Meridian

A multi-tenant telemetry platform for large GPS fleets — the infrastructure layer beneath Horizon. It handles ingestion, normalization, storage and distribution of real-time GPS data, designed around workloads reaching 80,000+ vehicles, hundreds of incoming positions per second, and multi-level tenant hierarchies.

1. GPS devices
2. GPS providers
3. Go ingestion services
4. Normalization
5. PostgreSQL / TimescaleDB
6. Rules + events
7. Centrifugo
8. Web / messaging / Horizon

Multi-tenant by design

Fleet platforms serve several layers of organizations at once:

1. Platform
2. Distributor
3. Integrator
4. Customer
5. Operators

The architecture supports complex organizational trees, role-based access and tenant isolation across every level.

Real-time distribution

A fleet platform gets significantly harder once thousands of connected units need to update interfaces simultaneously. Meridian uses a dedicated real-time layer so telemetry ingestion is never coupled to frontend connections — ingestion, business logic and user-facing apps scale independently.

Provider normalization

Integrations built or explored across Wialon, RedGPS, Traccar, Tracksolid, Webmaps, Lotreck, Navixy and 1NCE. Each ecosystem exposes a different data model and communication mechanism; Meridian creates a common operational layer above them.

I keep protocol and product deliberately separate — device protocols change on the manufacturer's schedule, not mine, and that churn shouldn't leak into product code.

1. Horizon Why does it matter? Investigation and reasoning
2. Event detection What is worth a look? Rules / algorithms / ML
3. Meridian this project What happened? Telemetry infrastructure

Stack Go · TypeScript · Hono · PostgreSQL · TimescaleDB · Redis · Centrifugo · React · TanStack · Docker · Northflank · Cloudflare