Case study
API / platform
Watcher24
Active developmentSolo builder — architecture & system design by me, AI-assisted ("vibe coded") implementation
0Updated 7/11/2026
A multi-tenant observability and audit logging platform — ingest telemetry from a JS or Python SDK, process it through a real-time pipeline (Go gateway → Redis Streams → Python analytics worker → ClickHouse), and inspect it live on a WebSocket-powered console.
Problem
Teams that want audit logging and observability for their own product usually end up gluing together an ingestion API, a queue, a time-series store, a realtime layer, and an auth/IAM system from scratch, per project.
Approach
Same process as FileNest — I'm the brain behind it: I designed the system before any code was written — the six-service split (a Go ingestion gateway, a Python analytics worker, a Go WebSocket realtime fan-out service, a Go notifier, a Next.js console, and a separate Next.js + better-auth IAM), the Redis Streams → ClickHouse event pipeline, and the pub/sub-driven realtime fan-out. The implementation itself was AI-assisted — "vibe coded": I gave the architecture, data flow, and specs as input, and an AI coding agent built the Go services, the Python worker, the Next.js console and IAM, and the JS/Python SDKs against that design.
Result
All six services (gateway, analytics worker, realtime, notifier, console, IAM) run locally via Docker Compose with per-service dev commands, backed by JavaScript and Python SDKs and full test suites for each service.