~/About~/Systems~/Foundry~/Blueprint~/Journal~/Projects
Book a Call

Event-Driven Notification Hub

The complete system record. Choose the proof lens that fits your role.

Choose The Proof Lens

One system record, three reader paths.

Feature-frozen3-act recordGitHub available
Business proofFounder / operator

One Notification Service for Every Project in the Portfolio

Product teams replace 8-10 hours of notification setup per app with one HTTP call, shared templates, preferences, and delivery tracking.

System FlowData ModelArchitecture LayersDecision Log
Open business proof7 min read
Architecture proofCTO / architect

Architectural Brief: Event-Driven Notification Hub

Kafka consumer sits behind a USE_KAFKA flag because Redpanda burns 150MB on a 1GB VPS. Direct HTTP ingestion and Kafka converge on one pipeline.

System TopologyInfrastructure DecisionsConstraintsDecision Log
Open architecture proof6 min read
Technical proofSenior engineer

The Signing Bug You Cannot See Until a Tenant Tries to Verify You

The Hub and the tenant hashed the same JSON payload to different signatures. JSON.stringify key order is non-deterministic, so callback verification silently failed.

Code SampleFailureConstraintSurprise
Open technical proof8 min read
Technical proofSenior engineer

The Cursor Pagination That Worked Until Five Rows Shared a Timestamp

Five suppression rows seeded in one transaction shared a microsecond-level timestamp. Cursor pagination returned three of them. Date.toISOString lost the rest.

Code SampleFailureConstraintScaling Limit
Open technical proof7 min read
Technical proofSenior engineer

The Per-Tenant Rate Limit That Wasn't Per-Tenant

Per-tenant rate limit silently capped every tenant at the default 200/minute. The rate check ran before auth populated the tenant on the request.

Code SampleFailureConstraintSurprise
Open technical proof8 min read
Technical proofSenior engineer

Why I Designed the Whole System Around Kafka and Then Deployed Without It

Redpanda burned 150-200MB of a 1GB VPS before the application even started. The Kafka consumer sits behind a flag until the RAM budget grows.

Code SampleFailureConstraintScaling Limit
Open technical proof7 min read

Related Systems

More proof in the same problem space.

View systems map