Torrix: observability without the database tax
A new LLM observability tool runs without PostgreSQL or Redis. That is not a feature list, that is an architecture decision.
Someone built Torrix, a self-hosted LLM observability platform that does not need PostgreSQL or Redis. That caught my attention because most observability tools ship with a seven-service docker-compose file before you log a single trace. The pitch is simple. You run one binary. It stores everything in SQLite and handles caching in-process. No separate database server, no Redis cluster, no Kafka for the event stream. For teams running a handful of LLM calls per second, that is enough. This matters because observability tools often cost more to run than the thing you are observing. A three-node PostgreSQL setup with Redis sentinels is overkill when you are debugging why your RAG pipeline returns the wrong chunk 12 percent of the time. Torrix bets that most people do not need distributed writes until they are handling thousands of requests per second. The trade-off is obvious. SQLite does not scale horizontally. If you grow past one server’s capacity, you will migrate to something heavier. But for early-stage projects or internal tools, starting with a single binary means you can ship observability in an afternoon instead of spending two days configuring connection pools. I have not run it yet, but the simplicity is appealing. Observability should not require a dedicated platform engineer until you actually need one.
Source: Show HN: Torrix, self hosted, LLM Observability,(no Postgres, no Redis)