Kubernetes, cloud-native infra, and the bits the docs leave out.
Taming Azure AI Foundry with LiteLLM: budgets and key rotation
Azure AI Foundry is amazing until you have to manage it for more than one team. Here is how LiteLLM became the gateway that fixed my key sprawl, capped everyone’s spend, and turned key rotation from a fire drill into a thirty second job.
Valkey or Redis: how I'd actually choose
Short version, because you’ve got a cluster to babysit: run Valkey. The exception is if you need vector search living in the same box as your cache — then look hard at Redis 8 and read the license before you commit. That’s the post. Everything under here is me showing my work. I run Valkey in prod. It’s the cache and rate-limit store in front of an LLM gateway, and the ingestion queue for a tracing stack, on AKS. So this isn’t a hot take about open-source ethics from someone who’s never had to kubectl rollout restart the thing at 1am. The ethics argument is over and you still have to ship something — this is about that part. ...