Case study · Realtime · Messaging
An anonymous messaging bot growing to 100k+ concurrent users, where losing a message or letting abuse through broke trust.
Who I worked with
End users chatting anonymously, and the operators keeping the space safe
Stack
BullMQRedisElysiaJSPrisma
→How it came together
01
Discover
Finding the real problem
- Identified the two failure modes that mattered most to users: lost messages and abuse
- Load patterns showed traffic spikes a request-response design would not survive
02
Build
Designing the solution
- An event-driven architecture: BullMQ queues, Redis pub/sub and PostgreSQL behind ElysiaJS
- Abuse detection running inside the message pipeline
- A Turborepo monorepo with Docker and Prisma for a fast, typed release loop
03
Ship
Getting it to people
- Rolled out to a live user base at scale with zero message loss
✓Measured results
100k+
concurrent users
0
messages lost
8k+
abuse incidents stopped daily