OpenLIT
OpenTelemetry-native LLM observability with GPU monitoring, guardrails, and evals.
About
OpenLIT (openlit/openlit, Apache-2.0) is an OpenTelemetry-native observability platform for LLM apps: auto-instrumentation for 20+ libraries, token/cost tracking, GPU monitoring, guardrails, and evaluations — all exportable to any OTEL backend, so there is no vendor lock-in.
Very active development (84 releases/12mo). Smaller community than Langfuse, but the OTel-native design is the right long-term bet for teams that already run OpenTelemetry.
What it can do
- Auto-instrumentation for 20+ LLM libraries
- Token, cost, and latency tracking
- GPU monitoring
- Guardrails and evaluations
- OTel-native export — no lock-in
Setup
pip install openlit.- Add
openlit.init()at your app entrypoint (or configure the OTEL exporter). - Point traces at the OpenLIT dashboard or your own OTEL backend.
- Enable guardrails and evals as needed.
Prerequisites
Python 3.8+; an OTEL backend or the OpenLIT dashboard.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier). Apache-2.0, actively developed (84 releases/12mo), official site live, and the OTel-native architecture is genuinely differentiating. Docked for smaller adoption (~2.8k stars) vs Langfuse. Verified tier on repo health + docs + independent coverage.
★ 2,844 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 6.7/10
- Maintenance · recency, releases, issue load
- 8.2/10
- Openness · license + docs
- 9.4/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · 2,844 GitHub stars (log-scaled vs catalog peers)
- · last push 1d ago; 84 releases/12mo; 166 open issues
- · manual review: OTel-native observability; very active releases; smaller than Langfuse; license: Apache-2.0
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.