`Torrix`, self-hosted LLM observability in one container

Cuts the usual Postgres/Redis stack down to SQLite in a single container. Good wedge for small production agent setups; scale limits are explicit, so the positioning is honest.

• —Setup is basically docker compose up, with all data stored in a local SQLite file; that removes most self-hosting friction.
• —Captures tokens, cost, latency, full prompt/response traces, and reasoning tokens via proxy or Python/Node SDKs.
• —Adds budget caps, cost forecasting, PII masking, evals, prompt versioning, and OpenTelemetry ingestion instead of stopping at logs.
• —The ceiling is clear: hundreds to low thousands of LLM calls per day, not a high-throughput observability backend.