Skip to content

SafeWatch Modules

KPI Dashboards

Each role lands on their dashboard after login. Route is determined by role claim.

sys_admin (/safewatch/sysadmin): - API error rate (4xx+5xx / total requests, last 24h) — from api_request_events - Average response time by route (top 10 slowest) — from api_request_events - Requests per hour (sparkline, last 24h) — from api_request_events - Auth failure count (last 24h) — from auth_events where success = false - Top error routes (table: route, count, avg duration) — from api_request_events

product (/safewatch/product): - DAU / WAU / MAU (distinct user_ids in period) — from api_request_events - Sessions uploaded per day (7d sparkline) — from session_events - Coach queries per day (7d sparkline) — from ai_events - Leaderboard submissions per day — from leaderboard_events - Feature usage breakdown (pie: sessions vs coach vs leaderboard) — from all three tables

support (/safewatch/support): - Recent registrations (last 10) — from auth_events where action = 'register' - Recent auth failures (last 10, with failure_reason) — from auth_events - Most active users today (top 10 by request count) — from api_request_events - Quick user search by user_id → links to User Explorer

ceo (/safewatch/ceo): - MAU (current month distinct users) — from api_request_events - Total registered users (cumulative) — from auth_events where action = 'register' - Conversion rate: registered → first session uploaded (users with ≥1 session_event / total registered) - AI usage cost total (last 30d) — from ai_events.cost_usd sum - Retention proxy D7: users who registered in the last 30d AND had at least one event ≥7 days after their registration date (D7 retained); same logic for D30 - Volume trend (sessions + coach queries + leaderboard, 30d bar chart) - Billing KPIs (MRR, ARR, churn, paying users): shown as "— available when billing is integrated"

Event Explorer (/safewatch/events)

Filterable, paginated table of all events across all 5 tables.

Filters: - Table (all / api_requests / auth / sessions / ai / leaderboard) - User ID (text input) - Time range (TimeRangePicker component) - Status / action (contextual per table)

Columns: timestamp, table, user_id, action/route, key metadata (varies by table), status.

Pagination: 50 rows per page, server-side. Newest first.

User Explorer (/safewatch/users/{userId?})

Search by user_id → full chronological event timeline for that user across all 5 tables.

Timeline entries show: timestamp, event type (colour-coded by table), action/route, key fields.

If no user_id in URL: shows search box. On submit: navigates to /safewatch/users/{userId}.

Feature Analytics (/safewatch/features)

Three feature usage metrics, selectable time range (7d / 30d / 90d):

  • Adoption: unique users who used each feature (sessions, coach, leaderboard) as % of MAU
  • Volume: daily usage count per feature (stacked bar chart)
  • Abandon rate: (for coach queries) requests where no subsequent event within 5 min — proxy for failed/abandoned queries

Data sources: session_events, ai_events, leaderboard_events.

Funnels (/safewatch/funnels)

Two pre-defined funnels (not configurable in this phase):

Onboarding funnel: 1. Registered (auth_events register) 2. First login (auth_events login_success, after register) 3. First session uploaded (session_events session_uploaded, first per user) 4. First coach query (ai_events coach_query, first per user) 5. First leaderboard submission (leaderboard_events lap_submitted, first per user)

Shows: count and % drop-off at each step. Time range: last 30d / 90d / all time.

Activation funnel: 1. Registered 2. Session uploaded within 7 days of registration 3. Coach query within 7 days of registration

Alerts Center (/safewatch/alerts)

Rule-based anomaly detection computed on each page load (no background job in this phase).

Alert rules (all thresholds configurable in appsettings): - Error rate spike: 4xx+5xx > 5% of requests in last 1h (vs baseline last 7d) - Auth failure spike: > 10 failures in any 5-minute window (last 1h) - AI cost spike: total cost_usd in last 1h > $1.00 - No activity: 0 API requests in last 30 min (potential outage)

Each alert shows: severity (warning / critical), description, affected metric, timestamp of detection, and a link to the relevant Explorer page.

Global Search (/safewatch/search)

Single search box. Searches in parallel: - Users: matches user_id prefix in any event table → links to User Explorer - Events: full-text match on route (api_requests), action (auth/session/ai/leaderboard) — returns top 20 most recent matching events - No free-text search into event metadata fields (too broad; filtered views are the correct tool)