with UPTOP_ENCRYPTION_KEY set never lands secrets in plaintext.
+func TestImportData_EncryptsAlertSettings(t *testing.T) {
+	s := newTestStore(t)
+	enc, err := NewEncryptor(strings.Repeat("ab", 32)) // 64 hex chars = 32 bytes
+	if err != nil {
+		t.Fatalf("NewEncryptor: %v", err)
+	}
+	s.SetEncryptor(enc)
+
+	backup := models.Backup{
+		Alerts: []models.AlertConfig{
+			{ID: 1, Name: "tg", Type: "telegram", Settings: map[string]string{"token": "123:SECRET", "chat_id": "42"}},
+		},
+	}
+	if err := s.ImportData(context.Background(), backup); err != nil {
+		t.Fatalf("ImportData: %v", err)
+	}
+
+	var raw string
+	if err := s.db.QueryRow("SELECT settings FROM alerts WHERE id = 1").Scan(&raw); err != nil {
+		t.Fatalf("query settings: %v", err)
+	}
+	if !strings.HasPrefix(raw, encryptedPrefix) {
+		t.Errorf("imported settings not encrypted: %q", raw)
+	}
+	if strings.Contains(raw, "SECRET") {
+		t.Errorf("plaintext secret found in stored column: %q", raw)
+	}
+
+	alerts, err := s.GetAllAlerts(context.Background())
+	if err != nil {
+		t.Fatalf("GetAllAlerts: %v", err)
+	}
+	if len(alerts) != 1 || alerts[0].Settings["token"] != "123:SECRET" {
+		t.Errorf("decrypt round-trip failed: %+v", alerts)
+	}
+}

internal/store/sqlstore_history.go

@@ -0,0 +1,116 @@
+package store
+
+import (
+	"context"
+	"fmt"
+	"time"
+
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+)
+
+func (s *SQLStore) SaveStateChange(ctx context.Context, siteID int, fromStatus, toStatus, errorReason string) error {
+	_, err := s.db.ExecContext(ctx, s.q("INSERT INTO state_changes (site_id, from_status, to_status, error_reason) VALUES (?, ?, ?, ?)"),
+		siteID, fromStatus, toStatus, errorReason)
+	return err
+}
+
+func (s *SQLStore) GetStateChanges(ctx context.Context, siteID int, limit int) ([]models.StateChange, error) {
+	rows, err := s.db.QueryContext(ctx, s.q("SELECT id, site_id, from_status, to_status, error_reason, changed_at FROM state_changes WHERE site_id = ? ORDER BY changed_at DESC LIMIT ?"), siteID, limit)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+	var changes []models.StateChange
+	for rows.Next() {
+		var sc models.StateChange
+		if err := rows.Scan(&sc.ID, &sc.SiteID, &sc.FromStatus, &sc.ToStatus, &sc.ErrorReason, &sc.ChangedAt); err != nil {
+			return changes, err
+		}
+		changes = append(changes, sc)
+	}
+	return changes, rows.Err()
+}
+
+func (s *SQLStore) GetStateChangesSince(ctx context.Context, siteID int, since time.Time) ([]models.StateChange, error) {
+	rows, err := s.db.QueryContext(ctx, s.q("SELECT id, site_id, from_status, to_status, error_reason, changed_at FROM state_changes WHERE site_id = ? AND changed_at >= ? ORDER BY changed_at DESC"), siteID, since)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+	var changes []models.StateChange
+	for rows.Next() {
+		var sc models.StateChange
+		if err := rows.Scan(&sc.ID, &sc.SiteID, &sc.FromStatus, &sc.ToStatus, &sc.ErrorReason, &sc.ChangedAt); err != nil {
+			return changes, err
+		}
+		changes = append(changes, sc)
+	}
+	return changes, rows.Err()
+}
+
+func (s *SQLStore) SaveCheck(ctx context.Context, siteID int, latencyNs int64, isUp bool) error {
+	return s.SaveCheckFromNode(ctx, siteID, "", latencyNs, isUp)
+}
+
+// SaveCheckFromNode inserts a single check row. Retention is handled out of
+// band by PruneCheckHistory on a timer, not per-insert, to keep the write hot
+// path a plain INSERT.
+func (s *SQLStore) SaveCheckFromNode(ctx context.Context, siteID int, nodeID string, latencyNs int64, isUp bool) error {
+	_, err := s.db.ExecContext(ctx, s.q("INSERT INTO check_history (site_id, node_id, latency_ns, is_up) VALUES (?, ?, ?, ?)"), siteID, nodeID, latencyNs, isUp)
+	return err
+}
+
+// PruneCheckHistory trims check_history to the newest maxCheckHistory rows per
+// site, across all sites, in one pass. Intended to run periodically.
+func (s *SQLStore) PruneCheckHistory(ctx context.Context) error {
+	q := fmt.Sprintf(`DELETE FROM check_history WHERE id IN (
+		SELECT id FROM (
+			SELECT id, ROW_NUMBER() OVER (PARTITION BY site_id ORDER BY checked_at DESC, id DESC) AS rn
+			FROM check_history
+		) ranked WHERE rn > %d
+	)`, maxCheckHistory)
+	_, err := s.db.ExecContext(ctx, s.q(q))
+	return err
+}
+
+// PruneStateChanges trims state_changes to the newest maxStateChangesPerSite
+// rows per site. Generous so realistic SLA windows are unaffected; bounds the
+// otherwise unbounded growth of a flapping monitor's history.
+func (s *SQLStore) PruneStateChanges(ctx context.Context) error {
+	q := fmt.Sprintf(`DELETE FROM state_changes WHERE id IN (
+		SELECT id FROM (
+			SELECT id, ROW_NUMBER() OVER (PARTITION BY site_id ORDER BY changed_at DESC, id DESC) AS rn
+			FROM state_changes
+		) ranked WHERE rn > %d
+	)`, maxStateChangesPerSite)
+	_, err := s.db.ExecContext(ctx, s.q(q))
+	return err
+}
+
+func (s *SQLStore) LoadAllHistory(ctx context.Context, limit int) (map[int][]models.CheckRecord, error) {
+	result := make(map[int][]models.CheckRecord)
+	rows, err := s.db.QueryContext(ctx, s.q(`
+		SELECT site_id, latency_ns, is_up FROM (
+			SELECT site_id, latency_ns, is_up,
+				ROW_NUMBER() OVER (PARTITION BY site_id ORDER BY checked_at DESC) AS rn
+			FROM check_history
+		) sub WHERE rn <= ?`), limit)
+	if err != nil {
+		return result, err
+	}
+	defer rows.Close()
+	for rows.Next() {
+		var r models.CheckRecord
+		if err := rows.Scan(&r.SiteID, &r.LatencyNs, &r.IsUp); err != nil {
+			return result, err
+		}
+		result[r.SiteID] = append(result[r.SiteID], r)
+	}
+	for id, records := range result {
+		for i, j := 0, len(records)-1; i < j; i, j = i+1, j-1 {
+			records[i], records[j] = records[j], records[i]
+		}
+		result[id] = records
+	}
+	return result, rows.Err()
+}

internal/store/sqlstore_history_test.go

@@ -0,0 +1,134 @@
+package store
+
+import (
+	"context"
+	"testing"
+	"time"
+
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+)
+
+func TestCheckHistory(t *testing.T) {
+	s := newTestStore(t)
+
+	if err := s.SaveCheck(context.Background(), 1, 5000000, true); err != nil {
+		t.Fatalf("SaveCheck: %v", err)
+	}
+	if err := s.SaveCheck(context.Background(), 1, 10000000, false); err != nil {
+		t.Fatalf("SaveCheck: %v", err)
+	}
+	if err := s.SaveCheck(context.Background(), 2, 3000000, true); err != nil {
+		t.Fatalf("SaveCheck site 2: %v", err)
+	}
+
+	history, err := s.LoadAllHistory(context.Background(), 10)
+	if err != nil {
+		t.Fatalf("LoadAllHistory: %v", err)
+	}
+	if len(history[1]) != 2 {
+		t.Fatalf("expected 2 records for site 1, got %d", len(history[1]))
+	}
+	if len(history[2]) != 1 {
+		t.Fatalf("expected 1 record for site 2, got %d", len(history[2]))
+	}
+
+	upCount := 0
+	for _, r := range history[1] {
+		if r.IsUp {
+			upCount++
+		}
+	}
+	if upCount != 1 {
+		t.Errorf("expected 1 up record for site 1, got %d", upCount)
+	}
+}
+
+func TestPruneCheckHistory(t *testing.T) {
+	s := newTestStore(t)
+
+	for i := 0; i < maxCheckHistory+5; i++ {
+		if err := s.SaveCheck(context.Background(), 1, int64(i), true); err != nil {
+			t.Fatalf("SaveCheck site 1: %v", err)
+		}
+	}
+	for i := 0; i < 3; i++ {
+		if err := s.SaveCheck(context.Background(), 2, int64(i), true); err != nil {
+			t.Fatalf("SaveCheck site 2: %v", err)
+		}
+	}
+
+	if err := s.PruneCheckHistory(context.Background()); err != nil {
+		t.Fatalf("PruneCheckHistory: %v", err)
+	}
+
+	history, err := s.LoadAllHistory(context.Background(), maxCheckHistory*2)
+	if err != nil {
+		t.Fatalf("LoadAllHistory: %v", err)
+	}
+	if len(history[1]) != maxCheckHistory {
+		t.Errorf("site 1: expected %d rows after prune, got %d", maxCheckHistory, len(history[1]))
+	}
+	if len(history[2]) != 3 {
+		t.Errorf("site 2: expected 3 rows untouched, got %d", len(history[2]))
+	}
+}
+
+func TestDeleteSiteCascade(t *testing.T) {
+	s := newTestStore(t)
+
+	site := models.SiteConfig{Name: "Cascade Test", URL: "https://example.com", Interval: 30}
+	if err := s.AddSite(context.Background(), site); err != nil {
+		t.Fatalf("AddSite: %v", err)
+	}
+	sites, err := s.GetSites(context.Background())
+	if err != nil {
+		t.Fatalf("GetSites: %v", err)
+	}
+	siteID := sites[0].ID
+
+	if err := s.SaveCheck(context.Background(), siteID, 1000, true); err != nil {
+		t.Fatalf("SaveCheck: %v", err)
+	}
+	if err := s.SaveStateChange(context.Background(), siteID, "UP", "DOWN", "timeout"); err != nil {
+		t.Fatalf("SaveStateChange: %v", err)
+	}
+	mw := models.MaintenanceWindow{
+		MonitorID: siteID,
+		Title:     "Test MW",
+		Type:      "maintenance",
+		StartTime: time.Now(),
+	}
+	if err := s.AddMaintenanceWindow(context.Background(), mw); err != nil {
+		t.Fatalf("AddMaintenanceWindow: %v", err)
+	}
+
+	if err := s.DeleteSite(context.Background(), siteID); err != nil {
+		t.Fatalf("DeleteSite: %v", err)
+	}
+
+	history, err := s.LoadAllHistory(context.Background(), 100)
+	if err != nil {
+		t.Fatalf("LoadAllHistory: %v", err)
+	}
+	if len(history[siteID]) != 0 {
+		t.Errorf("expected 0 check_history rows, got %d", len(history[siteID]))
+	}
+
+	changes, err := s.GetStateChanges(context.Background(), siteID, 100)
+	if err != nil {
+		t.Fatalf("GetStateChanges: %v", err)
+	}
+	if len(changes) != 0 {
+		t.Errorf("expected 0 state_changes rows, got %d", len(changes))
+	}
+
+	windows, err := s.GetActiveMaintenanceWindows(context.Background())
+	if err != nil {
+		t.Fatalf("GetActiveMaintenanceWindows: %v", err)
+	}
+	for _, w := range windows {
+		if w.MonitorID == siteID {
+			t.Errorf("orphaned maintenance window found: id=%d", w.ID)
+		}
+	}
+}

internal/store/sqlstore_maintenance.go

@@ -0,0 +1,148 @@
+package store
+
+import (
+	"context"
+	"database/sql"
+	"fmt"
+	"time"
+
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+)
+
+func (s *SQLStore) scanMaintenanceWindow(rows *sql.Rows) (models.MaintenanceWindow, error) {
+	var mw models.MaintenanceWindow
+	var endTime sql.NullTime
+	if err := rows.Scan(&mw.ID, &mw.MonitorID, &mw.Title, &mw.Description, &mw.Type, &mw.StartTime, &endTime, &mw.CreatedBy, &mw.CreatedAt); err != nil {
+		return mw, err
+	}
+	if endTime.Valid {
+		mw.EndTime = endTime.Time
+	}
+	return mw, nil
+}
+
+func (s *SQLStore) GetActiveMaintenanceWindows(ctx context.Context) ([]models.MaintenanceWindow, error) {
+	rows, err := s.db.QueryContext(ctx, s.q("SELECT id, monitor_id, title, description, type, start_time, end_time, created_by, created_at FROM maintenance_windows WHERE start_time <= CURRENT_TIMESTAMP AND (end_time IS NULL OR end_time > CURRENT_TIMESTAMP) ORDER BY start_time DESC"))
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+	var windows []models.MaintenanceWindow
+	for rows.Next() {
+		mw, err := s.scanMaintenanceWindow(rows)
+		if err != nil {
+			return windows, err
+		}
+		windows = append(windows, mw)
+	}
+	return windows, rows.Err()
+}
+
+func (s *SQLStore) GetAllMaintenanceWindows(ctx context.Context, limit int) ([]models.MaintenanceWindow, error) {
+	rows, err := s.db.QueryContext(ctx, s.q("SELECT id, monitor_id, title, description, type, start_time, end_time, created_by, created_at FROM maintenance_windows ORDER BY created_at DESC LIMIT ?"), limit)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+	var windows []models.MaintenanceWindow
+	for rows.Next() {
+		mw, err := s.scanMaintenanceWindow(rows)
+		if err != nil {
+			return windows, err
+		}
+		windows = append(windows, mw)
+	}
+	return windows, rows.Err()
+}
+
+func (s *SQLStore) GetOverlappingMaintenanceWindows(ctx context.Context, monitorID int, startTime, endTime time.Time) ([]models.MaintenanceWindow, error) {
+	var timeClause string
+	var args []interface{}
+
+	if endTime.IsZero() {
+		timeClause = "(end_time IS NULL OR end_time > ?)"
+		args = append(args, startTime)
+	} else {
+		timeClause = "(end_time IS NULL OR end_time > ?) AND start_time < ?"
+		args = append(args, startTime, endTime)
+	}
+
+	var scopeClause string
+	if monitorID == 0 {
+		scopeClause = "1=1"
+	} else {
+		scopeClause = "(monitor_id = ? OR monitor_id = 0 OR monitor_id IN (SELECT parent_id FROM sites WHERE id = ? AND parent_id > 0))"
+		args = append(args, monitorID, monitorID)
+	}
+
+	query := fmt.Sprintf(
+		"SELECT id, monitor_id, title, description, type, start_time, end_time, created_by, created_at FROM maintenance_windows WHERE %s AND %s ORDER BY start_time",
+		timeClause, scopeClause,
+	)
+
+	rows, err := s.db.QueryContext(ctx, s.q(query), args...)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+
+	var windows []models.MaintenanceWindow
+	for rows.Next() {
+		mw, err := s.scanMaintenanceWindow(rows)
+		if err != nil {
+			return nil, err
+		}
+		windows = append(windows, mw)
+	}
+	return windows, rows.Err()
+}
+
+func (s *SQLStore) AddMaintenanceWindow(ctx context.Context, mw models.MaintenanceWindow) error {
+	if mw.StartTime.IsZero() {
+		mw.StartTime = time.Now()
+	}
+	_, err := s.db.ExecContext(ctx, s.q("INSERT INTO maintenance_windows (monitor_id, title, description, type, start_time, end_time, created_by) VALUES (?, ?, ?, ?, ?, ?, ?)"),
+		mw.MonitorID, mw.Title, mw.Description, mw.Type, mw.StartTime, sql.NullTime{Time: mw.EndTime, Valid: !mw.EndTime.IsZero()}, mw.CreatedBy)
+	return err
+}
+
+func (s *SQLStore) EndMaintenanceWindow(ctx context.Context, id int) error {
+	_, err := s.db.ExecContext(ctx, s.q("UPDATE maintenance_windows SET end_time = CURRENT_TIMESTAMP WHERE id = ?"), id)
+	return err
+}
+
+func (s *SQLStore) DeleteMaintenanceWindow(ctx context.Context, id int) error {
+	_, err := s.db.ExecContext(ctx, s.q("DELETE FROM maintenance_windows WHERE id = ?"), id)
+	if err != nil {
+		return err
+	}
+	s.dialect.ResetSequenceOnEmpty(s.db, "maintenance_windows")
+	return nil
+}
+
+func (s *SQLStore) PruneExpiredMaintenanceWindows(ctx context.Context, retention time.Duration) (int64, error) {
+	cutoff := time.Now().Add(-retention)
+	result, err := s.db.ExecContext(ctx,
+		s.q("DELETE FROM maintenance_windows WHERE end_time IS NOT NULL AND end_time < ?"),
+		cutoff,
+	)
+	if err != nil {
+		return 0, err
+	}
+	return result.RowsAffected()
+}
+
+func (s *SQLStore) IsMonitorInMaintenance(ctx context.Context, monitorID int) (bool, error) {
+	var count int
+	err := s.db.QueryRowContext(ctx, s.q(`SELECT COUNT(*) FROM maintenance_windows
+		WHERE type = 'maintenance'
+		AND start_time <= CURRENT_TIMESTAMP
+		AND (end_time IS NULL OR end_time > CURRENT_TIMESTAMP)
+		AND (monitor_id = 0 OR monitor_id = ?
+			OR monitor_id IN (SELECT parent_id FROM sites WHERE id = ? AND parent_id > 0))`),
+		monitorID, monitorID).Scan(&count)
+	if err != nil {
+		return false, err
+	}
+	return count > 0, nil
+}

internal/store/sqlstore_maintenance_test.go

@@ -0,0 +1,168 @@
+package store
+
+import (
+	"context"
+	"testing"
+	"time"
+
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+)
+
+func TestPruneExpiredMaintenanceWindows(t *testing.T) {
+	s := newTestStore(t)
+
+	now := time.Now()
+
+	// Expired 10 days ago — should be pruned with 7d retention.
+	old := models.MaintenanceWindow{
+		MonitorID: 0,
+		Title:     "Old Window",
+		Type:      "maintenance",
+		StartTime: now.Add(-11 * 24 * time.Hour),
+		EndTime:   now.Add(-10 * 24 * time.Hour),
+	}
+	if err := s.AddMaintenanceWindow(context.Background(), old); err != nil {
+		t.Fatalf("AddMaintenanceWindow (old): %v", err)
+	}
+
+	// Expired 1 day ago — within 7d retention, should survive.
+	recent := models.MaintenanceWindow{
+		MonitorID: 0,
+		Title:     "Recent Window",
+		Type:      "maintenance",
+		StartTime: now.Add(-2 * 24 * time.Hour),
+		EndTime:   now.Add(-1 * 24 * time.Hour),
+	}
+	if err := s.AddMaintenanceWindow(context.Background(), recent); err != nil {
+		t.Fatalf("AddMaintenanceWindow (recent): %v", err)
+	}
+
+	// Ongoing — no end time, should survive.
+	ongoing := models.MaintenanceWindow{
+		MonitorID: 0,
+		Title:     "Ongoing Window",
+		Type:      "maintenance",
+		StartTime: now.Add(-1 * time.Hour),
+	}
+	if err := s.AddMaintenanceWindow(context.Background(), ongoing); err != nil {
+		t.Fatalf("AddMaintenanceWindow (ongoing): %v", err)
+	}
+
+	pruned, err := s.PruneExpiredMaintenanceWindows(context.Background(), 7*24*time.Hour)
+	if err != nil {
+		t.Fatalf("PruneExpiredMaintenanceWindows: %v", err)
+	}
+	if pruned != 1 {
+		t.Errorf("expected 1 pruned, got %d", pruned)
+	}
+
+	all, err := s.GetAllMaintenanceWindows(context.Background(), 100)
+	if err != nil {
+		t.Fatalf("GetAllMaintenanceWindows: %v", err)
+	}
+	if len(all) != 2 {
+		t.Fatalf("expected 2 remaining windows, got %d", len(all))
+	}
+	for _, w := range all {
+		if w.Title == "Old Window" {
+			t.Error("old window should have been pruned")
+		}
+	}
+}
+
+func TestGetOverlappingMaintenanceWindows(t *testing.T) {
+	s := newTestStore(t)
+	ctx := context.Background()
+
+	site := models.SiteConfig{Name: "web", URL: "https://example.com", Interval: 30}
+	if err := s.AddSite(ctx, site); err != nil {
+		t.Fatalf("AddSite: %v", err)
+	}
+
+	now := time.Now()
+
+	active := models.MaintenanceWindow{
+		MonitorID: 1,
+		Title:     "Deploy v2",
+		Type:      "maintenance",
+		StartTime: now.Add(-30 * time.Minute),
+		EndTime:   now.Add(30 * time.Minute),
+	}
+	if err := s.AddMaintenanceWindow(ctx, active); err != nil {
+		t.Fatalf("AddMaintenanceWindow: %v", err)
+	}
+
+	ended := models.MaintenanceWindow{
+		MonitorID: 1,
+		Title:     "Old deploy",
+		Type:      "maintenance",
+		StartTime: now.Add(-3 * time.Hour),
+		EndTime:   now.Add(-2 * time.Hour),
+	}
+	if err := s.AddMaintenanceWindow(ctx, ended); err != nil {
+		t.Fatalf("AddMaintenanceWindow: %v", err)
+	}
+
+	t.Run("same monitor overlaps", func(t *testing.T) {
+		overlaps, err := s.GetOverlappingMaintenanceWindows(ctx, 1, now, now.Add(1*time.Hour))
+		if err != nil {
+			t.Fatalf("unexpected error: %v", err)
+		}
+		if len(overlaps) != 1 {
+			t.Fatalf("expected 1 overlap, got %d", len(overlaps))
+		}
+		if overlaps[0].Title != "Deploy v2" {
+			t.Errorf("expected 'Deploy v2', got %q", overlaps[0].Title)
+		}
+	})
+
+	t.Run("different monitor no overlap", func(t *testing.T) {
+		overlaps, err := s.GetOverlappingMaintenanceWindows(ctx, 99, now, now.Add(1*time.Hour))
+		if err != nil {
+			t.Fatalf("unexpected error: %v", err)
+		}
+		if len(overlaps) != 0 {
+			t.Errorf("expected 0 overlaps, got %d", len(overlaps))
+		}
+	})
+
+	t.Run("global window overlaps all", func(t *testing.T) {
+		global := models.MaintenanceWindow{
+			MonitorID: 0,
+			Title:     "Global freeze",
+			Type:      "maintenance",
+			StartTime: now.Add(-10 * time.Minute),
+			EndTime:   now.Add(2 * time.Hour),
+		}
+		if err := s.AddMaintenanceWindow(ctx, global); err != nil {
+			t.Fatalf("AddMaintenanceWindow: %v", err)
+		}
+		overlaps, err := s.GetOverlappingMaintenanceWindows(ctx, 1, now, now.Add(1*time.Hour))
+		if err != nil {
+			t.Fatalf("unexpected error: %v", err)
+		}
+		if len(overlaps) != 2 {
+			t.Errorf("expected 2 overlaps (specific + global), got %d", len(overlaps))
+		}
+	})
+
+	t.Run("indefinite window overlaps", func(t *testing.T) {
+		overlaps, err := s.GetOverlappingMaintenanceWindows(ctx, 1, now, time.Time{})
+		if err != nil {
+			t.Fatalf("unexpected error: %v", err)
+		}
+		if len(overlaps) < 1 {
+			t.Error("expected at least 1 overlap for indefinite window")
+		}
+	})
+
+	t.Run("ended window excluded", func(t *testing.T) {
+		overlaps, err := s.GetOverlappingMaintenanceWindows(ctx, 1, now.Add(-4*time.Hour), now.Add(-3*time.Hour))
+		if err != nil {
+			t.Fatalf("unexpected error: %v", err)
+		}
+		if len(overlaps) != 0 {
+			t.Errorf("expected 0 overlaps for past range, got %d", len(overlaps))
+		}
+	})
+}

internal/store/sqlstore_test.go

@@ -3,9 +3,7 @@ package store
 import (
 	"context"
 	"fmt"
-	"strings"
 	"testing"
-	"time"
 
 	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
 )
@@ -74,60 +72,6 @@ func TestSiteCRUD(t *testing.T) {
 	}
 }
 
-func TestAlertCRUD(t *testing.T) {
-	s := newTestStore(t)
-
-	if err := s.AddAlert(context.Background(), "Discord", "discord", map[string]string{"url": "https://example.com/hook"}); err != nil {
-		t.Fatalf("AddAlert: %v", err)
-	}
-
-	alerts, err := s.GetAllAlerts(context.Background())
-	if err != nil {
-		t.Fatalf("GetAllAlerts: %v", err)
-	}
-	if len(alerts) != 1 {
-		t.Fatalf("expected 1 alert, got %d", len(alerts))
-	}
-	if alerts[0].Type != "discord" {
-		t.Errorf("expected type 'discord', got '%s'", alerts[0].Type)
-	}
-	if alerts[0].Settings["url"] != "https://example.com/hook" {
-		t.Errorf("settings url mismatch")
-	}
-
-	a, err := s.GetAlert(context.Background(), alerts[0].ID)
-	if err != nil {
-		t.Fatalf("GetAlert: %v", err)
-	}
-	if a.Name != "Discord" {
-		t.Errorf("expected name 'Discord', got '%s'", a.Name)
-	}
-
-	if err := s.UpdateAlert(context.Background(), a.ID, "Slack", "slack", map[string]string{"url": "https://slack.com/hook"}); err != nil {
-		t.Fatalf("UpdateAlert: %v", err)
-	}
-
-	a, err = s.GetAlert(context.Background(), a.ID)
-	if err != nil {
-		t.Fatalf("GetAlert: %v", err)
-	}
-	if a.Type != "slack" {
-		t.Errorf("expected type 'slack', got '%s'", a.Type)
-	}
-
-	if err := s.DeleteAlert(context.Background(), a.ID); err != nil {
-		t.Fatalf("DeleteAlert: %v", err)
-	}
-
-	alerts, err = s.GetAllAlerts(context.Background())
-	if err != nil {
-		t.Fatalf("GetAllAlerts: %v", err)
-	}
-	if len(alerts) != 0 {
-		t.Fatalf("expected 0 alerts after delete, got %d", len(alerts))
-	}
-}
-
 func TestUserCRUD(t *testing.T) {
 	s := newTestStore(t)
 
@@ -301,89 +245,6 @@ func TestImportData_NilUsersPreservesExisting(t *testing.T) {
 	}
 }
 
-func TestCheckHistory(t *testing.T) {
-	s := newTestStore(t)
-
-	if err := s.SaveCheck(context.Background(), 1, 5000000, true); err != nil {
-		t.Fatalf("SaveCheck: %v", err)
-	}
-	if err := s.SaveCheck(context.Background(), 1, 10000000, false); err != nil {
-		t.Fatalf("SaveCheck: %v", err)
-	}
-	if err := s.SaveCheck(context.Background(), 2, 3000000, true); err != nil {
-		t.Fatalf("SaveCheck site 2: %v", err)
-	}
-
-	history, err := s.LoadAllHistory(context.Background(), 10)
-	if err != nil {
-		t.Fatalf("LoadAllHistory: %v", err)
-	}
-	if len(history[1]) != 2 {
-		t.Fatalf("expected 2 records for site 1, got %d", len(history[1]))
-	}
-	if len(history[2]) != 1 {
-		t.Fatalf("expected 1 record for site 2, got %d", len(history[2]))
-	}
-
-	upCount := 0
-	for _, r := range history[1] {
-		if r.IsUp {
-			upCount++
-		}
-	}
-	if upCount != 1 {
-		t.Errorf("expected 1 up record for site 1, got %d", upCount)
-	}
-}
-
-func TestDeleteSiteCascade(t *testing.T) {
-	s := newTestStore(t)
-
-	site := models.SiteConfig{Name: "Cascade Test", URL: "https://example.com", Interval: 30}
-	if err := s.AddSite(context.Background(), site); err != nil {
-		t.Fatalf("AddSite: %v", err)
-	}
-	sites, _ := s.GetSites(context.Background())
-	siteID := sites[0].ID
-
-	if err := s.SaveCheck(context.Background(), siteID, 1000, true); err != nil {
-		t.Fatalf("SaveCheck: %v", err)
-	}
-	if err := s.SaveStateChange(context.Background(), siteID, "UP", "DOWN", "timeout"); err != nil {
-		t.Fatalf("SaveStateChange: %v", err)
-	}
-	mw := models.MaintenanceWindow{
-		MonitorID: siteID,
-		Title:     "Test MW",
-		Type:      "maintenance",
-		StartTime: time.Now(),
-	}
-	if err := s.AddMaintenanceWindow(context.Background(), mw); err != nil {
-		t.Fatalf("AddMaintenanceWindow: %v", err)
-	}
-
-	if err := s.DeleteSite(context.Background(), siteID); err != nil {
-		t.Fatalf("DeleteSite: %v", err)
-	}
-
-	history, _ := s.LoadAllHistory(context.Background(), 100)
-	if len(history[siteID]) != 0 {
-		t.Errorf("expected 0 check_history rows, got %d", len(history[siteID]))
-	}
-
-	changes, _ := s.GetStateChanges(context.Background(), siteID, 100)
-	if len(changes) != 0 {
-		t.Errorf("expected 0 state_changes rows, got %d", len(changes))
-	}
-
-	windows, _ := s.GetActiveMaintenanceWindows(context.Background())
-	for _, w := range windows {
-		if w.MonitorID == siteID {
-			t.Errorf("orphaned maintenance window found: id=%d", w.ID)
-		}
-	}
-}
-
 func TestPruneLogs(t *testing.T) {
 	s := newTestStore(t)
 
@@ -407,7 +268,7 @@ func TestPruneLogs(t *testing.T) {
 	// LoadLogs ordering ties when rows share a created_at second).
 	present := make(map[string]bool, len(logs))
 	for _, l := range logs {
-		present[l] = true
+		present[l.Message] = true
 	}
 	if !present[fmt.Sprintf("log %d", maxLogRows+50-1)] {
 		t.Error("newest log was pruned")
@@ -416,134 +277,3 @@ func TestPruneLogs(t *testing.T) {
 		t.Error("oldest log survived prune")
 	}
 }
-
-func TestPruneCheckHistory(t *testing.T) {
-	s := newTestStore(t)
-
-	for i := 0; i < maxCheckHistory+5; i++ {
-		if err := s.SaveCheck(context.Background(), 1, int64(i), true); err != nil {
-			t.Fatalf("SaveCheck site 1: %v", err)
-		}
-	}
-	for i := 0; i < 3; i++ {
-		if err := s.SaveCheck(context.Background(), 2, int64(i), true); err != nil {
-			t.Fatalf("SaveCheck site 2: %v", err)
-		}
-	}
-
-	if err := s.PruneCheckHistory(context.Background()); err != nil {
-		t.Fatalf("PruneCheckHistory: %v", err)
-	}
-
-	history, err := s.LoadAllHistory(context.Background(), maxCheckHistory*2)
-	if err != nil {
-		t.Fatalf("LoadAllHistory: %v", err)
-	}
-	if len(history[1]) != maxCheckHistory {
-		t.Errorf("site 1: expected %d rows after prune, got %d", maxCheckHistory, len(history[1]))
-	}
-	if len(history[2]) != 3 {
-		t.Errorf("site 2: expected 3 rows untouched, got %d", len(history[2]))
-	}
-}
-
-func TestPruneExpiredMaintenanceWindows(t *testing.T) {
-	s := newTestStore(t)
-
-	now := time.Now()
-
-	// Expired 10 days ago — should be pruned with 7d retention.
-	old := models.MaintenanceWindow{
-		MonitorID: 0,
-		Title:     "Old Window",
-		Type:      "maintenance",
-		StartTime: now.Add(-11 * 24 * time.Hour),
-		EndTime:   now.Add(-10 * 24 * time.Hour),
-	}
-	if err := s.AddMaintenanceWindow(context.Background(), old); err != nil {
-		t.Fatalf("AddMaintenanceWindow (old): %v", err)
-	}
-
-	// Expired 1 day ago — within 7d retention, should survive.
-	recent := models.MaintenanceWindow{
-		MonitorID: 0,
-		Title:     "Recent Window",
-		Type:      "maintenance",
-		StartTime: now.Add(-2 * 24 * time.Hour),
-		EndTime:   now.Add(-1 * 24 * time.Hour),
-	}
-	if err := s.AddMaintenanceWindow(context.Background(), recent); err != nil {
-		t.Fatalf("AddMaintenanceWindow (recent): %v", err)
-	}
-
-	// Ongoing — no end time, should survive.
-	ongoing := models.MaintenanceWindow{
-		MonitorID: 0,
-		Title:     "Ongoing Window",
-		Type:      "maintenance",
-		StartTime: now.Add(-1 * time.Hour),
-	}
-	if err := s.AddMaintenanceWindow(context.Background(), ongoing); err != nil {
-		t.Fatalf("AddMaintenanceWindow (ongoing): %v", err)
-	}
-
-	pruned, err := s.PruneExpiredMaintenanceWindows(context.Background(), 7*24*time.Hour)
-	if err != nil {
-		t.Fatalf("PruneExpiredMaintenanceWindows: %v", err)
-	}
-	if pruned != 1 {
-		t.Errorf("expected 1 pruned, got %d", pruned)
-	}
-
-	all, err := s.GetAllMaintenanceWindows(context.Background(), 100)
-	if err != nil {
-		t.Fatalf("GetAllMaintenanceWindows: %v", err)
-	}
-	if len(all) != 2 {
-		t.Fatalf("expected 2 remaining windows, got %d", len(all))
-	}
-	for _, w := range all {
-		if w.Title == "Old Window" {
-			t.Error("old window should have been pruned")
-		}
-	}
-}
-
-// ImportData must encrypt alert settings (like AddAlert/UpdateAlert) so a
-// restore with UPTOP_ENCRYPTION_KEY set never lands secrets in plaintext.
-func TestImportData_EncryptsAlertSettings(t *testing.T) {
-	s := newTestStore(t)
-	enc, err := NewEncryptor(strings.Repeat("ab", 32)) // 64 hex chars = 32 bytes
-	if err != nil {
-		t.Fatalf("NewEncryptor: %v", err)
-	}
-	s.SetEncryptor(enc)
-
-	backup := models.Backup{
-		Alerts: []models.AlertConfig{
-			{ID: 1, Name: "tg", Type: "telegram", Settings: map[string]string{"token": "123:SECRET", "chat_id": "42"}},
-		},
-	}
-	if err := s.ImportData(context.Background(), backup); err != nil {
-		t.Fatalf("ImportData: %v", err)
-	}
-
-	var raw string
-	if err := s.db.QueryRow("SELECT settings FROM alerts WHERE id = 1").Scan(&raw); err != nil {
-		t.Fatalf("query settings: %v", err)
-	}
-	if !strings.HasPrefix(raw, encryptedPrefix) {
-		t.Errorf("imported settings not encrypted: %q", raw)
-	}
-	if strings.Contains(raw, "SECRET") {
-		t.Errorf("plaintext secret found in stored column: %q", raw)
-	}
-
-	alerts, err := s.GetAllAlerts(context.Background())
-	if err != nil {
-		t.Fatalf("GetAllAlerts: %v", err)
-	}
-	if len(alerts) != 1 || alerts[0].Settings["token"] != "123:SECRET" {
-		t.Errorf("decrypt round-trip failed: %+v", alerts)
-	}
-}

internal/store/store.go

@@ -61,12 +61,13 @@ type Store interface {
 
 	// Logs
 	SaveLog(ctx context.Context, message string) error
-	LoadLogs(ctx context.Context, limit int) ([]string, error)
+	LoadLogs(ctx context.Context, limit int) ([]models.LogEntry, error)
 	PruneLogs(ctx context.Context) error
 
 	// Maintenance Windows
 	GetActiveMaintenanceWindows(ctx context.Context) ([]models.MaintenanceWindow, error)
 	GetAllMaintenanceWindows(ctx context.Context, limit int) ([]models.MaintenanceWindow, error)
+	GetOverlappingMaintenanceWindows(ctx context.Context, monitorID int, startTime, endTime time.Time) ([]models.MaintenanceWindow, error)
 	AddMaintenanceWindow(ctx context.Context, mw models.MaintenanceWindow) error
 	EndMaintenanceWindow(ctx context.Context, id int) error
 	DeleteMaintenanceWindow(ctx context.Context, id int) error

internal/store/storetest/mock.go

@@ -212,8 +212,8 @@ func (m *BaseMock) SaveLog(ctx context.Context, message string) error {
 	return nil
 }
 
-func (m *BaseMock) LoadLogs(_ context.Context, _ int) ([]string, error) { return nil, nil }
-func (m *BaseMock) PruneLogs(_ context.Context) error                   { return nil }
+func (m *BaseMock) LoadLogs(_ context.Context, _ int) ([]models.LogEntry, error) { return nil, nil }
+func (m *BaseMock) PruneLogs(_ context.Context) error                            { return nil }
 
 func (m *BaseMock) GetActiveMaintenanceWindows(ctx context.Context) ([]models.MaintenanceWindow, error) {
 	if m.GetActiveMaintenanceWindowsFunc != nil {
@@ -229,6 +229,10 @@ func (m *BaseMock) GetAllMaintenanceWindows(ctx context.Context, limit int) ([]m
 	return nil, nil
 }
 
+func (m *BaseMock) GetOverlappingMaintenanceWindows(_ context.Context, _ int, _, _ time.Time) ([]models.MaintenanceWindow, error) {
+	return nil, nil
+}
+
 func (m *BaseMock) AddMaintenanceWindow(_ context.Context, _ models.MaintenanceWindow) error {
 	return nil
 }

internal/tui/chart.go

@@ -0,0 +1,132 @@
+package tui
+
+import (
+	"fmt"
+	"strings"
+	"time"
+)
+
+func (m Model) latencyChart(latencies []time.Duration, statuses []bool, width, height int) string {
+	if len(latencies) == 0 || width < 20 || height < 1 {
+		return ""
+	}
+
+	var minMs, maxMs, sumMs int64
+	minMs = latencies[0].Milliseconds()
+	maxMs = minMs
+	upCount := 0
+	for i, l := range latencies {
+		ms := l.Milliseconds()
+		if i < len(statuses) && !statuses[i] {
+			continue
+		}
+		if ms < minMs {
+			minMs = ms
+		}
+		if ms > maxMs {
+			maxMs = ms
+		}
+		sumMs += ms
+		upCount++
+	}
+	var avgMs int64
+	if upCount > 0 {
+		avgMs = sumMs / int64(upCount)
+	}
+
+	maxLabel := fmt.Sprintf("%dms", maxMs)
+	minLabel := fmt.Sprintf("%dms", minMs)
+	labelW := len(maxLabel)
+	if len(minLabel) > labelW {
+		labelW = len(minLabel)
+	}
+	labelW += 1
+
+	chartW := width - labelW
+	if chartW > len(latencies) {
+		chartW = len(latencies)
+	}
+	if chartW < 10 {
+		chartW = 10
+	}
+
+	samples := latencies
+	sampledStatuses := statuses
+	if len(samples) > chartW {
+		samples = samples[len(samples)-chartW:]
+		if len(sampledStatuses) > chartW {
+			sampledStatuses = sampledStatuses[len(sampledStatuses)-chartW:]
+		}
+	}
+
+	spread := maxMs - minMs
+	if spread == 0 {
+		spread = 1
+	}
+	totalLevels := height * 8
+
+	levels := make([]int, len(samples))
+	for i, l := range samples {
+		ms := l.Milliseconds()
+		if i < len(sampledStatuses) && !sampledStatuses[i] {
+			levels[i] = 0
+			continue
+		}
+		lvl := int(float64(ms-minMs) / float64(spread) * float64(totalLevels-1))
+		if lvl < 1 && ms > 0 {
+			lvl = 1
+		}
+		if lvl >= totalLevels {
+			lvl = totalLevels - 1
+		}
+		levels[i] = lvl
+	}
+
+	labelStyle := m.st.subtleStyle
+
+	var rows []string
+	for row := height - 1; row >= 0; row-- {
+		var label string
+		switch row {
+		case height - 1:
+			label = fmt.Sprintf("%*s", labelW, maxLabel)
+		case 0:
+			label = fmt.Sprintf("%*s", labelW, minLabel)
+		default:
+			label = strings.Repeat(" ", labelW)
+		}
+
+		var sb strings.Builder
+		sb.WriteString(labelStyle.Render(label))
+
+		rowBase := row * 8
+		for i := range samples {
+			fill := levels[i] - rowBase
+			if fill <= 0 {
+				sb.WriteString(" ")
+				continue
+			}
+			if fill > 8 {
+				fill = 8
+			}
+			ch := string(sparkChars[fill-1])
+
+			isDown := i < len(sampledStatuses) && !sampledStatuses[i]
+			if isDown {
+				sb.WriteString(m.st.dangerStyle.Render(ch))
+			} else {
+				sb.WriteString(m.latencyStyle(samples[i].Milliseconds(), nil).Render(ch))
+			}
+		}
+		rows = append(rows, sb.String())
+	}
+
+	stats := fmt.Sprintf("%*s Min %s  Avg %s  Max %s",
+		labelW, "",
+		m.fmtLatency(time.Duration(minMs)*time.Millisecond),
+		m.fmtLatency(time.Duration(avgMs)*time.Millisecond),
+		m.fmtLatency(time.Duration(maxMs)*time.Millisecond))
+	rows = append(rows, stats)
+
+	return strings.Join(rows, "\n")
+}

internal/tui/const.go

@@ -0,0 +1,20 @@
+package tui
+
+import "time"
+
+const (
+	nodeOnlineThreshold = 60 * time.Second
+	nodeStaleThreshold  = 5 * time.Minute
+
+	uptimeGoodPct      = 99.0
+	uptimeExcellentPct = 99.9
+	uptimeWarnPct      = 95.0
+	uptimePrecisionPct = 99.99
+
+	stateHistoryLookback = 30 * 24 * time.Hour
+	stateHistoryDays     = 30
+	stateHistoryLimit    = 100
+
+	httpErrorThreshold = 400
+	errorDetailMaxLen  = 30
+)

internal/tui/data.go

@@ -8,13 +8,14 @@ import (
 	"time"
 
 	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/monitor"
 	"gitea.lerkolabs.com/lerkolabs/uptop/internal/store"
 	tea "github.com/charmbracelet/bubbletea"
 )
 
-func loadCollapsed(s store.Store) map[int]bool {
+func loadCollapsed(ctx context.Context, s store.Store) map[int]bool {
 	m := make(map[int]bool)
-	raw, err := s.GetPreference(context.Background(), "collapsed_groups")
+	raw, err := s.GetPreference(ctx, "collapsed_groups")
 	if err != nil || raw == "" {
 		return m
 	}
@@ -49,7 +50,7 @@ func writeCmd(op string, fn func() error) tea.Cmd {
 	}
 }
 
-func sortSitesForDisplay(allSites []models.Site, collapsed map[int]bool) []models.Site {
+func sortSitesForDisplay(allSites []models.Site, collapsed map[int]bool, sortCol int, sortAsc bool) []models.Site {
 	var groups, ungrouped []models.Site
 	children := make(map[int][]models.Site)
 	for _, s := range allSites {
@@ -62,14 +63,32 @@ func sortSitesForDisplay(allSites []models.Site, collapsed map[int]bool) []model
 		}
 	}
 	sort.Slice(groups, func(i, j int) bool { return groups[i].ID < groups[j].ID })
+
+	sortSlice := func(s []models.Site) {
+		sort.Slice(s, func(i, j int) bool { return s[i].ID < s[j].ID })
+		sort.SliceStable(s, func(i, j int) bool {
+			var less bool
+			switch sortCol {
+			case sortName:
+				less = strings.ToLower(s[i].Name) < strings.ToLower(s[j].Name)
+			case sortLatency:
+				less = s[i].Latency < s[j].Latency
+			default:
+				less = siteOrder(s[i]) < siteOrder(s[j])
+			}
+			if !sortAsc {
+				return less
+			}
+			return !less
+		})
+	}
+
 	for pid := range children {
 		c := children[pid]
-		sort.Slice(c, func(i, j int) bool { return c[i].ID < c[j].ID })
-		sort.SliceStable(c, func(i, j int) bool { return siteOrder(c[i]) < siteOrder(c[j]) })
+		sortSlice(c)
 		children[pid] = c
 	}
-	sort.Slice(ungrouped, func(i, j int) bool { return ungrouped[i].ID < ungrouped[j].ID })
-	sort.SliceStable(ungrouped, func(i, j int) bool { return siteOrder(ungrouped[i]) < siteOrder(ungrouped[j]) })
+	sortSlice(ungrouped)
 
 	var ordered []models.Site
 	for _, g := range groups {
@@ -99,14 +118,14 @@ func filterSites(sites []models.Site, needle string) []models.Site {
 // separately via loadTabDataCmd.
 func (m *Model) refreshLive() {
 	allSites := m.engine.GetAllSites()
-	ordered := sortSitesForDisplay(allSites, m.collapsed)
+	ordered := sortSitesForDisplay(allSites, m.collapsed, m.sortColumn, m.sortAsc)
 	if m.filterText != "" {
 		ordered = filterSites(ordered, m.filterText)
 	}
 	m.sites = ordered
 	m.refreshLogContent()
 
-	if m.currentTab == 0 && m.selectedID != 0 {
+	if m.currentTab == tabMonitors && m.selectedID != 0 {
 		for i, s := range m.sites {
 			if s.ID == m.selectedID {
 				m.cursor = i
@@ -118,7 +137,7 @@ func (m *Model) refreshLive() {
 }
 
 func (m *Model) syncSelectedID() {
-	if m.currentTab == 0 && m.cursor < len(m.sites) {
+	if m.currentTab == tabMonitors && m.cursor < len(m.sites) {
 		m.selectedID = m.sites[m.cursor].ID
 	}
 }
@@ -144,9 +163,9 @@ func (m *Model) loadTabDataCmd() tea.Cmd {
 	m.tabSeq++
 	seq := m.tabSeq
 	st := m.store
+	ctx := m.ctx
 	isAdmin := m.isAdmin
 	return func() tea.Msg {
-		ctx := context.Background()
 		alerts, err := st.GetAllAlerts(ctx)
 		if err != nil {
 			return tabDataMsg{seq: seq, err: err}
@@ -173,8 +192,20 @@ func (m *Model) loadTabDataCmd() tea.Cmd {
 // goroutine. View renders the cached result rather than querying the DB.
 func (m *Model) loadDetailCmd(siteID int) tea.Cmd {
 	eng := m.engine
+	ctx := m.ctx
+	var currentStatus models.Status
+	for _, s := range m.sites {
+		if s.ID == siteID {
+			currentStatus = s.Status
+			break
+		}
+	}
 	return func() tea.Msg {
-		return detailDataMsg{siteID: siteID, changes: eng.GetStateChanges(siteID, 5)}
+		changes := eng.GetStateChanges(ctx, siteID, 5)
+		now := time.Now()
+		allChanges := eng.GetStateChangesSince(ctx, siteID, now.Add(-stateHistoryLookback))
+		daily := monitor.ComputeDailyBreakdown(allChanges, currentStatus, stateHistoryDays, now)
+		return detailDataMsg{siteID: siteID, changes: changes, dailyDays: daily}
 	}
 }
 
@@ -182,8 +213,9 @@ func (m *Model) loadDetailCmd(siteID int) tea.Cmd {
 // the UI goroutine.
 func (m *Model) loadHistoryCmd(siteID int) tea.Cmd {
 	eng := m.engine
+	ctx := m.ctx
 	return func() tea.Msg {
-		return historyDataMsg{siteID: siteID, changes: eng.GetStateChanges(siteID, 100)}
+		return historyDataMsg{siteID: siteID, changes: eng.GetStateChanges(ctx, siteID, stateHistoryLimit)}
 	}
 }
 
@@ -192,12 +224,13 @@ func (m *Model) loadHistoryCmd(siteID int) tea.Cmd {
 // can be recognized and dropped.
 func (m *Model) loadSLACmd(siteID, periodIdx int) tea.Cmd {
 	eng := m.engine
+	ctx := m.ctx
 	since := time.Now().Add(-slaPeriods[periodIdx].duration)
 	return func() tea.Msg {
 		return slaDataMsg{
 			siteID:    siteID,
 			periodIdx: periodIdx,
-			changes:   eng.GetStateChangesSince(siteID, since),
+			changes:   eng.GetStateChangesSince(ctx, siteID, since),
 		}
 	}
 }

internal/tui/data_test.go

@@ -12,7 +12,7 @@ func TestSortSitesForDisplay_GroupsFirst(t *testing.T) {
 		{SiteConfig: models.SiteConfig{ID: 1, Name: "group-a", Type: "group"}, SiteState: models.SiteState{Status: "UP"}},
 		{SiteConfig: models.SiteConfig{ID: 2, Name: "child", Type: "http", ParentID: 1}, SiteState: models.SiteState{Status: "UP"}},
 	}
-	result := sortSitesForDisplay(sites, nil)
+	result := sortSitesForDisplay(sites, nil, sortStatus, false)
 	if len(result) != 3 {
 		t.Fatalf("expected 3 sites, got %d", len(result))
 	}
@@ -34,7 +34,7 @@ func TestSortSitesForDisplay_CollapsedHidesChildren(t *testing.T) {
 		{SiteConfig: models.SiteConfig{ID: 3, Name: "child-2", Type: "http", ParentID: 1}, SiteState: models.SiteState{Status: "UP"}},
 	}
 	collapsed := map[int]bool{1: true}
-	result := sortSitesForDisplay(sites, collapsed)
+	result := sortSitesForDisplay(sites, collapsed, sortStatus, false)
 	if len(result) != 1 {
 		t.Fatalf("collapsed group should hide children, got %d items", len(result))
 	}
@@ -49,7 +49,7 @@ func TestSortSitesForDisplay_StatusOrdering(t *testing.T) {
 		{SiteConfig: models.SiteConfig{ID: 2, Name: "down-site", Type: "http"}, SiteState: models.SiteState{Status: "DOWN"}},
 		{SiteConfig: models.SiteConfig{ID: 3, Name: "late-site", Type: "http"}, SiteState: models.SiteState{Status: "LATE"}},
 	}
-	result := sortSitesForDisplay(sites, nil)
+	result := sortSitesForDisplay(sites, nil, sortStatus, false)
 	if result[0].Status != "DOWN" {
 		t.Errorf("DOWN should sort first, got %s", result[0].Status)
 	}

internal/tui/errclass.go

@@ -58,7 +58,7 @@ func classifyError(errorReason string, siteType string, statusCode int) ErrorCat
 	if strings.HasPrefix(lower, "http ") || strings.Contains(lower, "keyword not found") {
 		return ErrCatHTTP
 	}
-	if statusCode >= 400 {
+	if statusCode >= httpErrorThreshold {
 		return ErrCatHTTP
 	}
 
@@ -145,7 +145,7 @@ func extractDetail(errorReason string, cat ErrorCategory, statusCode int) string
 			}
 		}
 		if detail == "" {
-			detail = limitStr(errorReason, 30)
+			detail = limitStr(errorReason, errorDetailMaxLen)
 		}
 	case ErrCatTCP:
 		for _, keyword := range []string{"connection refused", "connection reset", "no route to host", "network unreachable"} {
@@ -155,7 +155,7 @@ func extractDetail(errorReason string, cat ErrorCategory, statusCode int) string
 			}
 		}
 		if detail == "" {
-			detail = limitStr(errorReason, 30)
+			detail = limitStr(errorReason, errorDetailMaxLen)
 		}
 	case ErrCatTLS:
 		for _, keyword := range []string{"certificate expired", "certificate has expired", "handshake failure", "unknown authority"} {
@@ -166,16 +166,16 @@ func extractDetail(errorReason string, cat ErrorCategory, statusCode int) string
 		}
 		if detail == "" && strings.Contains(lower, "x509:") {
 			idx := strings.Index(lower, "x509:")
-			detail = limitStr(errorReason[idx:], 30)
+			detail = limitStr(errorReason[idx:], errorDetailMaxLen)
 		}
 		if detail == "" {
-			detail = limitStr(errorReason, 30)
+			detail = limitStr(errorReason, errorDetailMaxLen)
 		}
 	case ErrCatHTTP:
 		if statusCode > 0 {
 			detail = fmt.Sprintf("HTTP %d", statusCode)
 		} else {
-			detail = limitStr(errorReason, 30)
+			detail = limitStr(errorReason, errorDetailMaxLen)
 		}
 	case ErrCatTimeout:
 		if strings.Contains(lower, "i/o timeout") {
@@ -187,12 +187,12 @@ func extractDetail(errorReason string, cat ErrorCategory, statusCode int) string
 		if strings.Contains(lower, "no icmp response") {
 			detail = "no response"
 		} else {
-			detail = limitStr(errorReason, 30)
+			detail = limitStr(errorReason, errorDetailMaxLen)
 		}
 	case ErrCatPrivate:
 		detail = "private IP blocked"
 	default:
-		detail = limitStr(errorReason, 30)
+		detail = limitStr(errorReason, errorDetailMaxLen)
 	}
 
 	return detail

internal/tui/histogram.go

@@ -0,0 +1,89 @@
+package tui
+
+import (
+	"fmt"
+	"strings"
+	"time"
+
+	"github.com/charmbracelet/lipgloss"
+)
+
+type latencyBucket struct {
+	label string
+	min   int64
+	max   int64
+	count int
+}
+
+func (m Model) latencyHistogram(latencies []time.Duration, statuses []bool, width int) string {
+	if len(latencies) == 0 || width < 30 {
+		return ""
+	}
+
+	buckets := []latencyBucket{
+		{"0-50ms", 0, 50, 0},
+		{"50-100ms", 50, 100, 0},
+		{"100-200ms", 100, 200, 0},
+		{"200-500ms", 200, 500, 0},
+		{"500ms+", 500, 999999, 0},
+	}
+
+	for i, l := range latencies {
+		if i < len(statuses) && !statuses[i] {
+			continue
+		}
+		ms := l.Milliseconds()
+		for j := range buckets {
+			if ms >= buckets[j].min && ms < buckets[j].max {
+				buckets[j].count++
+				break
+			}
+		}
+	}
+
+	maxCount := 0
+	for _, b := range buckets {
+		if b.count > maxCount {
+			maxCount = b.count
+		}
+	}
+	if maxCount == 0 {
+		return ""
+	}
+
+	labelW := 10
+	countW := len(fmt.Sprintf("%d", maxCount))
+	barW := width - labelW - countW - 4
+	if barW < 5 {
+		barW = 5
+	}
+
+	var sb strings.Builder
+	for _, b := range buckets {
+		fill := 0
+		if maxCount > 0 {
+			fill = b.count * barW / maxCount
+		}
+
+		var barColor lipgloss.Style
+		switch {
+		case b.min < 100:
+			barColor = m.st.specialStyle
+		case b.min < 500:
+			barColor = m.st.warnStyle
+		default:
+			barColor = m.st.dangerStyle
+		}
+
+		bar := barColor.Render(strings.Repeat("█", fill))
+		if fill < barW {
+			bar += m.st.subtleStyle.Render(strings.Repeat("░", barW-fill))
+		}
+
+		label := fmt.Sprintf("%*s", labelW, b.label)
+		count := fmt.Sprintf("%*d", countW, b.count)
+		sb.WriteString("  " + m.st.subtleStyle.Render(label) + " " + bar + " " + count + "\n")
+	}
+
+	return sb.String()
+}

internal/tui/messages.go

@@ -4,6 +4,7 @@ import (
 	"time"
 
 	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/monitor"
 )
 
 // tabRefreshTTL bounds how often the DB-backed tab data (alerts, users, nodes,
@@ -34,8 +35,9 @@ type tabDataMsg struct {
 // on entry and refreshed on the tab-data cadence so View never touches the
 // database.
 type detailDataMsg struct {
-	siteID  int
-	changes []models.StateChange
+	siteID    int
+	changes   []models.StateChange
+	dailyDays []monitor.DayReport
 }
 
 // historyDataMsg carries the full state-change history for the history view.

internal/tui/panel.go

@@ -0,0 +1,52 @@
+package tui
+
+import (
+	"strings"
+
+	"github.com/charmbracelet/lipgloss"
+)
+
+func (m Model) titledPanel(title, content string, width int, focused bool) string {
+	borderColor := m.theme.Border
+	titleColor := m.theme.Muted
+	if focused {
+		borderColor = m.theme.Accent
+		titleColor = m.theme.Accent
+	}
+
+	bc := lipgloss.NewStyle().Foreground(borderColor)
+	tc := lipgloss.NewStyle().Foreground(titleColor).Bold(true)
+
+	innerW := width - 2
+	if innerW < 10 {
+		innerW = 10
+	}
+
+	titleRendered := tc.Render(" " + title + " ")
+	titleLen := len([]rune(title)) + 2
+	fillLen := innerW - titleLen - 1
+	if fillLen < 0 {
+		fillLen = 0
+	}
+
+	top := bc.Render("╭─") + titleRendered + bc.Render(strings.Repeat("─", fillLen)+"╮")
+
+	contentStyle := lipgloss.NewStyle().Width(innerW).MaxWidth(innerW)
+	inner := contentStyle.Render(content)
+
+	var lines []string
+	lines = append(lines, top)
+	for _, line := range strings.Split(inner, "\n") {
+		lines = append(lines, bc.Render("│")+line+strings.Repeat(" ", max(0, innerW-lipgloss.Width(line)))+bc.Render("│"))
+	}
+	lines = append(lines, bc.Render("╰"+strings.Repeat("─", innerW)+"╯"))
+
+	return strings.Join(lines, "\n")
+}
+
+func max(a, b int) int {
+	if a > b {
+		return a
+	}
+	return b
+}

internal/tui/sparkline.go

@@ -17,6 +17,17 @@ func parseHex(hex string) (r, g, b uint8) {
 	return
 }
 
+func trueColorHex(c lipgloss.TerminalColor) string {
+	switch v := c.(type) {
+	case lipgloss.CompleteColor:
+		return v.TrueColor
+	case lipgloss.Color:
+		return string(v)
+	default:
+		return ""
+	}
+}
+
 func dimColor(hex string, brightness float64) lipgloss.Color {
 	r, g, b := parseHex(hex)
 	f := 0.3 + brightness*0.7
@@ -27,35 +38,36 @@ func dimColor(hex string, brightness float64) lipgloss.Color {
 	))
 }
 
-func withBg(s lipgloss.Style, bg lipgloss.Color) lipgloss.Style {
-	if bg != "" {
+func withBg(s lipgloss.Style, bg lipgloss.TerminalColor) lipgloss.Style {
+	if bg != nil {
 		return s.Background(bg)
 	}
 	return s
 }
 
-func (m Model) latencyStyle(ms int64, bg lipgloss.Color) lipgloss.Style {
-	var hex string
+func (m Model) latencyStyle(ms int64, bg lipgloss.TerminalColor) lipgloss.Style {
+	var base lipgloss.TerminalColor
 	var t float64
 	switch {
 	case ms < 200:
-		hex = m.st.sparkSuccess
+		base = m.st.sparkSuccess
 		t = float64(ms) / 200
 	case ms < 500:
-		hex = m.st.sparkWarning
+		base = m.st.sparkWarning
 		t = float64(ms-200) / 300
 	default:
-		hex = m.st.sparkDanger
+		base = m.st.sparkDanger
 		t = float64(ms-500) / 1500
 		if t > 1 {
 			t = 1
 		}
 	}
+	hex := trueColorHex(base)
 	s := lipgloss.NewStyle().Foreground(dimColor(hex, t))
 	return withBg(s, bg)
 }
 
-func (m Model) latencySparkline(latencies []time.Duration, statuses []bool, width int, bg lipgloss.Color) string {
+func (m Model) latencySparkline(latencies []time.Duration, statuses []bool, width int, bg lipgloss.TerminalColor) string {
 	if len(latencies) == 0 {
 		return withBg(m.st.subtleStyle, bg).Render(strings.Repeat("·", width))
 	}
@@ -103,7 +115,7 @@ func (m Model) latencySparkline(latencies []time.Duration, statuses []bool, widt
 	return sb.String()
 }
 
-func (m Model) heartbeatSparkline(statuses []bool, width int, bg lipgloss.Color) string {
+func (m Model) heartbeatSparkline(statuses []bool, width int, bg lipgloss.TerminalColor) string {
 	if len(statuses) == 0 {
 		return withBg(m.st.subtleStyle, bg).Render(strings.Repeat("·", width))
 	}
@@ -143,7 +155,7 @@ func resolveSparklineIndex(x, sparkWidth, dataLen int) int {
 	return offset + (x - padding)
 }
 
-func (m Model) groupSparkline(groupID int, width int, bg lipgloss.Color) string {
+func (m Model) groupSparkline(groupID int, width int, bg lipgloss.TerminalColor) string {
 	allSites := m.engine.GetAllSites()
 	var childStatuses [][]bool
 	for _, s := range allSites {

internal/tui/sparkline_test.go

@@ -5,10 +5,12 @@ import (
 	"testing"
 	"time"
 	"unicode/utf8"
+
+	"github.com/charmbracelet/lipgloss"
 )
 
 func TestLatencySparkline_Empty(t *testing.T) {
-	got := styledModel.latencySparkline(nil, nil, 10, "")
+	got := styledModel.latencySparkline(nil, nil, 10, nil)
 	if !strings.Contains(got, "··········") {
 		t.Errorf("empty sparkline should be dots, got %q", got)
 	}
@@ -17,7 +19,7 @@ func TestLatencySparkline_Empty(t *testing.T) {
 func TestLatencySparkline_SingleValue(t *testing.T) {
 	latencies := []time.Duration{100 * time.Millisecond}
 	statuses := []bool{true}
-	got := styledModel.latencySparkline(latencies, statuses, 5, "")
+	got := styledModel.latencySparkline(latencies, statuses, 5, nil)
 	if len(got) == 0 {
 		t.Error("sparkline should not be empty")
 	}
@@ -33,7 +35,7 @@ func TestLatencySparkline_WidthTruncation(t *testing.T) {
 		latencies[i] = time.Duration(i*50) * time.Millisecond
 		statuses[i] = true
 	}
-	got := styledModel.latencySparkline(latencies, statuses, 5, "")
+	got := styledModel.latencySparkline(latencies, statuses, 5, nil)
 	if len(got) == 0 {
 		t.Error("sparkline should not be empty")
 	}
@@ -45,7 +47,7 @@ func TestLatencySparkline_WidthTruncation(t *testing.T) {
 func TestLatencySparkline_RelativeHeight(t *testing.T) {
 	latencies := []time.Duration{10 * time.Millisecond, 50 * time.Millisecond, 10 * time.Millisecond}
 	statuses := []bool{true, true, true}
-	out := stripANSI(styledModel.latencySparkline(latencies, statuses, 3, ""))
+	out := stripANSI(styledModel.latencySparkline(latencies, statuses, 3, nil))
 	runes := []rune(out)
 	if len(runes) < 3 {
 		t.Fatalf("expected 3 runes, got %d", len(runes))
@@ -57,14 +59,14 @@ func TestLatencySparkline_RelativeHeight(t *testing.T) {
 
 func TestLatencyStyle_BandsProduceDifferentColors(t *testing.T) {
 	st := newStyles(themeFlexokiDark)
-	st.sparkSuccess = "#00ff00"
-	st.sparkWarning = "#ffff00"
-	st.sparkDanger = "#ff0000"
+	st.sparkSuccess = lipgloss.Color("#00ff00")
+	st.sparkWarning = lipgloss.Color("#ffff00")
+	st.sparkDanger = lipgloss.Color("#ff0000")
 	m := Model{st: st}
 
-	green := m.latencyStyle(50, "")
-	yellow := m.latencyStyle(300, "")
-	red := m.latencyStyle(800, "")
+	green := m.latencyStyle(50, nil)
+	yellow := m.latencyStyle(300, nil)
+	red := m.latencyStyle(800, nil)
 
 	gfg := green.GetForeground()
 	yfg := yellow.GetForeground()
@@ -77,11 +79,11 @@ func TestLatencyStyle_BandsProduceDifferentColors(t *testing.T) {
 
 func TestLatencyStyle_BrightnessVariesWithinBand(t *testing.T) {
 	st := newStyles(themeFlexokiDark)
-	st.sparkSuccess = "#00ff00"
+	st.sparkSuccess = lipgloss.Color("#00ff00")
 	m := Model{st: st}
 
-	dim := m.latencyStyle(10, "")
-	bright := m.latencyStyle(190, "")
+	dim := m.latencyStyle(10, nil)
+	bright := m.latencyStyle(190, nil)
 
 	if dim.GetForeground() == bright.GetForeground() {
 		t.Error("10ms and 190ms should have different brightness within green band")
@@ -91,7 +93,7 @@ func TestLatencyStyle_BrightnessVariesWithinBand(t *testing.T) {
 func TestLatencySparkline_OutputWidth(t *testing.T) {
 	latencies := []time.Duration{100 * time.Millisecond, 200 * time.Millisecond, 300 * time.Millisecond}
 	statuses := []bool{true, true, true}
-	got := styledModel.latencySparkline(latencies, statuses, 5, "")
+	got := styledModel.latencySparkline(latencies, statuses, 5, nil)
 	count := utf8.RuneCountInString(stripANSI(got))
 	if count != 5 {
 		t.Errorf("expected 5 rune-width output, got %d from %q", count, got)
@@ -116,7 +118,7 @@ func stripANSI(s string) string {
 }
 
 func TestHeartbeatSparkline_Empty(t *testing.T) {
-	got := styledModel.heartbeatSparkline(nil, 10, "")
+	got := styledModel.heartbeatSparkline(nil, 10, nil)
 	if !strings.Contains(got, "··········") {
 		t.Errorf("empty heartbeat should be dots, got %q", got)
 	}
@@ -124,7 +126,7 @@ func TestHeartbeatSparkline_Empty(t *testing.T) {
 
 func TestHeartbeatSparkline_Mixed(t *testing.T) {
 	statuses := []bool{true, false, true, true, false}
-	got := styledModel.heartbeatSparkline(statuses, 5, "")
+	got := styledModel.heartbeatSparkline(statuses, 5, nil)
 	if len(got) == 0 {
 		t.Error("heartbeat sparkline should not be empty")
 	}
@@ -132,7 +134,7 @@ func TestHeartbeatSparkline_Mixed(t *testing.T) {
 
 func TestHeartbeatSparkline_PaddedWidth(t *testing.T) {
 	statuses := []bool{true, true}
-	got := styledModel.heartbeatSparkline(statuses, 5, "")
+	got := styledModel.heartbeatSparkline(statuses, 5, nil)
 	if !strings.Contains(got, "···") {
 		t.Errorf("should have dot padding for width > data, got %q", got)
 	}

internal/tui/tab_alerts.go

@@ -1,7 +1,6 @@
 package tui
 
 import (
-	"context"
 	"fmt"
 	neturl "net/url"
 	"sort"
@@ -185,7 +184,7 @@ func (m Model) viewAlertsTab() string {
 	nameW := widths[2]
 	cfgW := widths[4]
 
-	return m.renderTable(
+	tbl := m.renderTable(
 		headers,
 		len(m.alerts),
 		func(start, end int) [][]string {
@@ -206,6 +205,90 @@ func (m Model) viewAlertsTab() string {
 		},
 		widths, nil,
 	)
+
+	var totalSent, totalFail int
+	for _, a := range m.alerts {
+		h := m.engine.GetAlertHealth(a.ID)
+		totalSent += h.SendCount
+		totalFail += h.FailCount
+	}
+	types := make(map[string]bool)
+	for _, a := range m.alerts {
+		types[a.Type] = true
+	}
+	failLabel := "failures"
+	if totalFail == 1 {
+		failLabel = "failure"
+	}
+	summary := fmt.Sprintf("%d channels · %d types · %d sent · %d %s",
+		len(m.alerts), len(types), totalSent, totalFail, failLabel)
+
+	var detail string
+	if m.cursor < len(m.alerts) {
+		a := m.alerts[m.cursor]
+		h := m.engine.GetAlertHealth(a.ID)
+		detail = m.alertSelectedDetail(a, h)
+	}
+
+	return tbl + "\n  " + detail + "\n  " + m.st.subtleStyle.Render(summary)
+}
+
+func (m Model) alertSelectedDetail(a models.AlertConfig, h monitor.AlertHealth) string {
+	dot := m.st.subtleStyle.Render("  ·  ")
+	label := m.st.subtleStyle
+
+	parts := []string{fmtAlertType(a.Type)}
+	parts = append(parts, m.fmtAlertConfigFull(a))
+
+	if !h.LastSendAt.IsZero() {
+		if h.LastSendOK {
+			parts = append(parts, m.st.specialStyle.Render("●")+" "+label.Render("sent")+" "+m.fmtTimeAgo(h.LastSendAt))
+		} else {
+			parts = append(parts, m.st.dangerStyle.Render("●")+" "+label.Render("failed")+" "+m.fmtTimeAgo(h.LastSendAt))
+		}
+	}
+	if h.SendCount > 0 {
+		parts = append(parts, label.Render(fmt.Sprintf("%d sent, %d failed", h.SendCount, h.FailCount)))
+	}
+
+	return strings.Join(parts, dot)
+}
+
+func (m Model) fmtAlertConfigFull(alert models.AlertConfig) string {
+	switch alert.Type {
+	case "email":
+		host := alert.Settings["host"]
+		to := alert.Settings["to"]
+		if host != "" && to != "" {
+			return fmt.Sprintf("%s → %s", host, to)
+		}
+		if host != "" {
+			return host
+		}
+		return m.st.subtleStyle.Render("—")
+	case "ntfy":
+		topic := alert.Settings["topic"]
+		url := alert.Settings["url"]
+		if url != "" && topic != "" {
+			return fmt.Sprintf("%s/%s", url, topic)
+		}
+		return m.st.subtleStyle.Render("—")
+	case "telegram":
+		if id := alert.Settings["chat_id"]; id != "" {
+			return fmt.Sprintf("chat:%s", id)
+		}
+		return m.st.subtleStyle.Render("—")
+	case "gotify":
+		if url := alert.Settings["url"]; url != "" {
+			return url
+		}
+		return m.st.subtleStyle.Render("—")
+	default:
+		if val, ok := alert.Settings["url"]; ok && val != "" {
+			return maskWebhookURL(val)
+		}
+		return m.st.subtleStyle.Render("—")
+	}
 }
 
 func (m Model) viewAlertDetailPanel() string {
@@ -518,15 +601,16 @@ func (m *Model) submitAlertForm() tea.Cmd {
 	}
 
 	st := m.store
+	ctx := m.ctx
 	id := m.editID
 	name, aType := d.Name, d.AlertType
 	m.state = stateDashboard
 	if id > 0 {
 		return writeCmd("Update alert", func() error {
-			return st.UpdateAlert(context.Background(), id, name, aType, settings)
+			return st.UpdateAlert(ctx, id, name, aType, settings)
 		})
 	}
 	return writeCmd("Add alert", func() error {
-		return st.AddAlert(context.Background(), name, aType, settings)
+		return st.AddAlert(ctx, name, aType, settings)
 	})
 }

internal/tui/tab_logs.go

@@ -3,6 +3,9 @@ package tui
 import (
 	"fmt"
 	"strings"
+
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+	"github.com/charmbracelet/lipgloss"
 )
 
 type logSeverity int
@@ -15,8 +18,8 @@ const (
 	severitySystem
 )
 
-func classifyLog(line string) logSeverity {
-	lower := strings.ToLower(line)
+func classifyLog(msg string) logSeverity {
+	lower := strings.ToLower(msg)
 	switch {
 	case strings.Contains(lower, "confirmed down"),
 		strings.Contains(lower, "is down"),
@@ -63,67 +66,49 @@ func (m Model) renderLogTag(sev logSeverity) string {
 	}
 }
 
-func (m Model) renderLogLine(line string) string {
-	sev := classifyLog(line)
+func (m Model) renderLogLine(entry models.LogEntry) string {
+	sev := classifyLog(entry.Message)
 	tag := m.renderLogTag(sev)
-
-	ts := ""
-	msg := line
-	if len(line) > 10 && line[0] == '[' {
-		if idx := strings.Index(line, "]"); idx > 0 && idx < 12 {
-			ts = m.st.subtleStyle.Render(line[1:idx])
-			msg = strings.TrimSpace(line[idx+1:])
-		}
-	}
-
-	if ts != "" {
-		return fmt.Sprintf("  %s  %s  %s", ts, tag, msg)
-	}
-	return fmt.Sprintf("  %s  %s", tag, msg)
+	ts := m.st.subtleStyle.Render(entry.CreatedAt.Local().Format("01/02 15:04"))
+	return fmt.Sprintf("  %s  %s  %s", ts, tag, entry.Message)
+}
+
+func (m Model) viewLogsFullscreen() string {
+	header := m.st.subtleStyle.Render("  Logs") + "\n" + m.divider()
+
+	filterLabel := m.st.subtleStyle.Render("[f] All")
+	if m.logFilterImportant {
+		filterLabel = m.st.subtleStyle.Render("[f] Important only")
+	}
+	countLabel := m.st.subtleStyle.Render(fmt.Sprintf("%d/%d", m.logShown, m.logTotal))
+	footer := m.divider() + "\n  " + m.st.subtleStyle.Render("[Esc] Back") + "  " + filterLabel + "  " + countLabel
+
+	m.logViewport.Width = m.termWidth - chromePadH
+	m.logViewport.Height = m.termHeight - 8
+
+	return lipgloss.NewStyle().Padding(1, 2).Render(
+		header + "\n" + m.logViewport.View() + "\n" + footer)
 }
 
-// refreshLogContent rebuilds the log viewport from the full engine log list,
-// filtering before windowing so the entry count and "(n hidden)" reflect all
-// logs, not just the visible viewport slice.
 func (m *Model) refreshLogContent() {
 	var rendered []string
 	total := 0
 	shown := 0
 
-	for _, line := range m.engine.GetLogs() {
-		if strings.TrimSpace(line) == "" {
+	for _, entry := range m.engine.GetLogs() {
+		if strings.TrimSpace(entry.Message) == "" {
 			continue
 		}
 		total++
-		sev := classifyLog(line)
+		sev := classifyLog(entry.Message)
 		if m.logFilterImportant && !isImportantLog(sev) {
 			continue
 		}
 		shown++
-		rendered = append(rendered, m.renderLogLine(line))
+		rendered = append(rendered, m.renderLogLine(entry))
 	}
 
 	m.logTotal = total
 	m.logShown = shown
 	m.logViewport.SetContent(strings.Join(rendered, "\n"))
 }
-
-func (m Model) viewLogsTab() string {
-	if m.logTotal == 0 {
-		return m.emptyState("No log entries yet.", "Logs appear as monitors run checks")
-	}
-
-	filterLabel := "All"
-	if m.logFilterImportant {
-		filterLabel = "Important"
-	}
-
-	header := m.st.subtleStyle.Render(fmt.Sprintf(
-		"  %d entries  Filter: %s", m.logShown, filterLabel))
-
-	if m.logFilterImportant && m.logShown < m.logTotal {
-		header += m.st.subtleStyle.Render(fmt.Sprintf("  (%d hidden)", m.logTotal-m.logShown))
-	}
-
-	return "\n" + header + "\n\n" + m.logViewport.View()
-}

internal/tui/tab_logs_sidebar.go

@@ -0,0 +1,98 @@
+package tui
+
+import (
+	"strings"
+
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+	"github.com/charmbracelet/lipgloss"
+)
+
+func (m Model) renderCompactLogLine(entry models.LogEntry, maxW int) string {
+	sev := classifyLog(entry.Message)
+
+	var tag string
+	switch sev {
+	case severityDown:
+		tag = m.st.dangerStyle.Render("▼")
+	case severityUp:
+		tag = m.st.specialStyle.Render("▲")
+	case severityWarn:
+		tag = m.st.warnStyle.Render("◆")
+	case severitySystem:
+		tag = m.st.titleStyle.Render("●")
+	default:
+		tag = m.st.subtleStyle.Render("·")
+	}
+
+	ts := entry.CreatedAt.Local().Format("01/02 15:04")
+
+	msg := entry.Message
+	msg = strings.TrimPrefix(msg, "Monitor ")
+	msg = strings.TrimPrefix(msg, "Push ")
+
+	prefixW := len(ts) + 4
+	msgW := maxW - prefixW
+	if msgW < 5 {
+		msgW = 5
+	}
+	msg = limitStr(msg, msgW)
+
+	return " " + m.st.subtleStyle.Render(ts) + " " + tag + " " + msg
+}
+
+func (m Model) viewLogsSidebar(width, maxLines int) string {
+	logs := m.engine.GetLogs()
+	if len(logs) == 0 {
+		return m.st.subtleStyle.Render("  No logs yet")
+	}
+
+	sidebarStyle := lipgloss.NewStyle().Width(width).MaxWidth(width)
+
+	var all []string
+	for _, entry := range logs {
+		if strings.TrimSpace(entry.Message) == "" {
+			continue
+		}
+		if m.logFilterImportant && !isImportantLog(classifyLog(entry.Message)) {
+			continue
+		}
+		all = append(all, m.renderCompactLogLine(entry, width))
+	}
+
+	start := m.logScrollOffset
+	if start > len(all) {
+		start = len(all)
+	}
+	end := start + maxLines
+	if end > len(all) {
+		end = len(all)
+	}
+	visible := all[start:end]
+
+	return sidebarStyle.Render(strings.Join(visible, "\n"))
+}
+
+func (m *Model) scrollLogs(delta int) {
+	logs := m.engine.GetLogs()
+	total := 0
+	for _, entry := range logs {
+		if strings.TrimSpace(entry.Message) == "" {
+			continue
+		}
+		if m.logFilterImportant && !isImportantLog(classifyLog(entry.Message)) {
+			continue
+		}
+		total++
+	}
+
+	m.logScrollOffset += delta
+	if m.logScrollOffset < 0 {
+		m.logScrollOffset = 0
+	}
+	if m.logScrollOffset > total-1 {
+		m.logScrollOffset = total - 1
+	}
+	if m.logScrollOffset < 0 {
+		m.logScrollOffset = 0
+	}
+}

internal/tui/tab_maint.go

@@ -1,7 +1,6 @@
 package tui
 
 import (
-	"context"
 	"fmt"
 	"strconv"
 	"time"
@@ -107,7 +106,7 @@ func (m Model) viewMaintTab() string {
 	monW := widths[3]
 	timeW := widths[5]
 
-	return m.renderTable(
+	tbl := m.renderTable(
 		headers,
 		len(m.maintenanceWindows),
 		func(start, end int) [][]string {
@@ -130,6 +129,21 @@ func (m Model) viewMaintTab() string {
 		widths,
 		nil,
 	)
+
+	now := time.Now()
+	var active, scheduled, ended int
+	for _, mw := range m.maintenanceWindows {
+		if mw.StartTime.After(now) {
+			scheduled++
+		} else if !mw.EndTime.IsZero() && mw.EndTime.Before(now) {
+			ended++
+		} else {
+			active++
+		}
+	}
+	summary := fmt.Sprintf("%d active · %d scheduled · %d ended", active, scheduled, ended)
+
+	return tbl + "\n  " + m.st.subtleStyle.Render(summary)
 }
 
 func (m *Model) initMaintHuhForm() tea.Cmd {
@@ -239,8 +253,13 @@ func (m *Model) submitMaintForm() tea.Cmd {
 	}
 
 	st := m.store
+	ctx := m.ctx
 	m.state = stateDashboard
 	return writeCmd("Add maintenance window", func() error {
-		return st.AddMaintenanceWindow(context.Background(), mw)
+		overlaps, _ := st.GetOverlappingMaintenanceWindows(ctx, mw.MonitorID, mw.StartTime, mw.EndTime)
+		if len(overlaps) > 0 {
+			_ = st.SaveLog(ctx, fmt.Sprintf("Overlap: new window %q overlaps with existing %q", mw.Title, overlaps[0].Title))
+		}
+		return st.AddMaintenanceWindow(ctx, mw)
 	})
 }

internal/tui/tab_nodes.go

@@ -1,6 +1,8 @@
 package tui
 
 import (
+	"fmt"
+	"strings"
 	"time"
 )
 
@@ -20,7 +22,7 @@ func (m Model) viewNodesTab() string {
 	}
 	nameW := widths[0]
 
-	return m.renderTable(
+	tbl := m.renderTable(
 		headers,
 		len(m.nodes),
 		func(start, end int) [][]string {
@@ -48,6 +50,30 @@ func (m Model) viewNodesTab() string {
 		widths,
 		nil,
 	)
+
+	var online int
+	regions := make(map[string]bool)
+	var leader string
+	for _, n := range m.nodes {
+		if time.Since(n.LastSeen) < nodeOnlineThreshold {
+			online++
+		}
+		if n.Region != "" {
+			regions[n.Region] = true
+		}
+		if n.Name == "leader" {
+			leader = n.ID
+		}
+	}
+	parts := []string{fmt.Sprintf("%d/%d online", online, len(m.nodes))}
+	if leader != "" {
+		parts = append(parts, fmt.Sprintf("leader: %s", leader))
+	}
+	if len(regions) > 0 {
+		parts = append(parts, fmt.Sprintf("%d regions", len(regions)))
+	}
+
+	return tbl + "\n  " + m.st.subtleStyle.Render(strings.Join(parts, " · "))
 }
 
 func (m Model) fmtNodeStatus(lastSeen time.Time) string {
@@ -55,10 +81,10 @@ func (m Model) fmtNodeStatus(lastSeen time.Time) string {
 		return m.st.subtleStyle.Render("UNKNOWN")
 	}
 	ago := time.Since(lastSeen)
-	if ago < 60*time.Second {
+	if ago < nodeOnlineThreshold {
 		return m.st.specialStyle.Render("ONLINE")
 	}
-	if ago < 5*time.Minute {
+	if ago < nodeStaleThreshold {
 		return m.st.warnStyle.Render("STALE")
 	}
 	return m.st.dangerStyle.Render("OFFLINE")

internal/tui/tab_settings.go

@@ -0,0 +1,62 @@
+package tui
+
+import (
+	"fmt"
+
+	"github.com/charmbracelet/lipgloss"
+)
+
+func (m Model) viewSettingsTab() string {
+	maxSections := 2
+	if m.isAdmin {
+		maxSections = 3
+	}
+
+	sections := []string{"Alerts", "Nodes"}
+	if m.isAdmin {
+		sections = append(sections, "Users")
+	}
+	_ = maxSections
+
+	var tabs []string
+	for i, name := range sections {
+		var rendered string
+		if i == m.settingsSection {
+			rendered = m.st.activeTab.Render(name)
+		} else {
+			rendered = m.st.inactiveTab.Render(name)
+		}
+		tabs = append(tabs, m.zones.Mark(fmt.Sprintf("section-%d", i), rendered))
+	}
+	header := lipgloss.JoinHorizontal(lipgloss.Top, tabs...)
+
+	var content string
+	switch m.settingsSection {
+	case sectionAlerts:
+		content = m.viewAlertsTab()
+	case sectionNodes:
+		content = m.viewNodesTab()
+	case sectionUsers:
+		if m.isAdmin {
+			content = m.viewUsersTab()
+		}
+	}
+
+	return header + "\n" + content
+}
+
+func (m *Model) switchSettingsSection(idx int) {
+	max := 1
+	if m.isAdmin {
+		max = 2
+	}
+	if idx > max {
+		idx = 0
+	}
+	if idx < 0 {
+		idx = max
+	}
+	m.settingsSection = idx
+	m.cursor = 0
+	m.tableOffset = 0
+}

internal/tui/tab_sites.go

@@ -1,7 +1,6 @@
 package tui
 
 import (
-	"context"
 	"fmt"
 	"net/url"
 	"strconv"
@@ -82,8 +81,12 @@ func (m Model) computeLayout() tableLayout {
 	var widths []int
 	var fixed int
 
+	cw := m.contentWidth
+	if cw == 0 {
+		cw = m.termWidth
+	}
 	for _, c := range siteColumns {
-		if c.minTerm > 0 && m.termWidth < c.minTerm {
+		if c.minTerm > 0 && cw < c.minTerm {
 			continue
 		}
 		active = append(active, c.key)
@@ -102,9 +105,37 @@ func (m Model) computeLayout() tableLayout {
 		}
 	}
 
+	sortColMap := map[int]colKey{
+		sortStatus:  colStatus,
+		sortName:    colName,
+		sortLatency: colLatency,
+	}
+	sortableKeys := map[colKey]string{
+		colStatus:  "sort-status",
+		colName:    "sort-name",
+		colLatency: "sort-latency",
+	}
+	if sortedKey, ok := sortColMap[m.sortColumn]; ok {
+		arrow := "▼"
+		if m.sortAsc {
+			arrow = "▲"
+		}
+		for i, k := range active {
+			if k == sortedKey {
+				headers[i] = headers[i] + arrow
+				break
+			}
+		}
+	}
+	for i, k := range active {
+		if zoneID, ok := sortableKeys[k]; ok {
+			headers[i] = m.zones.Mark(zoneID, headers[i])
+		}
+	}
+
 	numCols := len(headers)
 	borderOverhead := 2 + (numCols - 1)
-	avail := m.termWidth - chromePadH - 2 - borderOverhead - fixed
+	avail := cw - chromePadH - 2 - borderOverhead - fixed
 	if avail < 20 {
 		avail = 20
 	}
@@ -204,7 +235,7 @@ func (m Model) viewSitesTab() string {
 			for i := start; i < end; i++ {
 				site := m.sites[i]
 				rowIdx := i - start
-				var rowBg lipgloss.Color
+				var rowBg lipgloss.TerminalColor
 				if i == m.cursor {
 					rowBg = m.theme.SelectedBg
 				} else if rowIdx%2 == 1 {
@@ -582,10 +613,11 @@ func (m *Model) submitSiteForm() tea.Cmd {
 	}
 
 	st := m.store
+	ctx := m.ctx
 	m.state = stateDashboard
 	if m.editID > 0 {
 		m.engine.UpdateSiteConfig(cfg)
-		return writeCmd("Update site", func() error { return st.UpdateSite(context.Background(), cfg) })
+		return writeCmd("Update site", func() error { return st.UpdateSite(ctx, cfg) })
 	}
-	return writeCmd("Add site", func() error { return st.AddSite(context.Background(), cfg) })
+	return writeCmd("Add site", func() error { return st.AddSite(ctx, cfg) })
 }

internal/tui/tab_users.go

@@ -1,7 +1,6 @@
 package tui
 
 import (
-	"context"
 	"fmt"
 
 	tea "github.com/charmbracelet/bubbletea"
@@ -114,15 +113,16 @@ func (m *Model) initUserHuhForm() tea.Cmd {
 func (m *Model) submitUserForm() tea.Cmd {
 	d := m.userFormData
 	st := m.store
+	ctx := m.ctx
 	id := m.editID
 	username, key, role := d.Username, d.PublicKey, d.Role
-	m.state = stateUsers
+	m.state = stateDashboard
 	if id > 0 {
 		return writeCmd("Update user", func() error {
-			return st.UpdateUser(context.Background(), id, username, key, role)
+			return st.UpdateUser(ctx, id, username, key, role)
 		})
 	}
 	return writeCmd("Add user", func() error {
-		return st.AddUser(context.Background(), username, key, role)
+		return st.AddUser(ctx, username, key, role)
 	})
 }

internal/tui/table_helpers.go

@@ -13,7 +13,11 @@ const (
 )
 
 func (m Model) isWide() bool {
-	return m.termWidth >= wideBreakpoint
+	w := m.contentWidth
+	if w == 0 {
+		w = m.termWidth
+	}
+	return w >= wideBreakpoint
 }
 
 func (m Model) renderTable(headers []string, items int, buildRows func(start, end int) [][]string, colWidths []int, styleOverride StyleOverride) string {
@@ -35,7 +39,11 @@ func (m Model) renderTable(headers []string, items int, buildRows func(start, en
 	}
 	borderOverhead := 2 + len(colWidths) - 1
 	tableWidth := colTotal + borderOverhead
-	maxWidth := m.termWidth - chromePadH - 2
+	cw := m.contentWidth
+	if cw == 0 {
+		cw = m.termWidth
+	}
+	maxWidth := cw - chromePadH - 2
 	if tableWidth > maxWidth {
 		tableWidth = maxWidth
 	}
@@ -44,8 +52,7 @@ func (m Model) renderTable(headers []string, items int, buildRows func(start, en
 	}
 
 	t := table.New().
-		Border(lipgloss.RoundedBorder()).
-		BorderStyle(m.st.tableBorderStyle).
+		Border(lipgloss.HiddenBorder()).
 		Width(tableWidth).
 		Headers(headers...).
 		Rows(rows...).
@@ -86,5 +93,5 @@ func (m Model) renderTable(headers []string, items int, buildRows func(start, en
 			return base
 		})
 
-	return "\n" + t.Render()
+	return t.Render()
 }

internal/tui/theme.go

@@ -5,35 +5,43 @@ import (
 	"github.com/charmbracelet/lipgloss"
 )
 
+func cc(hex, ansi string) lipgloss.CompleteColor {
+	return lipgloss.CompleteColor{
+		TrueColor: hex,
+		ANSI256:   hex,
+		ANSI:      ansi,
+	}
+}
+
 type Theme struct {
 	Name string
 
 	// Base layers
-	Bg      lipgloss.Color
-	Surface lipgloss.Color
-	Panel   lipgloss.Color
-	Border  lipgloss.Color
+	Bg      lipgloss.TerminalColor
+	Surface lipgloss.TerminalColor
+	Panel   lipgloss.TerminalColor
+	Border  lipgloss.TerminalColor
 
 	// Text
-	Fg     lipgloss.Color
-	Muted  lipgloss.Color
-	Subtle lipgloss.Color
+	Fg     lipgloss.TerminalColor
+	Muted  lipgloss.TerminalColor
+	Subtle lipgloss.TerminalColor
 
 	// Semantic
-	Success lipgloss.Color
-	Warning lipgloss.Color
-	Stale   lipgloss.Color
-	Danger  lipgloss.Color
-	Info    lipgloss.Color
-	Accent  lipgloss.Color
-	Purple  lipgloss.Color
+	Success lipgloss.TerminalColor
+	Warning lipgloss.TerminalColor
+	Stale   lipgloss.TerminalColor
+	Danger  lipgloss.TerminalColor
+	Info    lipgloss.TerminalColor
+	Accent  lipgloss.TerminalColor
+	Purple  lipgloss.TerminalColor
 
 	// Table
-	ZebraBg lipgloss.Color
+	ZebraBg lipgloss.TerminalColor
 
 	// Selection
-	SelectedFg lipgloss.Color
-	SelectedBg lipgloss.Color
+	SelectedFg lipgloss.TerminalColor
+	SelectedBg lipgloss.TerminalColor
 }
 
 var themes = []Theme{
@@ -46,107 +54,107 @@ var themes = []Theme{
 
 var themeFlexokiDark = Theme{
 	Name:       "Flexoki Dark",
-	Bg:         "#1C1B1A",
-	Surface:    "#282726",
-	Panel:      "#343331",
-	Border:     "#575653",
-	Fg:         "#CECDC3",
-	Muted:      "#878580",
-	Subtle:     "#6F6E69",
-	Success:    "#879A39",
-	Warning:    "#D0A215",
-	Stale:      "#DA702C",
-	Danger:     "#D14D41",
-	Info:       "#4385BE",
-	Accent:     "#3AA99F",
-	Purple:     "#8B7EC8",
-	ZebraBg:    "#222120",
-	SelectedFg: "#FFFCF0",
-	SelectedBg: "#403E3C",
+	Bg:         cc("#1C1B1A", ""),
+	Surface:    cc("#282726", ""),
+	Panel:      cc("#343331", ""),
+	Border:     cc("#575653", "8"),
+	Fg:         cc("#CECDC3", "15"),
+	Muted:      cc("#878580", "7"),
+	Subtle:     cc("#6F6E69", "7"),
+	Success:    cc("#879A39", "10"),
+	Warning:    cc("#D0A215", "11"),
+	Stale:      cc("#DA702C", "3"),
+	Danger:     cc("#D14D41", "9"),
+	Info:       cc("#4385BE", "12"),
+	Accent:     cc("#3AA99F", "14"),
+	Purple:     cc("#8B7EC8", "13"),
+	ZebraBg:    cc("#222120", ""),
+	SelectedFg: cc("#FFFCF0", "15"),
+	SelectedBg: cc("#403E3C", "4"),
 }
 
 var themeTokyoNight = Theme{
 	Name:       "Tokyo Night",
-	Bg:         "#1a1b26",
-	Surface:    "#24283b",
-	Panel:      "#292e42",
-	Border:     "#3b4261",
-	Fg:         "#c0caf5",
-	Muted:      "#a9b1d6",
-	Subtle:     "#565f89",
-	Success:    "#9ece6a",
-	Warning:    "#e0af68",
-	Stale:      "#ff9e64",
-	Danger:     "#f7768e",
-	Info:       "#7aa2f7",
-	Accent:     "#7dcfff",
-	Purple:     "#bb9af7",
-	ZebraBg:    "#1c1d28",
-	SelectedFg: "#c0caf5",
-	SelectedBg: "#292e42",
+	Bg:         cc("#1a1b26", ""),
+	Surface:    cc("#24283b", ""),
+	Panel:      cc("#292e42", ""),
+	Border:     cc("#3b4261", "8"),
+	Fg:         cc("#c0caf5", "15"),
+	Muted:      cc("#a9b1d6", "7"),
+	Subtle:     cc("#565f89", "7"),
+	Success:    cc("#9ece6a", "10"),
+	Warning:    cc("#e0af68", "11"),
+	Stale:      cc("#ff9e64", "3"),
+	Danger:     cc("#f7768e", "9"),
+	Info:       cc("#7aa2f7", "12"),
+	Accent:     cc("#7dcfff", "14"),
+	Purple:     cc("#bb9af7", "13"),
+	ZebraBg:    cc("#1c1d28", ""),
+	SelectedFg: cc("#c0caf5", "15"),
+	SelectedBg: cc("#292e42", "4"),
 }
 
 var themeGruvbox = Theme{
 	Name:       "Gruvbox",
-	Bg:         "#282828",
-	Surface:    "#3c3836",
-	Panel:      "#504945",
-	Border:     "#665c54",
-	Fg:         "#ebdbb2",
-	Muted:      "#bdae93",
-	Subtle:     "#7c6f64",
-	Success:    "#b8bb26",
-	Warning:    "#fabd2f",
-	Stale:      "#fe8019",
-	Danger:     "#fb4934",
-	Info:       "#83a598",
-	Accent:     "#8ec07c",
-	Purple:     "#d3869b",
-	ZebraBg:    "#2a2a2a",
-	SelectedFg: "#fbf1c7",
-	SelectedBg: "#504945",
+	Bg:         cc("#282828", ""),
+	Surface:    cc("#3c3836", ""),
+	Panel:      cc("#504945", ""),
+	Border:     cc("#665c54", "8"),
+	Fg:         cc("#ebdbb2", "15"),
+	Muted:      cc("#bdae93", "7"),
+	Subtle:     cc("#7c6f64", "7"),
+	Success:    cc("#b8bb26", "10"),
+	Warning:    cc("#fabd2f", "11"),
+	Stale:      cc("#fe8019", "3"),
+	Danger:     cc("#fb4934", "9"),
+	Info:       cc("#83a598", "12"),
+	Accent:     cc("#8ec07c", "14"),
+	Purple:     cc("#d3869b", "13"),
+	ZebraBg:    cc("#2a2a2a", ""),
+	SelectedFg: cc("#fbf1c7", "15"),
+	SelectedBg: cc("#504945", "4"),
 }
 
 var themeCatppuccinMocha = Theme{
 	Name:       "Catppuccin Mocha",
-	Bg:         "#1e1e2e",
-	Surface:    "#313244",
-	Panel:      "#45475a",
-	Border:     "#585b70",
-	Fg:         "#cdd6f4",
-	Muted:      "#a6adc8",
-	Subtle:     "#6c7086",
-	Success:    "#a6e3a1",
-	Warning:    "#f9e2af",
-	Stale:      "#fab387",
-	Danger:     "#f38ba8",
-	Info:       "#89b4fa",
-	Accent:     "#94e2d5",
-	Purple:     "#cba6f7",
-	ZebraBg:    "#232334",
-	SelectedFg: "#cdd6f4",
-	SelectedBg: "#45475a",
+	Bg:         cc("#1e1e2e", ""),
+	Surface:    cc("#313244", ""),
+	Panel:      cc("#45475a", ""),
+	Border:     cc("#585b70", "8"),
+	Fg:         cc("#cdd6f4", "15"),
+	Muted:      cc("#a6adc8", "7"),
+	Subtle:     cc("#6c7086", "7"),
+	Success:    cc("#a6e3a1", "10"),
+	Warning:    cc("#f9e2af", "11"),
+	Stale:      cc("#fab387", "3"),
+	Danger:     cc("#f38ba8", "9"),
+	Info:       cc("#89b4fa", "12"),
+	Accent:     cc("#94e2d5", "14"),
+	Purple:     cc("#cba6f7", "13"),
+	ZebraBg:    cc("#232334", ""),
+	SelectedFg: cc("#cdd6f4", "15"),
+	SelectedBg: cc("#45475a", "4"),
 }
 
 var themeNord = Theme{
 	Name:       "Nord",
-	Bg:         "#2e3440",
-	Surface:    "#3b4252",
-	Panel:      "#434c5e",
-	Border:     "#4c566a",
-	Fg:         "#d8dee9",
-	Muted:      "#d8dee9",
-	Subtle:     "#4c566a",
-	Success:    "#a3be8c",
-	Warning:    "#ebcb8b",
-	Stale:      "#d08770",
-	Danger:     "#bf616a",
-	Info:       "#81a1c1",
-	Accent:     "#88c0d0",
-	Purple:     "#b48ead",
-	ZebraBg:    "#323845",
-	SelectedFg: "#eceff4",
-	SelectedBg: "#434c5e",
+	Bg:         cc("#2e3440", ""),
+	Surface:    cc("#3b4252", ""),
+	Panel:      cc("#434c5e", ""),
+	Border:     cc("#4c566a", "8"),
+	Fg:         cc("#d8dee9", "15"),
+	Muted:      cc("#d8dee9", "7"),
+	Subtle:     cc("#4c566a", "7"),
+	Success:    cc("#a3be8c", "10"),
+	Warning:    cc("#ebcb8b", "11"),
+	Stale:      cc("#d08770", "3"),
+	Danger:     cc("#bf616a", "9"),
+	Info:       cc("#81a1c1", "12"),
+	Accent:     cc("#88c0d0", "14"),
+	Purple:     cc("#b48ead", "13"),
+	ZebraBg:    cc("#323845", ""),
+	SelectedFg: cc("#eceff4", "15"),
+	SelectedBg: cc("#434c5e", "4"),
 }
 
 func (t Theme) HuhTheme() *huh.Theme {

internal/tui/tui.go

@@ -30,9 +30,9 @@ type styles struct {
 	activeTab    lipgloss.Style
 	inactiveTab  lipgloss.Style
 
-	sparkSuccess string
-	sparkWarning string
-	sparkDanger  string
+	sparkSuccess lipgloss.TerminalColor
+	sparkWarning lipgloss.TerminalColor
+	sparkDanger  lipgloss.TerminalColor
 
 	tableHeaderStyle   lipgloss.Style
 	tableCellStyle     lipgloss.Style
@@ -46,23 +46,23 @@ type styles struct {
 
 func newStyles(t Theme) *styles {
 	return &styles{
-		subtleStyle:  lipgloss.NewStyle().Foreground(t.Subtle),
+		subtleStyle:  lipgloss.NewStyle().Foreground(t.Subtle).Faint(true),
 		specialStyle: lipgloss.NewStyle().Foreground(t.Success),
-		warnStyle:    lipgloss.NewStyle().Foreground(t.Warning),
-		staleStyle:   lipgloss.NewStyle().Foreground(t.Stale),
-		dangerStyle:  lipgloss.NewStyle().Foreground(t.Danger),
+		warnStyle:    lipgloss.NewStyle().Foreground(t.Warning).Bold(true),
+		staleStyle:   lipgloss.NewStyle().Foreground(t.Stale).Faint(true),
+		dangerStyle:  lipgloss.NewStyle().Foreground(t.Danger).Bold(true),
 		titleStyle:   lipgloss.NewStyle().Foreground(t.Accent).Bold(true),
 		activeTab:    lipgloss.NewStyle().Background(t.Surface).Foreground(t.Accent).Bold(true).Padding(0, 1),
-		inactiveTab:  lipgloss.NewStyle().Padding(0, 1).Foreground(t.Muted),
+		inactiveTab:  lipgloss.NewStyle().Padding(0, 1).Foreground(t.Muted).Faint(true),
 
-		sparkSuccess: string(t.Success),
-		sparkWarning: string(t.Warning),
-		sparkDanger:  string(t.Danger),
+		sparkSuccess: t.Success,
+		sparkWarning: t.Warning,
+		sparkDanger:  t.Danger,
 
 		tableHeaderStyle:   lipgloss.NewStyle().Foreground(t.Accent).Bold(true).Padding(0, 1),
 		tableCellStyle:     lipgloss.NewStyle().Padding(0, 1),
 		tableSelectedStyle: lipgloss.NewStyle().Padding(0, 1).Bold(true).Foreground(t.SelectedFg).Background(t.SelectedBg),
-		tableBorderStyle:   lipgloss.NewStyle().Foreground(t.Border),
+		tableBorderStyle:   lipgloss.NewStyle().Foreground(t.Border).Faint(true),
 		tableZebraStyle:    lipgloss.NewStyle().Padding(0, 1).Background(t.ZebraBg),
 
 		siteGroupStyle: lipgloss.NewStyle().Padding(0, 1).Bold(true).Foreground(t.Accent),
@@ -84,6 +84,31 @@ const (
 	detailSparkWidth = 40
 )
 
+const (
+	tabMonitors = 0
+	tabMaint    = 1
+	tabSettings = 2
+)
+
+const (
+	sectionAlerts = 0
+	sectionNodes  = 1
+	sectionUsers  = 2
+)
+
+const (
+	panelMonitors = 0
+	panelLogs     = 1
+	panelDetail   = 2
+)
+
+const (
+	sortStatus  = 0
+	sortName    = 1
+	sortLatency = 2
+	sortMax     = 3
+)
+
 type sessionState int
 
 const (
@@ -102,16 +127,22 @@ const (
 )
 
 type Model struct {
-	state        sessionState
-	currentTab   int
-	cursor       int
-	selectedID   int
-	tableOffset  int
-	maxTableRows int
-	termWidth    int
-	termHeight   int
-	editID       int
-	editToken    string
+	state           sessionState
+	currentTab      int
+	settingsSection int
+	cursor          int
+	selectedID      int
+	sortColumn      int
+	sortAsc         bool
+	tableOffset     int
+	maxTableRows    int
+	termWidth       int
+	termHeight      int
+	contentWidth    int
+	focusedPanel    int
+	logScrollOffset int
+	editID          int
+	editToken       string
 
 	huhForm       *huh.Form
 	siteFormData  *siteFormData
@@ -144,6 +175,7 @@ type Model struct {
 	deleteName string
 	deleteTab  int
 
+	ctx        context.Context
 	collapsed  map[int]bool
 	store      store.Store
 	engine     *monitor.Engine
@@ -165,9 +197,12 @@ type Model struct {
 	lastTabLoad        time.Time // last dispatch of loadTabDataCmd (throttle)
 	tabSeq             int       // seq of the newest issued tab-data load
 
-	// detail-panel state-change history, loaded on enter so View does no DB IO
+	logsOpen            bool
+	detailOpen          bool
 	detailChanges       []models.StateChange
 	detailChangesSiteID int
+	detailDailyDays     []monitor.DayReport
+	detailViewport      viewport.Model
 
 	filterMode bool
 	filterText string
@@ -180,14 +215,14 @@ type Model struct {
 	version  string
 }
 
-func InitialModel(isAdmin bool, s store.Store, eng *monitor.Engine, version string) Model {
+func InitialModel(ctx context.Context, isAdmin bool, s store.Store, eng *monitor.Engine, version string) Model {
 	vpLogs := viewport.New(100, 20)
 	vpLogs.SetContent("Waiting for logs...")
 	z := zone.New()
 	spring := harmonica.NewSpring(harmonica.FPS(10), 6.0, 0.4)
-	collapsed := loadCollapsed(s)
+	collapsed := loadCollapsed(ctx, s)
 
-	themeName, _ := s.GetPreference(context.Background(), "theme")
+	themeName, _ := s.GetPreference(ctx, "theme")
 	theme := themeByName(themeName)
 	themeIdx := 0
 	for i, t := range themes {
@@ -197,7 +232,10 @@ func InitialModel(isAdmin bool, s store.Store, eng *monitor.Engine, version stri
 		}
 	}
 
+	detailPref, _ := s.GetPreference(ctx, "detail_open")
+
 	return Model{
+		ctx:             ctx,
 		state:           stateDashboard,
 		logViewport:     vpLogs,
 		maxTableRows:    5,
@@ -210,6 +248,8 @@ func InitialModel(isAdmin bool, s store.Store, eng *monitor.Engine, version stri
 		theme:           theme,
 		themeIndex:      themeIdx,
 		st:              newStyles(theme),
+		logsOpen:        true,
+		detailOpen:      detailPref == "true",
 		demoMode:        os.Getenv("UPTOP_DEMO") == "1",
 		version:         version,
 		sparkTooltipIdx: -1,

internal/tui/update.go

@@ -1,7 +1,6 @@
 package tui
 
 import (
-	"context"
 	"fmt"
 	"time"
 
@@ -28,6 +27,7 @@ func (m Model) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 		}
 		m.detailChanges = msg.changes
 		m.detailChangesSiteID = msg.siteID
+		m.detailDailyDays = msg.dailyDays
 		return m, nil
 	case historyDataMsg:
 		if msg.siteID != m.historySiteID {
@@ -53,6 +53,9 @@ func (m Model) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 	if m.state == stateFormSite || m.state == stateFormAlert || m.state == stateFormUser || m.state == stateFormMaint {
 		return m.handleFormMsg(msg)
 	}
+	if m.state == stateLogs {
+		return m.handleLogsFullscreen(msg)
+	}
 
 	switch msg := msg.(type) {
 	case tea.MouseMsg:
@@ -75,34 +78,32 @@ func (m *Model) handleConfirmDelete(msg tea.Msg) (tea.Model, tea.Cmd) {
 		// writeDoneMsg reload converges the UI back to the DB state (and the
 		// engine poll loop re-adds a site that is still in the DB).
 		st := m.store
+		ctx := m.ctx
 		id := m.deleteID
 		var cmd tea.Cmd
 		switch m.deleteTab {
-		case 0:
-			cmd = writeCmd("Delete site", func() error { return st.DeleteSite(context.Background(), id) })
+		case tabMonitors:
+			cmd = writeCmd("Delete site", func() error { return st.DeleteSite(ctx, id) })
 			m.engine.RemoveSite(id)
 			m.adjustCursor(len(m.sites) - 1)
-		case 1:
-			cmd = writeCmd("Delete alert", func() error { return st.DeleteAlert(context.Background(), id) })
-			m.adjustCursor(len(m.alerts) - 1)
-		case 4:
-			cmd = writeCmd("Delete maintenance window", func() error { return st.DeleteMaintenanceWindow(context.Background(), id) })
+		case tabMaint:
+			cmd = writeCmd("Delete maintenance window", func() error { return st.DeleteMaintenanceWindow(ctx, id) })
 			m.adjustCursor(len(m.maintenanceWindows) - 1)
-		case 5:
-			cmd = writeCmd("Delete user", func() error { return st.DeleteUser(context.Background(), id) })
-			m.adjustCursor(len(m.users) - 1)
+		case tabSettings:
+			switch m.settingsSection {
+			case sectionAlerts:
+				cmd = writeCmd("Delete alert", func() error { return st.DeleteAlert(ctx, id) })
+				m.adjustCursor(len(m.alerts) - 1)
+			case sectionUsers:
+				cmd = writeCmd("Delete user", func() error { return st.DeleteUser(ctx, id) })
+				m.adjustCursor(len(m.users) - 1)
+			}
 		}
 		m.refreshLive()
 		m.state = stateDashboard
-		if m.deleteTab == 5 {
-			m.state = stateUsers
-		}
 		return m, cmd
 	case "n", "N", "esc":
 		m.state = stateDashboard
-		if m.deleteTab == 5 {
-			m.state = stateUsers
-		}
 	case "ctrl+c":
 		return m, tea.Quit
 	}
@@ -117,9 +118,6 @@ func (m *Model) handleFormMsg(msg tea.Msg) (tea.Model, tea.Cmd) {
 		if keyMsg.String() == "esc" {
 			m.huhForm = nil
 			m.state = stateDashboard
-			if m.currentTab == 5 {
-				m.state = stateUsers
-			}
 			return m, nil
 		}
 	}
@@ -148,11 +146,16 @@ func (m *Model) handleFormMsg(msg tea.Msg) (tea.Model, tea.Cmd) {
 	return m, nil
 }
 
+const detailInlineHeight = 12
+
 func (m *Model) recalcLayout() {
 	chrome := chromeBase
 	if m.filterMode || m.filterText != "" {
 		chrome++
 	}
+	if m.detailOpen && m.currentTab == tabMonitors {
+		chrome += detailInlineHeight
+	}
 	m.maxTableRows = m.termHeight - chrome
 	if m.maxTableRows < 1 {
 		m.maxTableRows = 1
@@ -232,14 +235,49 @@ func (m *Model) handleTabData(msg tabDataMsg) (tea.Model, tea.Cmd) {
 // surfaces through the engine log (picked up by the next refreshLive).
 func (m *Model) testAlertCmd(id int, name string) tea.Cmd {
 	eng := m.engine
+	ctx := m.ctx
 	return func() tea.Msg {
-		if err := eng.TestAlert(id); err != nil {
+		if err := eng.TestAlert(ctx, id); err != nil {
 			eng.AddLog(fmt.Sprintf("Test alert failed (%s): %v", name, err))
 		}
 		return nil
 	}
 }
 
+func (m *Model) handleLogsFullscreen(msg tea.Msg) (tea.Model, tea.Cmd) {
+	switch msg := msg.(type) {
+	case tea.KeyMsg:
+		switch msg.String() {
+		case "esc", "q":
+			m.state = stateDashboard
+			m.focusedPanel = panelLogs
+		case "ctrl+c":
+			return m, tea.Quit
+		case "f":
+			m.logFilterImportant = !m.logFilterImportant
+			m.refreshLogContent()
+		case "up", "k":
+			m.logViewport.ScrollUp(1)
+		case "down", "j":
+			m.logViewport.ScrollDown(1)
+		case "pgup":
+			m.logViewport.ScrollUp(m.logViewport.Height)
+		case "pgdown":
+			m.logViewport.ScrollDown(m.logViewport.Height)
+		}
+	case tea.MouseMsg:
+		switch msg.Button {
+		case tea.MouseButtonWheelUp:
+			m.logViewport.ScrollUp(3)
+		case tea.MouseButtonWheelDown:
+			m.logViewport.ScrollDown(3)
+		}
+	case tickMsg:
+		m.refreshLogContent()
+	}
+	return m, nil
+}
+
 func (m *Model) handleMouse(msg tea.MouseMsg) (tea.Model, tea.Cmd) {
 	if m.state == stateHistory {
 		switch msg.Button {
@@ -265,7 +303,7 @@ func (m *Model) handleMouse(msg tea.MouseMsg) (tea.Model, tea.Cmd) {
 		}
 		return m, nil
 	}
-	if m.state != stateDashboard && m.state != stateLogs && m.state != stateUsers {
+	if m.state != stateDashboard {
 		return m, nil
 	}
 	if msg.Action == tea.MouseActionPress && msg.Button == tea.MouseButtonLeft {
@@ -275,11 +313,11 @@ func (m *Model) handleMouse(msg tea.MouseMsg) (tea.Model, tea.Cmd) {
 		return m, nil
 	}
 
-	if m.state == stateLogs {
+	if m.currentTab == tabMonitors && m.focusedPanel == panelLogs {
 		if msg.Button == tea.MouseButtonWheelUp {
-			m.logViewport.ScrollUp(3)
+			m.scrollLogs(-3)
 		} else {
-			m.logViewport.ScrollDown(3)
+			m.scrollLogs(3)
 		}
 		return m, nil
 	}
@@ -301,6 +339,9 @@ func (m *Model) handleMouse(msg tea.MouseMsg) (tea.Model, tea.Cmd) {
 		}
 	}
 	m.syncSelectedID()
+	if m.detailOpen && m.currentTab == tabMonitors && m.cursor < len(m.sites) {
+		return m, m.loadDetailCmd(m.sites[m.cursor].ID)
+	}
 	return m, nil
 }
 
@@ -325,7 +366,7 @@ func (m *Model) handleKey(msg tea.KeyMsg) (tea.Model, tea.Cmd) {
 		return m.handleSLAKey(msg)
 	case stateAlertDetail:
 		return m.handleAlertDetailKey(msg)
-	case stateDashboard, stateLogs, stateUsers:
+	case stateDashboard:
 		return m.handleDashboardKey(msg)
 	}
 	return m, nil
@@ -365,6 +406,18 @@ func (m *Model) handleFilterKey(msg tea.KeyMsg) (tea.Model, tea.Cmd) {
 
 func (m *Model) handleDetailKey(msg tea.KeyMsg) (tea.Model, tea.Cmd) {
 	switch msg.String() {
+	case "up", "k":
+		m.detailViewport.ScrollUp(1)
+		return m, nil
+	case "down", "j":
+		m.detailViewport.ScrollDown(1)
+		return m, nil
+	case "pgup":
+		m.detailViewport.ScrollUp(m.detailViewport.Height / 2)
+		return m, nil
+	case "pgdown":
+		m.detailViewport.ScrollDown(m.detailViewport.Height / 2)
+		return m, nil
 	case "esc":
 		if m.sparkTooltipIdx >= 0 {
 			m.sparkTooltipIdx = -1
@@ -521,100 +574,200 @@ func (m *Model) handleAlertDetailKey(msg tea.KeyMsg) (tea.Model, tea.Cmd) {
 }
 
 func (m *Model) handleDashboardKey(msg tea.KeyMsg) (tea.Model, tea.Cmd) {
-	var cmd tea.Cmd
 	switch msg.String() {
 	case "q":
 		return m, tea.Quit
 	case "/":
-		if m.currentTab == 0 {
+		if m.currentTab == tabMonitors {
 			m.filterMode = true
 			m.recalcLayout()
 			return m, nil
 		}
-	case "f":
-		if m.state == stateLogs {
-			m.logFilterImportant = !m.logFilterImportant
-			m.refreshLogContent()
-			return m, nil
+	case ">", ".":
+		if m.currentTab == tabMonitors {
+			m.sortColumn = (m.sortColumn + 1) % sortMax
+			m.sortAsc = false
+			m.refreshLive()
+		}
+	case "<", ",":
+		if m.currentTab == tabMonitors {
+			m.sortColumn = (m.sortColumn - 1 + sortMax) % sortMax
+			m.sortAsc = false
+			m.refreshLive()
+		}
+	case "r":
+		if m.currentTab == tabMonitors {
+			m.sortAsc = !m.sortAsc
+			m.refreshLive()
 		}
 	case "tab":
 		m.switchTab(m.currentTab + 1)
-	case "pgup", "pgdown":
-		if m.state == stateLogs {
-			m.logViewport, cmd = m.logViewport.Update(msg)
-			return m, cmd
+	case "left":
+		if m.currentTab == tabSettings {
+			m.switchSettingsSection(m.settingsSection - 1)
+		}
+	case "right":
+		if m.currentTab == tabSettings {
+			m.switchSettingsSection(m.settingsSection + 1)
+		}
+	case "l":
+		switch m.currentTab {
+		case tabSettings:
+			m.switchSettingsSection(m.settingsSection + 1)
+		case tabMonitors:
+			if m.focusedPanel == panelLogs {
+				m.logsOpen = false
+				m.focusedPanel = panelMonitors
+			} else {
+				m.logsOpen = true
+				m.focusedPanel = panelLogs
+			}
+			m.recalcLayout()
 		}
 	case "up", "k":
-		if m.state == stateLogs {
-			m.logViewport.ScrollUp(1)
-		} else if m.cursor > 0 {
+		if m.currentTab == tabMonitors && m.focusedPanel == panelLogs {
+			m.scrollLogs(-1)
+			return m, nil
+		}
+		if m.cursor > 0 {
 			m.cursor--
 			if m.cursor < m.tableOffset {
 				m.tableOffset = m.cursor
 			}
 			m.syncSelectedID()
+			if m.detailOpen && m.currentTab == tabMonitors && m.cursor < len(m.sites) {
+				return m, m.loadDetailCmd(m.sites[m.cursor].ID)
+			}
 		}
 	case "down", "j":
-		if m.state == stateLogs {
-			m.logViewport.ScrollDown(1)
-		} else {
-			max := m.currentListLen() - 1
-			if m.cursor < max {
-				m.cursor++
-				if m.cursor >= m.tableOffset+m.maxTableRows {
-					m.tableOffset++
-				}
-				m.syncSelectedID()
+		if m.currentTab == tabMonitors && m.focusedPanel == panelLogs {
+			m.scrollLogs(1)
+			return m, nil
+		}
+		max := m.currentListLen() - 1
+		if m.cursor < max {
+			m.cursor++
+			if m.cursor >= m.tableOffset+m.maxTableRows {
+				m.tableOffset++
+			}
+			m.syncSelectedID()
+			if m.detailOpen && m.currentTab == tabMonitors && m.cursor < len(m.sites) {
+				return m, m.loadDetailCmd(m.sites[m.cursor].ID)
 			}
 		}
 	case "n":
 		return m.handleNewItem()
-	case "e", "enter":
+	case "enter":
+		if m.currentTab == tabMonitors && m.focusedPanel == panelLogs {
+			m.refreshLogContent()
+			m.logViewport.GotoTop()
+			m.state = stateLogs
+			return m, nil
+		}
+		if m.currentTab == tabMonitors && len(m.sites) > 0 {
+			m.state = stateDetail
+			return m, m.loadDetailCmd(m.sites[m.cursor].ID)
+		}
+		return m.handleEditItem()
+	case "e":
 		return m.handleEditItem()
 	case "t":
-		if m.currentTab == 1 && len(m.alerts) > 0 {
+		if m.currentTab == tabSettings && m.settingsSection == sectionAlerts && len(m.alerts) > 0 {
 			a := m.alerts[m.cursor]
 			return m, m.testAlertCmd(a.ID, a.Name)
 		}
 	case " ":
-		if m.currentTab == 0 && len(m.sites) > 0 && m.sites[m.cursor].Type == "group" {
+		if m.currentTab == tabMonitors && len(m.sites) > 0 && m.sites[m.cursor].Type == "group" {
 			gid := m.sites[m.cursor].ID
 			m.collapsed[gid] = !m.collapsed[gid]
 			payload := collapsedJSON(m.collapsed)
 			st := m.store
+			ctx := m.ctx
 			m.refreshLive()
 			return m, writeCmd("Save collapsed groups", func() error {
-				return st.SetPreference(context.Background(), "collapsed_groups", payload)
+				return st.SetPreference(ctx, "collapsed_groups", payload)
 			})
 		}
 	case "p":
-		if m.currentTab == 0 && len(m.sites) > 0 {
+		if m.currentTab == tabMonitors && len(m.sites) > 0 {
 			id := m.sites[m.cursor].ID
 			paused := m.engine.ToggleSitePause(id)
 			st := m.store
+			ctx := m.ctx
 			m.refreshLive()
 			return m, writeCmd("Update pause state", func() error {
-				return st.UpdateSitePaused(context.Background(), id, paused)
+				return st.UpdateSitePaused(ctx, id, paused)
 			})
 		}
 	case "i":
-		if m.currentTab == 0 && len(m.sites) > 0 {
-			m.state = stateDetail
-			return m, m.loadDetailCmd(m.sites[m.cursor].ID)
-		} else if m.currentTab == 1 && len(m.alerts) > 0 {
+		if m.currentTab == tabMonitors && len(m.sites) > 0 {
+			m.detailOpen = !m.detailOpen
+			m.recalcLayout()
+			st := m.store
+			ctx := m.ctx
+			open := m.detailOpen
+			var cmd tea.Cmd
+			if m.detailOpen {
+				cmd = m.loadDetailCmd(m.sites[m.cursor].ID)
+			}
+			saveCmd := writeCmd("Save detail preference", func() error {
+				v := "false"
+				if open {
+					v = "true"
+				}
+				return st.SetPreference(ctx, "detail_open", v)
+			})
+			if cmd != nil {
+				return m, tea.Batch(cmd, saveCmd)
+			}
+			return m, saveCmd
+		} else if m.currentTab == tabSettings && m.settingsSection == sectionAlerts && len(m.alerts) > 0 {
 			m.state = stateAlertDetail
 		}
+	case "esc":
+		if m.currentTab == tabMonitors {
+			if m.focusedPanel != panelMonitors {
+				m.focusedPanel = panelMonitors
+			} else if m.detailOpen {
+				m.detailOpen = false
+				m.recalcLayout()
+				st := m.store
+				ctx := m.ctx
+				return m, writeCmd("Save detail preference", func() error {
+					return st.SetPreference(ctx, "detail_open", "false")
+				})
+			}
+		}
+	case "h":
+		if m.detailOpen && m.currentTab == tabMonitors && m.cursor < len(m.sites) {
+			site := m.sites[m.cursor]
+			m.historySiteName = site.Name
+			m.historySiteID = site.ID
+			m.historyChanges = nil
+			m.historyViewport = viewport.New(
+				m.termWidth-chromePadH,
+				m.termHeight-10,
+			)
+			m.historyViewport.SetContent("\n  Loading state history...")
+			m.state = stateHistory
+			return m, m.loadHistoryCmd(site.ID)
+		}
+	case "s":
+		if m.detailOpen && m.currentTab == tabMonitors && m.cursor < len(m.sites) {
+			return m, m.openSLAView(m.sites[m.cursor])
+		}
 	case "x":
-		if m.currentTab == 4 && len(m.maintenanceWindows) > 0 {
+		if m.currentTab == tabMaint && len(m.maintenanceWindows) > 0 {
 			mw := m.maintenanceWindows[m.cursor]
 			now := time.Now()
 			isActive := !mw.StartTime.After(now) && (mw.EndTime.IsZero() || mw.EndTime.After(now))
 			if isActive {
 				st := m.store
+				ctx := m.ctx
 				id := mw.ID
 				m.refreshLive()
 				return m, writeCmd("End maintenance", func() error {
-					return st.EndMaintenanceWindow(context.Background(), id)
+					return st.EndMaintenanceWindow(ctx, id)
 				})
 			}
 		}
@@ -623,9 +776,10 @@ func (m *Model) handleDashboardKey(msg tea.KeyMsg) (tea.Model, tea.Cmd) {
 		m.theme = themes[m.themeIndex]
 		m.st = newStyles(m.theme)
 		st := m.store
+		ctx := m.ctx
 		name := m.theme.Name
 		return m, writeCmd("Save theme", func() error {
-			return st.SetPreference(context.Background(), "theme", name)
+			return st.SetPreference(ctx, "theme", name)
 		})
 	case "d":
 		return m.handleDeleteItem()
@@ -637,19 +791,22 @@ func (m *Model) handleNewItem() (tea.Model, tea.Cmd) {
 	m.editID = 0
 	m.editToken = ""
 	switch m.currentTab {
-	case 0:
+	case tabMonitors:
 		m.state = stateFormSite
 		return m, m.initSiteHuhForm()
-	case 1:
-		m.state = stateFormAlert
-		return m, m.initAlertHuhForm()
-	case 4:
+	case tabMaint:
 		m.state = stateFormMaint
 		return m, m.initMaintHuhForm()
-	case 5:
-		if m.isAdmin {
-			m.state = stateFormUser
-			return m, m.initUserHuhForm()
+	case tabSettings:
+		switch m.settingsSection {
+		case sectionAlerts:
+			m.state = stateFormAlert
+			return m, m.initAlertHuhForm()
+		case sectionUsers:
+			if m.isAdmin {
+				m.state = stateFormUser
+				return m, m.initUserHuhForm()
+			}
 		}
 	}
 	return m, nil
@@ -657,24 +814,27 @@ func (m *Model) handleNewItem() (tea.Model, tea.Cmd) {
 
 func (m *Model) handleEditItem() (tea.Model, tea.Cmd) {
 	switch m.currentTab {
-	case 0:
+	case tabMonitors:
 		if len(m.sites) > 0 {
 			m.editID = m.sites[m.cursor].ID
 			m.editToken = m.sites[m.cursor].Token
 			m.state = stateFormSite
 			return m, m.initSiteHuhForm()
 		}
-	case 1:
-		if len(m.alerts) > 0 {
-			m.editID = m.alerts[m.cursor].ID
-			m.state = stateFormAlert
-			return m, m.initAlertHuhForm()
-		}
-	case 5:
-		if m.isAdmin && len(m.users) > 0 {
-			m.editID = m.users[m.cursor].ID
-			m.state = stateFormUser
-			return m, m.initUserHuhForm()
+	case tabSettings:
+		switch m.settingsSection {
+		case sectionAlerts:
+			if len(m.alerts) > 0 {
+				m.editID = m.alerts[m.cursor].ID
+				m.state = stateFormAlert
+				return m, m.initAlertHuhForm()
+			}
+		case sectionUsers:
+			if m.isAdmin && len(m.users) > 0 {
+				m.editID = m.users[m.cursor].ID
+				m.state = stateFormUser
+				return m, m.initUserHuhForm()
+			}
 		}
 	}
 	return m, nil
@@ -682,43 +842,43 @@ func (m *Model) handleEditItem() (tea.Model, tea.Cmd) {
 
 func (m *Model) handleDeleteItem() (tea.Model, tea.Cmd) {
 	switch m.currentTab {
-	case 0:
+	case tabMonitors:
 		if len(m.sites) > 0 {
 			m.deleteID = m.sites[m.cursor].ID
 			m.deleteName = m.sites[m.cursor].Name
-			m.deleteTab = 0
+			m.deleteTab = tabMonitors
 			m.state = stateConfirmDelete
 		}
-	case 1:
-		if len(m.alerts) > 0 {
-			m.deleteID = m.alerts[m.cursor].ID
-			m.deleteName = m.alerts[m.cursor].Name
-			m.deleteTab = 1
-			m.state = stateConfirmDelete
-		}
-	case 4:
+	case tabMaint:
 		if len(m.maintenanceWindows) > 0 {
 			m.deleteID = m.maintenanceWindows[m.cursor].ID
 			m.deleteName = m.maintenanceWindows[m.cursor].Title
-			m.deleteTab = 4
+			m.deleteTab = tabMaint
 			m.state = stateConfirmDelete
 		}
-	case 5:
-		if m.isAdmin && len(m.users) > 0 {
-			m.deleteID = m.users[m.cursor].ID
-			m.deleteName = m.users[m.cursor].Username
-			m.deleteTab = 5
-			m.state = stateConfirmDelete
+	case tabSettings:
+		switch m.settingsSection {
+		case sectionAlerts:
+			if len(m.alerts) > 0 {
+				m.deleteID = m.alerts[m.cursor].ID
+				m.deleteName = m.alerts[m.cursor].Name
+				m.deleteTab = tabSettings
+				m.state = stateConfirmDelete
+			}
+		case sectionUsers:
+			if m.isAdmin && len(m.users) > 0 {
+				m.deleteID = m.users[m.cursor].ID
+				m.deleteName = m.users[m.cursor].Username
+				m.deleteTab = tabSettings
+				m.state = stateConfirmDelete
+			}
 		}
 	}
 	return m, nil
 }
 
 func (m *Model) handleClick(msg tea.MouseMsg) (tea.Model, tea.Cmd) {
-	tabCount := 5
-	if m.isAdmin {
-		tabCount = 6
-	}
+	tabCount := tabSettings + 1
 	for i := 0; i < tabCount; i++ {
 		if m.zones.Get(fmt.Sprintf("tab-%d", i)).InBounds(msg) {
 			m.switchTab(i)
@@ -726,6 +886,58 @@ func (m *Model) handleClick(msg tea.MouseMsg) (tea.Model, tea.Cmd) {
 		}
 	}
 
+	if m.currentTab == tabSettings {
+		maxSec := 1
+		if m.isAdmin {
+			maxSec = 2
+		}
+		for i := 0; i <= maxSec; i++ {
+			if m.zones.Get(fmt.Sprintf("section-%d", i)).InBounds(msg) {
+				m.switchSettingsSection(i)
+				return m, nil
+			}
+		}
+	}
+
+	if m.currentTab == tabMonitors {
+		sortZones := []struct {
+			zone string
+			col  int
+		}{
+			{"sort-status", sortStatus},
+			{"sort-name", sortName},
+			{"sort-latency", sortLatency},
+		}
+		for _, sz := range sortZones {
+			if m.zones.Get(sz.zone).InBounds(msg) {
+				if m.sortColumn == sz.col {
+					m.sortAsc = !m.sortAsc
+				} else {
+					m.sortColumn = sz.col
+					m.sortAsc = false
+				}
+				m.refreshLive()
+				return m, nil
+			}
+		}
+
+		if m.zones.Get("panel-monitors").InBounds(msg) {
+			m.focusedPanel = panelMonitors
+		} else if m.zones.Get("panel-logs").InBounds(msg) {
+			if m.focusedPanel == panelLogs {
+				m.logsOpen = false
+				m.focusedPanel = panelMonitors
+				m.recalcLayout()
+			} else {
+				m.focusedPanel = panelLogs
+			}
+			return m, nil
+		} else if m.detailOpen && m.zones.Get("panel-detail").InBounds(msg) {
+			m.focusedPanel = panelDetail
+			return m, nil
+		}
+	}
+
 	prefix, listLen := m.currentZonePrefix()
 	end := m.tableOffset + m.maxTableRows
 	if end > listLen {
@@ -735,6 +947,9 @@ func (m *Model) handleClick(msg tea.MouseMsg) (tea.Model, tea.Cmd) {
 		if m.zones.Get(fmt.Sprintf("%s-%d", prefix, i)).InBounds(msg) {
 			m.cursor = i
 			m.syncSelectedID()
+			if m.detailOpen && m.currentTab == tabMonitors {
+				return m, m.loadDetailCmd(m.sites[m.cursor].ID)
+			}
 			return m, nil
 		}
 	}
@@ -743,24 +958,14 @@ func (m *Model) handleClick(msg tea.MouseMsg) (tea.Model, tea.Cmd) {
 }
 
 func (m *Model) switchTab(idx int) {
-	maxTabs := 4
-	if m.isAdmin {
-		maxTabs = 5
-	}
+	maxTabs := tabSettings
 	if idx > maxTabs {
 		idx = 0
 	}
 	m.currentTab = idx
 	m.cursor = 0
 	m.tableOffset = 0
-	switch idx {
-	case 2:
-		m.state = stateLogs
-	case 5:
-		m.state = stateUsers
-	default:
-		m.state = stateDashboard
-	}
+	m.state = stateDashboard
 }
 
 func (m *Model) adjustCursor(_ int) {
@@ -791,30 +996,36 @@ func (m *Model) submitForm() tea.Cmd {
 
 func (m Model) currentListLen() int {
 	switch m.currentTab {
-	case 1:
-		return len(m.alerts)
-	case 3:
-		return len(m.nodes)
-	case 4:
-		return len(m.maintenanceWindows)
-	case 5:
-		return len(m.users)
-	default:
+	case tabMonitors:
 		return len(m.sites)
+	case tabMaint:
+		return len(m.maintenanceWindows)
+	case tabSettings:
+		switch m.settingsSection {
+		case sectionAlerts:
+			return len(m.alerts)
+		case sectionNodes:
+			return len(m.nodes)
+		case sectionUsers:
+			return len(m.users)
+		}
 	}
+	return 0
 }
 
 func (m Model) currentZonePrefix() (string, int) {
 	switch m.currentTab {
-	case 0:
+	case tabMonitors:
 		return "site", len(m.sites)
-	case 1:
-		return "alert", len(m.alerts)
-	case 4:
+	case tabMaint:
 		return "maint", len(m.maintenanceWindows)
-	case 5:
-		return "user", len(m.users)
-	default:
-		return "site", 0
+	case tabSettings:
+		switch m.settingsSection {
+		case sectionAlerts:
+			return "alert", len(m.alerts)
+		case sectionUsers:
+			return "user", len(m.users)
+		}
 	}
+	return "site", 0
 }

internal/tui/update_test.go

@@ -300,7 +300,7 @@ func TestWriteDoneMsg_LogsErrorAndReloads(t *testing.T) {
 	mm := updated.(Model)
 	found := false
 	for _, line := range mm.engine.GetLogs() {
-		if strings.Contains(line, "Delete site failed: boom") {
+		if strings.Contains(line.Message, "Delete site failed: boom") {
 			found = true
 		}
 	}

internal/tui/uptime_timeline.go

@@ -0,0 +1,50 @@
+package tui
+
+import (
+	"fmt"
+	"strings"
+
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/monitor"
+)
+
+func (m Model) uptimeTimeline(days []monitor.DayReport, width int) string {
+	if len(days) == 0 {
+		return m.st.subtleStyle.Render("No uptime data")
+	}
+
+	maxDays := width - 10
+	if maxDays < 10 {
+		maxDays = 10
+	}
+
+	display := days
+	if len(display) > maxDays {
+		display = display[len(display)-maxDays:]
+	}
+
+	var sb strings.Builder
+	for _, d := range display {
+		ch := "█"
+		switch {
+		case d.UptimePct >= uptimeGoodPct:
+			sb.WriteString(m.st.specialStyle.Render(ch))
+		case d.UptimePct >= uptimeWarnPct:
+			sb.WriteString(m.st.warnStyle.Render(ch))
+		case d.UptimePct > 0:
+			sb.WriteString(m.st.dangerStyle.Render(ch))
+		default:
+			sb.WriteString(m.st.subtleStyle.Render("░"))
+		}
+	}
+
+	pct := days[len(days)-1].UptimePct
+	pctStyle := m.st.specialStyle
+	if pct < uptimeGoodPct {
+		pctStyle = m.st.dangerStyle
+	} else if pct < uptimeExcellentPct {
+		pctStyle = m.st.warnStyle
+	}
+	sb.WriteString("  " + pctStyle.Render(fmt.Sprintf("%.2f%%", pct)))
+
+	return sb.String()
+}

internal/tui/view_dashboard.go

@@ -92,6 +92,8 @@ func (m Model) View() string {
 		return ""
 	case stateDetail:
 		return m.zones.Scan(m.viewDetailPanel())
+	case stateLogs:
+		return m.viewLogsFullscreen()
 	case stateHistory:
 		return m.viewHistoryPanel()
 	case stateSLA:
@@ -130,7 +132,7 @@ func (m Model) computeStats() dashboardStats {
 		}
 	}
 	for _, n := range m.nodes {
-		if !n.LastSeen.IsZero() && time.Since(n.LastSeen) > 5*time.Minute {
+		if !n.LastSeen.IsZero() && time.Since(n.LastSeen) > nodeStaleThreshold {
 			s.offlineNodes++
 		}
 	}
@@ -151,20 +153,52 @@ func (m Model) viewDashboard() string {
 
 	var content string
 	switch m.currentTab {
-	case 0:
-		content = m.viewSitesTab()
-	case 1:
-		content = m.viewAlertsTab()
-	case 2:
-		content = m.viewLogsTab()
-	case 3:
-		content = m.viewNodesTab()
-	case 4:
-		content = m.viewMaintTab()
-	case 5:
-		if m.isAdmin {
-			content = m.viewUsersTab()
+	case tabMonitors:
+		showSidebar := m.termWidth >= wideBreakpoint && m.logsOpen
+		if showSidebar {
+			availW := m.termWidth - chromePadH
+			leftW := availW * 70 / 100
+			rightW := availW - leftW
+			m.contentWidth = leftW - 2
+			monitors := m.viewSitesTab()
+			monPanel := m.zones.Mark("panel-monitors", m.titledPanel("Monitors", monitors, leftW, m.focusedPanel == panelMonitors))
+			sidebarContent := m.viewLogsSidebar(rightW-2, m.maxTableRows)
+			logPanel := m.zones.Mark("panel-logs", m.titledPanel("Logs", sidebarContent, rightW, m.focusedPanel == panelLogs))
+			top := lipgloss.JoinHorizontal(lipgloss.Top, monPanel, logPanel)
+			if m.detailOpen {
+				site := ""
+				if m.cursor < len(m.sites) {
+					site = m.sites[m.cursor].Name
+				}
+				detail := m.viewDetailInline(availW - 2)
+				detailPanel := m.zones.Mark("panel-detail", m.titledPanel(site, detail, availW, m.focusedPanel == panelDetail))
+				content = top + "\n" + detailPanel
+			} else {
+				content = top
+			}
+		} else {
+			m.contentWidth = m.termWidth - 2
+			monitors := m.viewSitesTab()
+			availW := m.termWidth - chromePadH
+			monPanel := m.zones.Mark("panel-monitors", m.titledPanel("Monitors", monitors, availW, m.focusedPanel == panelMonitors))
+			if m.detailOpen {
+				site := ""
+				if m.cursor < len(m.sites) {
+					site = m.sites[m.cursor].Name
+				}
+				detail := m.viewDetailInline(availW - 2)
+				detailPanel := m.zones.Mark("panel-detail", m.titledPanel(site, detail, availW, m.focusedPanel == panelDetail))
+				content = monPanel + "\n" + detailPanel
+			} else {
+				content = monPanel
+			}
 		}
+	case tabMaint:
+		m.contentWidth = m.termWidth
+		content = m.viewMaintTab()
+	case tabSettings:
+		m.contentWidth = m.termWidth
+		content = m.viewSettingsTab()
 	}
 
 	content = strings.TrimSpace(content)
@@ -177,13 +211,19 @@ func (m Model) viewDashboard() string {
 		availHeight = 5
 	}
 
-	contentHeight := availHeight - lipgloss.Height(header) - lipgloss.Height(footer)
+	divW := m.termWidth - chromePadH
+	if divW < 40 {
+		divW = 40
+	}
+	tabDivider := m.st.subtleStyle.Render(strings.Repeat("─", divW))
+
+	contentHeight := availHeight - lipgloss.Height(header) - 1 - lipgloss.Height(footer)
 	if contentHeight < 1 {
 		contentHeight = 1
 	}
 	paddedContent := lipgloss.NewStyle().Height(contentHeight).MaxHeight(contentHeight).Render(content)
 
-	return outerPad.Render(lipgloss.JoinVertical(lipgloss.Top, header, paddedContent, footer))
+	return outerPad.Render(lipgloss.JoinVertical(lipgloss.Top, header, tabDivider, paddedContent, footer))
 }
 
 type tabEntry struct {
@@ -193,15 +233,21 @@ type tabEntry struct {
 }
 
 func (m Model) renderTabBar(stats dashboardStats) string {
-	tabs := []tabEntry{
-		{"Sites", stats.totalMonitors, stats.downCount + stats.lateCount},
-		{"Alerts", len(m.alerts), 0},
-		{"Logs", 0, 0},
-		{"Nodes", len(m.nodes), stats.offlineNodes},
-		{"Maint", len(m.maintenanceWindows), stats.activeMaint},
+	settingsCount := len(m.alerts) + len(m.nodes)
+	settingsWarn := stats.offlineNodes
+	for _, a := range m.alerts {
+		h := m.engine.GetAlertHealth(a.ID)
+		if !h.LastSendOK && !h.LastSendAt.IsZero() {
+			settingsWarn++
+		}
 	}
 	if m.isAdmin {
-		tabs = append(tabs, tabEntry{"Users", len(m.users), 0})
+		settingsCount += len(m.users)
+	}
+	tabs := []tabEntry{
+		{"Monitors", stats.totalMonitors, stats.downCount + stats.lateCount},
+		{"Maint", len(m.maintenanceWindows), stats.activeMaint},
+		{"Settings", settingsCount, settingsWarn},
 	}
 
 	countStyle := lipgloss.NewStyle().Foreground(m.theme.Muted)
@@ -250,7 +296,7 @@ func (m Model) renderFooter(stats dashboardStats) string {
 	if len(m.nodes) > 0 {
 		online := 0
 		for _, n := range m.nodes {
-			if !n.LastSeen.IsZero() && time.Since(n.LastSeen) < 60*time.Second {
+			if !n.LastSeen.IsZero() && time.Since(n.LastSeen) < nodeOnlineThreshold {
 				online++
 			}
 		}
@@ -264,24 +310,38 @@ func (m Model) renderFooter(stats dashboardStats) string {
 
 	var keys string
 	switch m.currentTab {
-	case 0:
-		keys = "[/]Filter [n]New [e]Edit [i]Info [d]Del [p]Pause [Space]Collapse [T]Theme [Tab]Switch [q]Quit"
-	case 1:
-		keys = "[n]New [e]Edit [i]Info [d]Del [t]Test [T]Theme [Tab]Switch [q]Quit"
-	case 2:
-		keys = "[↑/↓]Scroll  [PgUp/PgDn]Page  [f]Filter  [T]Theme  [Tab]Switch  [q]Quit"
-	case 4:
+	case tabMonitors:
+		if m.focusedPanel == panelLogs {
+			keys = "[↑/↓]Scroll [Enter]Expand [l/Esc]Back [T]Theme [q]Quit"
+		} else if m.detailOpen {
+			keys = "[i]Close [Enter]Expand [h]History [s]SLA [e]Edit [l]Logs [↑/↓]Select [T]Theme [q]Quit"
+		} else {
+			keys = "[/]Filter [i]Info [Enter]Detail [</>]Sort [r]Reverse [n]New [e]Edit [d]Del [l]Logs [T]Theme [Tab]Switch [q]Quit"
+		}
+	case tabMaint:
 		keys = "[n]New [x]End [d]Del [T]Theme [Tab]Switch [q]Quit"
-	case 5:
-		keys = "[n]Add [d]Revoke [T]Theme [Tab]Switch [q]Quit"
+	case tabSettings:
+		switch m.settingsSection {
+		case sectionAlerts:
+			keys = "[n]New [e]Edit [i]Info [d]Del [t]Test [←/→]Section [T]Theme [Tab]Switch [q]Quit"
+		case sectionUsers:
+			keys = "[n]Add [d]Revoke [←/→]Section [T]Theme [Tab]Switch [q]Quit"
+		default:
+			keys = "[←/→]Section [T]Theme [Tab]Switch [q]Quit"
+		}
 	default:
 		keys = "[T]Theme [Tab]Switch [q]Quit"
 	}
 
 	ver := m.st.subtleStyle.Render("v" + m.version)
-	footer := statusLine + "  " + m.st.subtleStyle.Render(keys) + "  " + ver
-	if m.filterText != "" && m.currentTab == 0 {
-		footer = m.st.subtleStyle.Render(fmt.Sprintf("filter: %s", m.filterText)) + "  " + statusLine + "  " + m.st.subtleStyle.Render(keys) + "  " + ver
+	line := statusLine + "  " + m.st.subtleStyle.Render(keys) + "  " + ver
+	if m.filterText != "" && m.currentTab == tabMonitors {
+		line = m.st.subtleStyle.Render(fmt.Sprintf("filter: %s", m.filterText)) + "  " + statusLine + "  " + m.st.subtleStyle.Render(keys) + "  " + ver
 	}
-	return footer
+
+	divW := m.termWidth - chromePadH
+	if divW < 40 {
+		divW = 40
+	}
+	return m.st.subtleStyle.Render(strings.Repeat("─", divW)) + "\n" + line
 }

internal/tui/view_detail.go

@@ -8,7 +8,6 @@ import (
 	"time"
 
 	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
-	"gitea.lerkolabs.com/lerkolabs/uptop/internal/monitor"
 	"github.com/charmbracelet/lipgloss"
 )
 
@@ -20,45 +19,120 @@ func (m Model) viewDetailPanel() string {
 	hist, _ := m.engine.GetHistory(site.ID)
 
 	var b strings.Builder
+	totalW := m.termWidth - chromePadH
 
 	var breadcrumb string
 	if site.ParentID > 0 {
 		for _, s := range m.sites {
 			if s.ID == site.ParentID {
-				breadcrumb = m.st.subtleStyle.Render("  Sites > "+s.Name+" > ") + m.st.titleStyle.Render(site.Name)
+				breadcrumb = m.st.subtleStyle.Render("  Monitors > "+s.Name+" > ") + m.st.titleStyle.Render(site.Name)
 				break
 			}
 		}
 	}
 	if breadcrumb == "" {
-		breadcrumb = m.st.subtleStyle.Render("  Sites > ") + m.st.titleStyle.Render(site.Name)
+		breadcrumb = m.st.subtleStyle.Render("  Monitors > ") + m.st.titleStyle.Render(site.Name)
 	}
 	b.WriteString(breadcrumb + "\n")
 	b.WriteString(m.divider() + "\n")
 
-	row := func(label, value string) {
-		fmt.Fprintf(&b, "  %-16s %s\n", m.st.subtleStyle.Render(label), value)
+	// Two-column layout for key info
+	colW := (totalW - 4) / 2
+	if colW < 30 {
+		colW = 30
 	}
 
-	section := func(label string) {
-		b.WriteString("\n" + m.st.subtleStyle.Render("  "+label) + "\n")
+	row := func(label, value string) string {
+		return fmt.Sprintf("  %-16s %s", m.st.subtleStyle.Render(label), value)
 	}
 
-	row("Status", m.fmtStatus(site.Status, site.Paused, m.isMonitorInMaintenance(site.ID)))
+	divW := totalW - 4
+	if divW < 20 {
+		divW = 20
+	}
+	sectionDiv := m.st.subtleStyle.Render(strings.Repeat("─", divW))
+	sectionHead := func(title string) string {
+		return m.st.titleStyle.Render("  "+title) + " " + m.st.subtleStyle.Render(strings.Repeat("─", divW-len(title)-3))
+	}
+
+	// Left column: endpoint details
+	var left []string
+	left = append(left, m.st.titleStyle.Render("  ENDPOINT"))
+	left = append(left, row("Type", site.Type))
+	if site.URL != "" {
+		left = append(left, row("URL", limitStr(site.URL, colW-19)))
+	}
+	if site.Hostname != "" {
+		left = append(left, row("Host", site.Hostname))
+	}
+	if site.Port > 0 {
+		left = append(left, row("Port", strconv.Itoa(site.Port)))
+	}
+	left = append(left, row("Interval", fmt.Sprintf("%ds", site.Interval)))
+	if site.MaxRetries > 0 {
+		left = append(left, row("Retries", m.fmtRetries(site)))
+	}
+	if site.Regions != "" {
+		left = append(left, row("Regions", site.Regions))
+	}
+	if site.Description != "" {
+		left = append(left, row("Description", limitStr(site.Description, colW-19)))
+	}
+
+	// Right column: status + timing + HTTP
+	var right []string
+	right = append(right, m.st.titleStyle.Render("  STATUS"))
+	right = append(right, row("Status", m.fmtStatus(site.Status, site.Paused, m.isMonitorInMaintenance(site.ID))))
+	right = append(right, row("Latency", m.fmtLatency(site.Latency)))
+	right = append(right, row("Uptime", m.fmtUptime(hist.Statuses)))
+	if !site.StatusChangedAt.IsZero() {
+		dur := time.Since(site.StatusChangedAt)
+		right = append(right, row("State Since", fmtDuration(dur)+" ago"))
+	}
+	if !site.LastCheck.IsZero() {
+		right = append(right, row("Last Check", m.fmtTimeAgo(site.LastCheck)))
+	}
+	if !site.LastSuccessAt.IsZero() {
+		right = append(right, row("Last Success", m.fmtTimeAgo(site.LastSuccessAt)))
+	}
 
 	if (site.Status == models.StatusDown || site.Status == models.StatusSSLExp || site.Status == models.StatusLate || site.Status == models.StatusStale) && site.LastError != "" {
-		errWidth := m.termWidth - chromePadH - 19
-		if errWidth < 30 {
-			errWidth = 30
+		errW := colW - 19
+		if errW < 20 {
+			errW = 20
 		}
-		wrapped := lipgloss.NewStyle().Width(errWidth).Render(site.LastError)
-		row("Error", m.st.dangerStyle.Render(wrapped))
+		right = append(right, row("Error", m.st.dangerStyle.Render(limitStr(site.LastError, errW))))
 	}
 
-	if site.Type == "http" && site.StatusCode > 0 {
-		row("HTTP Code", strconv.Itoa(site.StatusCode))
+	if site.Type == "http" {
+		if site.StatusCode > 0 {
+			right = append(right, row("HTTP Code", strconv.Itoa(site.StatusCode)))
+		}
+		codes := site.AcceptedCodes
+		if codes == "" {
+			codes = "200-299"
+		}
+		right = append(right, row("Codes", codes))
+		right = append(right, row("SSL", m.fmtSSL(site)))
+		if site.Method != "" && site.Method != "GET" {
+			right = append(right, row("Method", site.Method))
+		}
 	}
 
+	// Pad shorter column
+	for len(left) < len(right) {
+		left = append(left, "")
+	}
+	for len(right) < len(left) {
+		right = append(right, "")
+	}
+
+	leftCol := lipgloss.NewStyle().Width(colW).Render(strings.Join(left, "\n"))
+	rightCol := lipgloss.NewStyle().Width(colW).Render(strings.Join(right, "\n"))
+	b.WriteString(lipgloss.JoinHorizontal(lipgloss.Top, leftCol, rightCol) + "\n")
+	b.WriteString("\n" + sectionDiv + "\n")
+
+	// Connection chain (full width, only on errors)
 	if (site.Status == models.StatusDown || site.Status == models.StatusSSLExp) && site.LastError != "" {
 		chain := connectionChain(site.LastError, site.Type, site.StatusCode, strings.HasPrefix(site.URL, "https"))
 		if len(chain) > 0 {
@@ -87,81 +161,22 @@ func (m Model) viewDetailPanel() string {
 		}
 	}
 
-	if !site.StatusChangedAt.IsZero() {
-		dur := time.Since(site.StatusChangedAt)
-		row("State Since", site.StatusChangedAt.Format("2006-01-02 15:04:05")+" ("+fmtDuration(dur)+")")
-	}
-
-	if !site.LastSuccessAt.IsZero() {
-		ago := time.Since(site.LastSuccessAt)
-		row("Last Success", site.LastSuccessAt.Format("15:04:05")+" ("+fmtDuration(ago)+" ago)")
-	}
-
+	// Maintenance
 	if m.isMonitorInMaintenance(site.ID) {
 		for _, mw := range m.maintenanceWindows {
 			if mw.Type == "maintenance" && (mw.MonitorID == 0 || mw.MonitorID == site.ID || mw.MonitorID == site.ParentID) {
-				row("Maintenance", m.st.maintStyle.Render(mw.Title))
+				fmt.Fprintf(&b, "  %-16s %s\n", m.st.subtleStyle.Render("Maintenance"), m.st.maintStyle.Render(mw.Title))
 				break
 			}
 		}
 	}
 
-	section("ENDPOINT")
-	row("Type", site.Type)
+	// Push token
 	if site.Type == "push" && site.Token != "" {
-		row("Token", site.Token)
-		row("Push", "curl -X POST -H 'Authorization: Bearer "+site.Token+"' <host>/api/push")
-	}
-	if site.URL != "" {
-		row("URL", site.URL)
-	}
-	if site.Hostname != "" {
-		row("Host", site.Hostname)
-	}
-	if site.Port > 0 {
-		row("Port", strconv.Itoa(site.Port))
-	}
-
-	section("TIMING")
-	row("Interval", fmt.Sprintf("%ds", site.Interval))
-	if site.Timeout > 0 {
-		row("Timeout", fmt.Sprintf("%ds", site.Timeout))
-	}
-	row("Latency", m.fmtLatency(site.Latency))
-	row("Uptime", m.fmtUptime(hist.Statuses))
-	if !site.LastCheck.IsZero() {
-		row("Last Check", m.fmtTimeAgo(site.LastCheck))
-	}
-
-	if site.Type == "http" {
-		section("HTTP")
-		if site.Method != "" && site.Method != "GET" {
-			row("Method", site.Method)
-		}
-		codes := site.AcceptedCodes
-		if codes == "" {
-			codes = "200-299"
-		}
-		row("Codes", codes)
-		row("SSL", m.fmtSSL(site))
-		if site.IgnoreTLS {
-			row("TLS Verify", m.st.dangerStyle.Render("disabled"))
-		}
-	}
-
-	if site.MaxRetries > 0 || site.Regions != "" || site.Description != "" {
-		section("CONFIG")
-		if site.MaxRetries > 0 {
-			row("Retries", m.fmtRetries(site))
-		}
-		if site.Regions != "" {
-			row("Regions", site.Regions)
-		}
-		if site.Description != "" {
-			row("Description", site.Description)
-		}
+		fmt.Fprintf(&b, "  %-16s %s\n", m.st.subtleStyle.Render("Token"), site.Token)
 	}
 
+	// Probe results
 	probeResults := m.engine.GetProbeResults(site.ID)
 	if len(probeResults) > 0 {
 		nodeIDs := make([]string, 0, len(probeResults))
@@ -169,7 +184,7 @@ func (m Model) viewDetailPanel() string {
 			nodeIDs = append(nodeIDs, id)
 		}
 		sort.Strings(nodeIDs)
-		b.WriteString("\n" + m.st.subtleStyle.Render("  PROBE RESULTS") + "\n")
+		b.WriteString("\n" + sectionHead("PROBE RESULTS") + "\n")
 		for _, nodeID := range nodeIDs {
 			result := probeResults[nodeID]
 			status := m.st.specialStyle.Render("UP")
@@ -186,36 +201,30 @@ func (m Model) viewDetailPanel() string {
 		}
 	}
 
-	// Loaded on panel-enter (loadDetailCmd) and cached, so View does no DB IO.
-	var stateChanges []models.StateChange
-	if m.detailChangesSiteID == site.ID {
-		stateChanges = m.detailChanges
+	// Bottom two-column: graphs left, state changes right
+	graphW := (totalW - 4) * 70 / 100
+	changeW := totalW - 4 - graphW
+	if graphW < 30 {
+		graphW = 30
 	}
-	if len(stateChanges) > 0 {
-		b.WriteString("\n" + m.st.subtleStyle.Render("  STATE CHANGES") + "\n")
-		for i, sc := range stateChanges {
-			ago := fmtDuration(time.Since(sc.ChangedAt))
-			arrow := m.st.subtleStyle.Render(sc.FromStatus) + " → "
-			if sc.ToStatus == string(models.StatusUp) {
-				arrow += m.st.specialStyle.Render(sc.ToStatus)
-			} else {
-				arrow += m.st.dangerStyle.Render(sc.ToStatus)
-			}
-			line := fmt.Sprintf("  %s  %s", arrow, m.st.subtleStyle.Render(ago+" ago"))
-			if dur := computeOutageDuration(stateChanges, i); dur > 0 {
-				line += "  " + m.st.warnStyle.Render("outage "+fmtDuration(dur))
-			}
-			if sc.ErrorReason != "" && sc.ToStatus != string(models.StatusUp) {
-				line += "  " + m.st.dangerStyle.Render(sc.ErrorReason)
-			}
-			b.WriteString(line + "\n")
-		}
-		b.WriteString("  " + m.st.subtleStyle.Render("[h] History") + "\n")
+	if changeW < 20 {
+		changeW = 20
+	}
+	bottomColW := graphW
+
+	// Left: latency + histogram
+	var graphLines []string
+	sectionLabel := func(title string) string {
+		return m.st.titleStyle.Render("  " + title)
 	}
 
-	b.WriteString(m.divider() + "\n")
+	graphLines = append(graphLines, sectionLabel("LATENCY"))
 	if site.Type == "push" {
-		b.WriteString("  " + m.zones.Mark("spark-heartbeat", m.heartbeatSparkline(hist.Statuses, detailSparkWidth, "")))
+		sparkW := bottomColW - 4
+		if sparkW > detailSparkWidth {
+			sparkW = detailSparkWidth
+		}
+		graphLines = append(graphLines, "  "+m.heartbeatSparkline(hist.Statuses, sparkW, nil))
 		if len(hist.Statuses) > 0 {
 			up := 0
 			for _, s := range hist.Statuses {
@@ -223,12 +232,15 @@ func (m Model) viewDetailPanel() string {
 					up++
 				}
 			}
-			fmt.Fprintf(&b, "\n  %s %d/%d checks up",
-				m.st.subtleStyle.Render("Heartbeats"),
-				up, len(hist.Statuses))
+			graphLines = append(graphLines, fmt.Sprintf("  %s %d/%d checks up",
+				m.st.subtleStyle.Render("Heartbeats"), up, len(hist.Statuses)))
 		}
 	} else {
-		b.WriteString("  " + m.zones.Mark("spark-latency", m.latencySparkline(hist.Latencies, hist.Statuses, detailSparkWidth, "")))
+		sparkW := bottomColW - 4
+		if sparkW > detailSparkWidth {
+			sparkW = detailSparkWidth
+		}
+		graphLines = append(graphLines, "  "+m.latencySparkline(hist.Latencies, hist.Statuses, sparkW, nil))
 		var minL, maxL, total time.Duration
 		count := 0
 		for i, l := range hist.Latencies {
@@ -247,60 +259,76 @@ func (m Model) viewDetailPanel() string {
 		}
 		if count > 0 {
 			avg := total / time.Duration(count)
-			fmt.Fprintf(&b, "\n  %s %dms  %s %dms  %s %dms",
+			graphLines = append(graphLines, fmt.Sprintf("  %s %dms  %s %dms  %s %dms",
 				m.st.subtleStyle.Render("Min"), minL.Milliseconds(),
 				m.st.subtleStyle.Render("Avg"), avg.Milliseconds(),
-				m.st.subtleStyle.Render("Max"), maxL.Milliseconds())
+				m.st.subtleStyle.Render("Max"), maxL.Milliseconds()))
 		}
 	}
 
-	if m.sparkTooltipIdx >= 0 {
-		b.WriteString("\n" + m.renderSparkTooltip(site, hist, detailSparkWidth))
+	if site.Type != "push" && len(hist.Latencies) > 5 {
+		graphLines = append(graphLines, "")
+		graphLines = append(graphLines, sectionLabel("DISTRIBUTION"))
+		graphLines = append(graphLines, m.latencyHistogram(hist.Latencies, hist.Statuses, bottomColW))
 	}
 
+	// Right: state changes
+	var changeLines []string
+	var stateChanges []models.StateChange
+	if m.detailChangesSiteID == site.ID {
+		stateChanges = m.detailChanges
+	}
+	changeLines = append(changeLines, sectionLabel("STATE CHANGES"))
+	if len(stateChanges) > 0 {
+		for i, sc := range stateChanges {
+			from := m.fmtStatusWord(string(sc.FromStatus))
+			to := m.fmtStatusWord(string(sc.ToStatus))
+			ago := fmtDuration(time.Since(sc.ChangedAt))
+			line := fmt.Sprintf("  %s → %s  %s ago", from, to, ago)
+			if sc.ToStatus == "UP" {
+				dur := computeOutageDuration(stateChanges, i)
+				if dur > 0 {
+					line += "  " + m.st.warnStyle.Render("outage "+fmtDuration(dur))
+				}
+			}
+			if sc.ErrorReason != "" {
+				line += "  " + m.st.dangerStyle.Render(limitStr(sc.ErrorReason, changeW-30))
+			}
+			changeLines = append(changeLines, line)
+		}
+	} else {
+		changeLines = append(changeLines, m.st.subtleStyle.Render("  No state changes"))
+	}
+
+	// Pad and join
+	for len(graphLines) < len(changeLines) {
+		graphLines = append(graphLines, "")
+	}
+	for len(changeLines) < len(graphLines) {
+		changeLines = append(changeLines, "")
+	}
+
+	graphCol := lipgloss.NewStyle().Width(graphW).Render(strings.Join(graphLines, "\n"))
+	changeCol := lipgloss.NewStyle().Width(changeW).Render(strings.Join(changeLines, "\n"))
+	b.WriteString(lipgloss.JoinHorizontal(lipgloss.Top, graphCol, changeCol) + "\n")
+
 	b.WriteString("\n")
 	b.WriteString(m.divider() + "\n")
 	b.WriteString(m.st.subtleStyle.Render("  [q/Esc] Back  [e] Edit  [h] History  [s] SLA  [click] Inspect"))
 
-	return lipgloss.NewStyle().Padding(1, 2).Render(b.String())
-}
-
-func (m Model) renderSparkTooltip(site models.Site, hist monitor.SiteHistory, sparkWidth int) string {
-	idx := m.sparkTooltipIdx
-
-	var dataLen int
-	if site.Type == "push" {
-		dataLen = len(hist.Statuses)
-	} else {
-		dataLen = len(hist.Latencies)
-	}
-	if idx < 0 || idx >= dataLen {
-		return ""
-	}
-
-	var parts []string
-
-	checksAgo := dataLen - 1 - idx
-	approxSecs := checksAgo * site.Interval
-	if approxSecs == 0 {
-		parts = append(parts, "latest")
-	} else {
-		parts = append(parts, "~"+fmtDuration(time.Duration(approxSecs)*time.Second)+" ago")
-	}
-
-	if site.Type != "push" && idx < len(hist.Latencies) {
-		parts = append(parts, m.fmtLatency(hist.Latencies[idx]))
-	}
-
-	if idx < len(hist.Statuses) {
-		if hist.Statuses[idx] {
-			parts = append(parts, m.st.specialStyle.Render("UP"))
-		} else {
-			parts = append(parts, m.st.dangerStyle.Render("DOWN"))
-		}
-	}
-
-	sep := m.st.subtleStyle.Render(" | ")
-	pos := m.st.subtleStyle.Render(fmt.Sprintf("[%d/%d]", idx+1, dataLen))
-	return "  " + strings.Join(parts, sep) + "  " + pos
+	// Wrap in a viewport for scrolling
+	content := b.String()
+	contentH := m.termHeight - 4
+	if contentH < 10 {
+		contentH = 10
+	}
+	lines := strings.Split(content, "\n")
+	if len(lines) > contentH {
+		m.detailViewport.SetContent(content)
+		m.detailViewport.Width = totalW
+		m.detailViewport.Height = contentH
+		return lipgloss.NewStyle().Padding(1, 2).Render(m.detailViewport.View())
+	}
+
+	return lipgloss.NewStyle().Padding(1, 2).Render(content)
 }

internal/tui/view_detail_inline.go

@@ -0,0 +1,311 @@
+package tui
+
+import (
+	"fmt"
+	"strings"
+	"time"
+
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/models"
+	"gitea.lerkolabs.com/lerkolabs/uptop/internal/monitor"
+	"github.com/charmbracelet/lipgloss"
+)
+
+const detailTwoColMinWidth = 80
+
+func (m Model) viewDetailInline(width int) string {
+	if m.cursor >= len(m.sites) {
+		return ""
+	}
+	site := m.sites[m.cursor]
+	hist, _ := m.engine.GetHistory(site.ID)
+
+	if width < detailTwoColMinWidth {
+		return m.viewDetailSingleCol(site, hist, width)
+	}
+	return m.viewDetailTwoCol(site, hist, width)
+}
+
+func (m Model) viewDetailTwoCol(site models.Site, hist monitor.SiteHistory, width int) string {
+	leftW := width * 55 / 100
+	rightW := width - leftW - 3 // 3 for " │ " divider
+
+	left := m.detailLeftCol(site, hist, leftW)
+	right := m.detailRightCol(site, hist, rightW)
+
+	leftLines := strings.Split(left, "\n")
+	rightLines := strings.Split(right, "\n")
+
+	lineCount := len(leftLines)
+	if len(rightLines) > lineCount {
+		lineCount = len(rightLines)
+	}
+	for len(leftLines) < lineCount {
+		leftLines = append(leftLines, "")
+	}
+	for len(rightLines) < lineCount {
+		rightLines = append(rightLines, "")
+	}
+
+	divChar := m.st.subtleStyle.Render("│")
+	leftStyle := lipgloss.NewStyle().Width(leftW).MaxWidth(leftW)
+	rightStyle := lipgloss.NewStyle().Width(rightW).MaxWidth(rightW)
+
+	var b strings.Builder
+	for i := range lineCount {
+		l := leftStyle.Render(leftLines[i])
+		r := rightStyle.Render(rightLines[i])
+		b.WriteString(l + " " + divChar + " " + r + "\n")
+	}
+
+	keys := m.st.subtleStyle.Render("[h] History  [s] SLA  [e] Edit  [esc] Close")
+	b.WriteString("  " + keys + "\n")
+
+	return lipgloss.NewStyle().Width(width).MaxWidth(width).Render(b.String())
+}
+
+func (m Model) detailLeftCol(site models.Site, hist monitor.SiteHistory, width int) string {
+	var b strings.Builder
+
+	if len(hist.Latencies) > 0 {
+		chartW := width - 2
+		if chartW < 20 {
+			chartW = 20
+		}
+		chart := m.latencyChart(hist.Latencies, hist.Statuses, chartW, 3)
+		if chart != "" {
+			b.WriteString(chart + "\n")
+		}
+	}
+
+	if len(m.detailDailyDays) > 0 && m.detailChangesSiteID == site.ID {
+		timelineW := width - 2
+		if timelineW < 20 {
+			timelineW = 20
+		}
+		b.WriteString("  " + m.st.subtleStyle.Render("30d") + " " + m.uptimeTimeline(m.detailDailyDays, timelineW) + "\n")
+	}
+
+	return strings.TrimRight(b.String(), "\n")
+}
+
+func (m Model) detailRightCol(site models.Site, hist monitor.SiteHistory, width int) string {
+	dot := m.st.subtleStyle.Render("  ·  ")
+	label := m.st.subtleStyle
+
+	var b strings.Builder
+
+	// Line 1: status + latency + last check
+	status := m.fmtStatus(site.Status, site.Paused, m.isMonitorInMaintenance(site.ID))
+	parts := []string{status}
+	if site.Latency > 0 {
+		parts = append(parts, m.fmtLatency(site.Latency))
+	}
+	if !site.LastCheck.IsZero() {
+		parts = append(parts, m.fmtTimeAgo(site.LastCheck))
+	}
+	b.WriteString(strings.Join(parts, dot) + "\n")
+
+	// Line 2: type-specific details
+	typeParts := m.detailTypeLine(site)
+	if len(typeParts) > 0 {
+		b.WriteString(strings.Join(typeParts, dot) + "\n")
+	}
+
+	// Line 3: uptime + retries
+	uptimeParts := []string{label.Render("Uptime") + " " + m.fmtUptime(hist.Statuses)}
+	if site.Type != "group" && site.MaxRetries > 0 {
+		uptimeParts = append(uptimeParts, label.Render("Retries")+" "+m.fmtRetries(site))
+	}
+	b.WriteString(strings.Join(uptimeParts, dot) + "\n")
+
+	// Error line (if down/broken)
+	if (site.Status == models.StatusDown || site.Status == models.StatusSSLExp ||
+		site.Status == models.StatusLate || site.Status == models.StatusStale) && site.LastError != "" {
+		errW := width - 8
+		if errW < 20 {
+			errW = 20
+		}
+		b.WriteString(label.Render("Error") + " " + m.st.dangerStyle.Render(limitStr(site.LastError, errW)) + "\n")
+	}
+
+	// Blank line before state changes
+	b.WriteString("\n")
+
+	// State changes (one per line, compact)
+	var stateChanges []models.StateChange
+	if m.detailChangesSiteID == site.ID {
+		stateChanges = m.detailChanges
+	}
+	if len(stateChanges) > 0 {
+		limit := 5
+		if len(stateChanges) < limit {
+			limit = len(stateChanges)
+		}
+		for _, sc := range stateChanges[:limit] {
+			ago := fmtDuration(time.Since(sc.ChangedAt))
+			arrow := m.st.subtleStyle.Render("→")
+			from := m.fmtStatusWord(sc.FromStatus)
+			to := m.fmtStatusWord(sc.ToStatus)
+			entry := from + " " + arrow + " " + to + " " + m.st.subtleStyle.Render(ago+" ago")
+			if sc.ErrorReason != "" {
+				reasonW := width - 30
+				if reasonW < 15 {
+					reasonW = 15
+				}
+				entry += "  " + m.st.dangerStyle.Render(limitStr(sc.ErrorReason, reasonW))
+			}
+			b.WriteString(entry + "\n")
+		}
+	}
+
+	return strings.TrimRight(b.String(), "\n")
+}
+
+func (m Model) detailTypeLine(site models.Site) []string {
+	label := m.st.subtleStyle
+	var parts []string
+
+	switch site.Type {
+	case "http":
+		if site.StatusCode > 0 {
+			codeStr := fmt.Sprintf("HTTP %d", site.StatusCode)
+			if site.StatusCode >= httpErrorThreshold {
+				parts = append(parts, m.st.dangerStyle.Render(codeStr))
+			} else {
+				parts = append(parts, m.st.specialStyle.Render(codeStr))
+			}
+		}
+		if site.CheckSSL && site.HasSSL {
+			days := int(time.Until(site.CertExpiry).Hours() / 24)
+			sslStr := fmt.Sprintf("SSL %dd", days)
+			switch {
+			case days <= 0:
+				parts = append(parts, m.st.dangerStyle.Render("SSL EXPIRED"))
+			case days <= site.ExpiryThreshold:
+				parts = append(parts, m.st.warnStyle.Render(sslStr))
+			default:
+				parts = append(parts, m.st.specialStyle.Render(sslStr))
+			}
+		}
+		if site.URL != "" {
+			parts = append(parts, label.Render(limitStr(site.URL, 40)))
+		}
+	case "push":
+		parts = append(parts, label.Render("Push"))
+		if site.Interval > 0 {
+			parts = append(parts, label.Render(fmt.Sprintf("every %s", fmtDuration(time.Duration(site.Interval)*time.Second))))
+		}
+		if !site.LastSuccessAt.IsZero() {
+			parts = append(parts, label.Render("last")+" "+m.fmtTimeAgo(site.LastSuccessAt))
+		}
+	case "ping":
+		parts = append(parts, label.Render("Ping"))
+		if site.Hostname != "" {
+			parts = append(parts, label.Render(site.Hostname))
+		}
+	case "port":
+		parts = append(parts, label.Render("Port"))
+		if site.Hostname != "" {
+			target := site.Hostname
+			if site.Port > 0 {
+				target = fmt.Sprintf("%s:%d", site.Hostname, site.Port)
+			}
+			parts = append(parts, label.Render(target))
+		}
+	case "dns":
+		parts = append(parts, label.Render("DNS"))
+		if site.DNSResolveType != "" {
+			parts = append(parts, label.Render(site.DNSResolveType))
+		}
+		if site.DNSServer != "" {
+			parts = append(parts, label.Render(site.DNSServer))
+		}
+	}
+
+	return parts
+}
+
+// viewDetailSingleCol is the narrow-terminal fallback (original stacked layout).
+func (m Model) viewDetailSingleCol(site models.Site, hist monitor.SiteHistory, width int) string {
+	var b strings.Builder
+	dot := m.st.subtleStyle.Render("  ·  ")
+
+	status := m.fmtStatus(site.Status, site.Paused, m.isMonitorInMaintenance(site.ID))
+	parts := []string{status}
+	if site.Latency > 0 {
+		parts = append(parts, m.fmtLatency(site.Latency))
+	}
+	parts = append(parts, fmt.Sprintf("Uptime %s", m.fmtUptime(hist.Statuses)))
+	if !site.LastCheck.IsZero() {
+		parts = append(parts, m.fmtTimeAgo(site.LastCheck))
+	}
+	b.WriteString("  " + strings.Join(parts, dot) + "\n")
+
+	if (site.Status == models.StatusDown || site.Status == models.StatusSSLExp ||
+		site.Status == models.StatusLate || site.Status == models.StatusStale) && site.LastError != "" {
+		errW := width - 12
+		if errW < 20 {
+			errW = 20
+		}
+		b.WriteString("  " + m.st.subtleStyle.Render("Error") + "  " + m.st.dangerStyle.Render(limitStr(site.LastError, errW)) + "\n")
+	}
+
+	var stateChanges []models.StateChange
+	if m.detailChangesSiteID == site.ID {
+		stateChanges = m.detailChanges
+	}
+	if len(stateChanges) > 0 {
+		limit := 3
+		if len(stateChanges) < limit {
+			limit = len(stateChanges)
+		}
+		var scParts []string
+		for _, sc := range stateChanges[:limit] {
+			ago := fmtDuration(time.Since(sc.ChangedAt))
+			arrow := m.st.subtleStyle.Render("→")
+			from := m.fmtStatusWord(sc.FromStatus)
+			to := m.fmtStatusWord(sc.ToStatus)
+			entry := from + " " + arrow + " " + to + " " + m.st.subtleStyle.Render(ago+" ago")
+			if sc.ErrorReason != "" {
+				entry += "  " + m.st.dangerStyle.Render(limitStr(sc.ErrorReason, 30))
+			}
+			scParts = append(scParts, entry)
+		}
+		b.WriteString("  " + strings.Join(scParts, dot) + "\n")
+	}
+
+	if len(hist.Latencies) > 0 {
+		chartW := width - 4
+		if chartW < 20 {
+			chartW = 20
+		}
+		chart := m.latencyChart(hist.Latencies, hist.Statuses, chartW, 3)
+		if chart != "" {
+			b.WriteString(chart + "\n")
+		}
+	}
+
+	if len(m.detailDailyDays) > 0 && m.detailChangesSiteID == site.ID {
+		timelineW := width - 4
+		if timelineW < 20 {
+			timelineW = 20
+		}
+		b.WriteString("  " + m.st.subtleStyle.Render("30d") + " " + m.uptimeTimeline(m.detailDailyDays, timelineW) + "\n")
+	}
+
+	keys := m.st.subtleStyle.Render("[h] History  [s] SLA  [e] Edit  [esc] Close")
+	b.WriteString("  " + keys + "\n")
+
+	return lipgloss.NewStyle().Width(width).MaxWidth(width).Render(b.String())
+}
+
+func (m Model) fmtStatusWord(status string) string {
+	switch status {
+	case "DOWN":
+		return m.st.dangerStyle.Render("DOWN")
+	case "UP":
+		return m.st.specialStyle.Render("UP")
+	default:
+		return m.st.subtleStyle.Render(status)
+	}
+}

internal/tui/view_sla.go

@@ -40,10 +40,10 @@ func (m Model) viewSLAPanel() string {
 	}
 	bar := m.uptimeBar(r.UptimePct, barWidth)
 	uptimeColor := m.st.specialStyle
-	if r.UptimePct < 99.9 {
+	if r.UptimePct < uptimeExcellentPct {
 		uptimeColor = m.st.warnStyle
 	}
-	if r.UptimePct < 99.0 {
+	if r.UptimePct < uptimeGoodPct {
 		uptimeColor = m.st.dangerStyle
 	}
 	fmt.Fprintf(&b, "  %-16s %s  %s\n", m.st.subtleStyle.Render("Uptime"), uptimeColor.Render(fmt.Sprintf("%s%%", fmtPct(r.UptimePct))), bar)
@@ -94,10 +94,10 @@ func (m Model) buildSLADailyContent() string {
 		pctStr := fmtPct(day.UptimePct) + "%"
 
 		color := m.st.specialStyle
-		if day.UptimePct < 99.9 {
+		if day.UptimePct < uptimeExcellentPct {
 			color = m.st.warnStyle
 		}
-		if day.UptimePct < 99.0 {
+		if day.UptimePct < uptimeGoodPct {
 			color = m.st.dangerStyle
 		}
 
@@ -128,7 +128,7 @@ func fmtPct(pct float64) string {
 	if pct == 100 {
 		return "100.00"
 	}
-	if pct >= 99.99 {
+	if pct >= uptimePrecisionPct {
 		return fmt.Sprintf("%.3f", pct)
 	}
 	return fmt.Sprintf("%.2f", pct)