[2026] OpenTelemetry for LLM Observability — Self-Hosted Setup

I've been running a small AI automation shop — just me, a handful of agents, and a self-hosted stack that needs to stay observable without blowing the budget. When I started instrumenting my LLM pipel