June 10, 2026ArchitectureRAGAI Infrastructure
File-Based RAG Without Apology
Why I chose a 200-line knowledge module over pgvector — and when you should too.
Writing
Architecture notes and engineering decisions from real projects.
June 7, 2026InfrastructureAPI Design
Rate Limit Postmortem
How setting CHAT_MAX_RPM_PER_IP to 2 broke my portfolio chat.
June 4, 2026InfrastructureSelf-HealingEdge Computing
The Watchdog That Doesn't Bark
Building a self-healing monitor that recovers silently — and when to actually wake up a human.
June 1, 2026GoInfrastructureEdge Computing
Inferencia: A Smart LLM Router in Go
Building a Go-based LLM proxy that routes chat, TTS, and embeddings to different backends — and why I hosted it on a Raspberry Pi 5.