Every LLM observability tool shows you that costs spiked. None of them tell you which deploy caused it.
LLMTrace is a self-hosted Go proxy that sits in front of your Anthropic/OpenAI calls and logs every request with cost, user, and deploy SHA into your own Postgres. When your bill jumps, you can point to the exact commit.
No SaaS, no data leaving your infra. Drop in the docker-compose and you're logging in under 5 minutes.
