Notes on architecture, distributed systems, reliability, data ingestion, and software delivery.
RAG still describes part of this system. It stopped describing the interesting part once retrieved context became durable state that could constrain the next decision.
A pragmatic backend write-up about using Postgres as a lightweight staging and coordination mechanism for a snapshot ingestion pipeline, without introducing Kafka before the workload requires it.
August 2026 update: This article documents a reproducible pre-production/lab deployment, not a production-ready environment. The public-subnet/debugging shortcuts and missing …
We moved from a synchronous REST insert to an event-driven pipeline: batch POST → domain events → private Kafka topic → consumer maps back to domain and processes as before. Here’s the why, the how, and the code.
August 2026 update: This post records the delivery setup as it existed in September 2025. The project has evolved since then; the current Order Tracking case study is the …