Case study
Loading case study
Case study
Loading case study
Case study
A real-time workspace where a team can run an incident together without stepping on each other.
During an incident several people are in the same record at once: editing the write-up, changing status, handing it over. Last-write-wins loses work, and a page refresh to see what others did is too slow.
SyncBoard is built as a multi-tenant product, so the same system also has to keep each customer's data invisible to every other customer, and meter usage for billing, without any of that slowing the live connection down.
Shows the architecture SyncBoard is being built to. The project is in progress, so individual pieces are at different stages.
Socket.io runs as a cluster of nodes that share events through Redis Pub/Sub, so a user connected to one node still sees changes made through another.
Collaborative editing uses Yjs, which merges concurrent edits without a referee. Structured fields use optimistic locking: a stale write gets a 409 and the client rolls back.
Stripe usage metering is pushed onto BullMQ queues, so billing work never blocks a socket message.
RS256 JWTs with refresh-token rotation kept in Redis, written directly instead of through an auth library.
A request for another tenant's resource always returns 404, never 403, so the response doesn't confirm the resource exists. Ownership changes only through one dedicated, atomic route.