🏗️ Platform Architecture | Updated: 2025-10-25 | Version: v2.3.0
Agentic Maintenance OS Architecture
AI-Powered Maintenance Operating System architecture: autonomous agents for task automation, predictive maintenance engine, intelligent scheduling algorithms, self-evolving learning capabilities, and cross-fleet intelligence
🏗️ Technology Stack (Production-Verified)
Frontend Architecture
| Technology |
Version |
Purpose & Implementation |
| React |
18.3.1 |
Modern concurrent rendering with Suspense boundaries for code splitting |
| TypeScript |
5.5.3 (strict mode) |
100% type safety across codebase, zero any types policy |
| Vite |
5.4.1 |
Lightning-fast HMR during development, optimized production builds with 38% bundle reduction |
| TanStack Query |
5.56 (React Query) |
Advanced server state management with maritime domain-specific caching strategies |
| shadcn/ui + Radix UI |
Latest |
Enterprise-grade accessible component library with ARIA compliance |
| Tailwind CSS |
3.4 |
Utility-first styling with custom maritime design tokens and responsive breakpoints |
| React Router |
6.x |
Client-side routing with code-split lazy loading for optimal performance |
| react-window |
Latest |
Virtualization for efficient rendering of large equipment and task lists |
Backend Infrastructure
| Component |
Technology |
Implementation Details |
| Database |
Supabase PostgreSQL |
Enterprise-grade database with advanced indexing and materialized views |
| Authentication |
Supabase Auth |
Secure JWT-based authentication with multi-tenant user management |
| Row-Level Security |
PostgreSQL RLS |
Comprehensive data isolation with dual-access patterns (system admin + org users) |
| Real-Time |
WebSocket Subscriptions |
Live data synchronization with <200ms latency, selective org filtering |
| Storage |
Supabase Storage |
Secure file management with organization-based access control, orphan protection |
| Edge Functions |
Deno runtime |
Serverless compute for log ingestion and custom business logic |
| Connection Pooling |
PgBouncer |
Efficient database connection management for scalability |
🗄️ Database Architecture: 5-Schema Design
Schema 01: Multi-Tenant Organization Structure
Purpose: Root-level organization hierarchy with complete data isolation
- organizations: Root entities with subscription tiers, feature flags, regulatory jurisdiction
- system_users: Global access users (support/admin) with
can_access_all_data flag
- organization_users: Organization-isolated users with role-based permissions
- organization_divisions: Business units, departments, operational divisions
- organization_locations: Global offices, facilities, geographic locations
Schema 02: Vessel & Fleet Management
- vessels: IMO number, vessel type, operational status, flag, classification society
- vessel_types: Vessel categories with specifications and regulatory requirements
- vessel_equipment_installations: Serial number, installation date, position, criticality
- equipment_definitions: Manufacturer, model, system code, specifications (global catalog)
- equipment_parts: Part numbers, manufacturers, critical spare flags, average pricing
- operating_zones: Geographic bounds, regulatory requirements, environmental conditions
Schema 03: PMS (Planned Maintenance System) ⭐ Revolutionary
Industry First: Schedule-specific hours tracking with independent baselines per maintenance activity
- schedule_working_hours: Per-schedule hour tracking with reset history audit trail
- pms_schedules: Maintenance templates with dual-interval logic (hours + time)
- pms_tasks: Individual tasks with status workflow (pending → in_progress → for_review → completed)
- pms_task_history: Complete audit trail with user attribution and timestamps
- pms_task_time_entries: Labor time tracking per task with user accountability
- maintenance_parts_consumption: Parts usage per task with cost tracking
- pms_task_attachments: File attachments with 24-hour orphan protection
- equipment_maintenance_recommendations: OEM manufacturer recommendations (MAN B&W, Wärtsilä, etc.)
- pms_alerts: Automated alert generation with severity classification
Schema 04: Events & Incidents Management
- events: Event tracking with type, severity, investigation status, resolution
- event_types: Categorized events with investigation requirements and notification flags
- event_severity_levels: Severity classification with response time SLAs
- event_work_requests: Event-to-task workflow with priority and assignment
- event_files: Photo/video/document evidence with orphan protection system
Schema 05: Storage & File Management
- storage.objects: Supabase storage integration with metadata tracking
- Buckets: vessel-files, event-files, pms-files (organization-scoped)
- RLS Policies: Organization-based access control on all storage operations
- Orphan Protection: 24-hour grace period before automatic cleanup
🤖 AI Assistant: Learn the Platform Conversationally
Never Read Documentation Again: The fleetcore AI assistant knows every platform feature. Ask how to do something, get instant answers with examples.
Platform Help & Guidance
| What You're Trying To Do |
Just Ask the AI |
What You Get |
| Set up a new vessel |
"How do I add a new vessel to the system?" |
Step-by-step walkthrough of vessel setup, required fields, best practices |
| Configure maintenance schedule |
"How do I create a 500-hour oil change schedule?" |
Explanation of schedule-specific hours, intervals, alert thresholds, with screenshots references |
| Understand alert system |
"Why am I getting a critical alert?" |
Explanation of alert triggers, how to clear them, when to reset hours |
| Track parts inventory |
"How does parts consumption tracking work?" |
Guide to recording parts used, reorder points, cost tracking per task |
| Export compliance report |
"How do I generate a report for PSC inspection?" |
Instructions for filtering data, exporting maintenance logs, audit trail access |
Maritime Knowledge + Technical Specs
Beyond Platform Help: The AI also answers maritime technical questions while you work:
- "What's the SOLAS requirement for this equipment?" - Get regulatory context for maintenance tasks
- "Recommended oil for Caterpillar 3516B?" - OEM specs while creating maintenance schedules
- "How often should I service this pump type?" - Industry best practices for interval planning
- "What's the classification for this safety equipment?" - Criticality ratings for SOLAS compliance
Intelligent Research for Vessel Data
Enable "Online Research" toggle when you need current vessel or equipment information:
| Scenario |
How AI Helps |
| Adding a vessel you're taking over management for |
Ask: "What are the specifications for MV [Vessel Name]?" - AI searches 28 optimized maritime sources (vessel registries, class societies) and gives you dimensions, DWT, equipment list with citations |
| Need OEM part numbers for procurement |
Ask: "Wärtsilä 32 fuel injector part number" - AI finds manufacturer catalogs and cross-references |
| Comparing equipment for specification |
Ask: "Compare Caterpillar vs MAN B&W auxiliary generators" - AI analyzes multiple sources and gives factual comparison |
| Fleet benchmarking research |
Ask: "What vessels does [Competitor] operate?" - AI finds fleet lists, vessel types, and operational data |
Why 28 Sources Are Optimal
We don't search the entire web randomly. We use highly optimized search criteria targeting 28 carefully selected authoritative maritime sources:
- Tier 1 Official: MarineTraffic, VesselFinder, Equasis, DNV, Lloyd's Register, ABS, IMO (vessel registries & class societies)
- Tier 2 Manufacturers: Caterpillar, Wärtsilä, MAN Energy Solutions, Rolls-Royce, Cummins (OEM documentation)
- Tier 3 Technical: Manufacturer catalogs, technical databases, maritime engineering resources
- Smart Filtering: AI prioritizes official sources over news, manufacturer docs over forums, excludes unreliable sites
Result: Fast, accurate, verified answers - not overwhelming you with 100+ random web pages.
Real Operational Scenarios
- Chief Engineer at 2 AM: Equipment alarm sounds. Opens fleetcore, asks AI: "Fuel pressure drop on Wärtsilä 32 troubleshooting" - gets systematic diagnostic steps immediately, no waiting for office hours
- Technical Superintendent Planning: Creating annual maintenance budget. Asks AI: "What's the recommended overhaul interval for MAN B&W 6S50MC-C?" - gets OEM recommendation with source citation, adds to planning
- New Fleet Manager Onboarding: Joining company, needs to understand vessels. Asks AI about each vessel in fleet - gets full briefing on capabilities, equipment, maintenance status
- PSC Inspection Tomorrow: Asks AI: "SOLAS Chapter III requirements for life-saving equipment" - gets regulation text, practical checklist, what inspector will verify
- Procurement Decision: Needs to replace generator. Asks AI: "Compare specs and maintenance costs for Cat 3516B vs Cummins QSK60" - gets data-driven comparison
Available Everywhere
Chat icon visible on every page. Working on maintenance schedules? Ask a question. Reviewing compliance? Ask a question. Planning procurement? Ask a question. Your maritime expert is always there.
🔄 Real-Time System Architecture
WebSocket Subscriptions
| Data Type |
Update Latency |
Filtering Strategy |
| Equipment Hours |
<200ms |
Organization + vessel filtering |
| Maintenance Task Status |
<200ms |
Organization + assignment filtering |
| Alert Generation |
<200ms |
Organization + severity filtering |
| File Attachments |
<200ms |
Organization + task filtering |
| AI Assistant Responses |
<1s first token |
Streaming SSE with chunk delivery |
Intelligent Cache Management
Maritime Domain-Specific Caching Strategy optimized for vessel operations data access patterns
| Data Type |
Stale Time |
Cache Time |
Rationale |
| Vessels |
5 minutes |
30 minutes |
Vessel data changes infrequently, keep cached for quick access |
| Equipment |
2 minutes |
15 minutes |
Moderate change frequency, balance freshness vs performance |
| Maintenance Tasks |
1 minute |
10 minutes |
High change frequency, aggressive refetch for critical operational data |
| Organizations |
15 minutes |
2 hours |
Very stable data, long-term cache with manual invalidation |
| Compliance Data |
10 minutes |
60 minutes |
Important but stable regulatory tracking data |
🔒 Security Architecture
Row-Level Security (RLS) Implementation
Dual-Access Pattern: System admins (global oversight) + Organization users (isolated access)
Standard RLS Policy Pattern (All Tables)
CREATE POLICY "organization_access" ON table_name
FOR ALL TO authenticated
USING (
-- System Admin: Global access for support
EXISTS (
SELECT 1 FROM system_users
WHERE id = auth.uid()
AND role = 'system_admin'
AND can_access_all_data = true
)
OR
-- Organization User: Isolated to assigned org
organization_id IN (
SELECT organization_id FROM organization_users
WHERE auth_user_id = auth.uid()
)
);
Security Features
- JWT-Based Authentication: Secure token-based user authentication
- HTTPS Enforcement: All API communication encrypted in transit
- Data Encryption at Rest: PostgreSQL encryption for stored data
- Security Definer Functions: Prevent infinite recursion in RLS policies
- Audit Trail: Complete activity logging with user attribution and timestamps
- GDPR Compliance: Data retention policies, right to access, right to erasure
⚡ Performance Optimization
Frontend Performance
- Code Splitting: 38% bundle size reduction through intelligent route-based chunking
- Lazy Loading: All major routes dynamically imported with React.lazy()
- Tree Shaking: Vite eliminates unused code during build process
- Asset Optimization: Image compression, SVG optimization, font subsetting
- Virtualization: react-window for efficient rendering of 10,000+ item lists
Database Performance
- Strategic Indexing: All query-heavy columns indexed for sub-100ms queries
- Materialized Views: Pre-computed analytics for instant dashboard loading
- Query Optimization: EXPLAIN ANALYZE on all critical queries
- Connection Pooling: PgBouncer for efficient connection management
- Batch Operations: Bulk inserts and updates for efficiency
Performance Targets (Production)
| Metric |
Target |
Measurement Method |
| First Contentful Paint |
<1.2 seconds |
Lighthouse audit |
| Largest Contentful Paint |
<2.5 seconds |
Lighthouse audit |
| Cumulative Layout Shift |
<0.1 |
Lighthouse audit |
| Lighthouse Score |
>95 |
Google Lighthouse |
| API Response Time |
<100ms (95th percentile) |
Application monitoring |
| Database Query Time |
<50ms (average) |
Database monitoring |
| Real-Time Update Latency |
<200ms |
WebSocket monitoring |
| Uptime SLA |
99.99% |
Infrastructure monitoring |
📈 Scalability Architecture
Horizontal Scaling Capabilities
- Concurrent Users: Designed for 1000+ simultaneous users per organization
- Data Processing: Handle 10,000+ maintenance tasks per month per vessel
- Storage Growth: Linear scaling with organization count (no performance degradation)
- Regional Distribution: Global CDN edge locations for optimal latency
- Database Scaling: Supabase infrastructure auto-scales with demand
Multi-Tenant Design Benefits
- Data Isolation: Complete organization-level data separation
- Performance Independence: One organization's load doesn't affect others
- Linear Scaling: Add unlimited organizations without performance impact
- Security Boundaries: RLS ensures no cross-organization data access
🛠️ Automated Database Triggers
Maintenance Workflow Automation
- Initial Task Creation: Automatic task generation when new schedule created
- Recurring Task Generation: Create next task immediately when current task completed
- Hours Update Propagation: Update all schedule-specific hours when equipment hours change
- Schedule Reset Baseline: Set reset baseline hours when maintenance task completed
- Alert Generation: Automatic alert creation when tasks approach due dates/hours
📊 Analytics & Reporting
Real-Time Analytics Dashboard
- Equipment Health Scoring: 0-100 health score based on PMS compliance and overdue tasks
- Overdue Task Detection: Automated identification with criticality weighting
- Compliance Tracking: Real-time SOLAS/MARPOL/ISM compliance percentage
- Fleet-Wide Visibility: Cross-vessel performance monitoring and benchmarking
- Resource Planning: Maintenance workload forecasting and crew allocation
🚀 Enterprise Architecture Value
Platform Status
- ✅ Core Architecture: Production-ready with comprehensive testing
- ✅ Security Framework: Multi-tenant RLS validated and verified
- ✅ Performance Optimization: Code splitting, caching, virtualization implemented
- ✅ Database Design: 5-schema architecture with automated triggers
- ✅ Real-Time Systems: WebSocket subscriptions operational
- 🔄 Final Validation: Maritime operator testing and feedback integration
Contact: https://fleetcore.ai/contact
Technical Documentation: Available upon request for enterprise prospects