Product / UX / Telemetry & observability

OpenLIT

OpenTelemetry-native AI engineering platform covering LLM observability, GPU monitoring, guardrails, evaluations, and prompt management. Integrates with 50+ LLM providers, vector databases, and agent frameworks. Vault for managing API keys. Differentiated by breadth of integrations and GPU-level monitoring. 2.5K GitHub stars.

openlit PyPI v1.42.0 (8 May 2026); OTel GPU Collector 0.0.6 (3 Jun 2026). Confirmed live on GitHub + PyPI June 2026.

Openness

5 high confidence

5.0

license: Apache-2.0(OSI)
source: public
core-gated: ungated

Fully Apache-2.0, OpenTelemetry-native, no feature-gated core; vendor-neutral by design (exports to any OTel backend).

https://github.com/openlit/openlit recorded 2026-06-04
Apache-2.0 license; OTel-native SDKs; full feature source

Adoption

2 medium confidence

2.0

~590K openlit PyPI downloads/mo; modest but real. As an OTel auto-instrumentation lib some downloads are CI/transitive, so placed conservatively at level 2 (early adopters). ~2.5k GitHub stars corroborates a smaller community vs Langfuse/MLflow.

https://pypistats.org/packages/openlit recorded 2026-06-04
~590,377 openlit downloads in last month

Capability

4 high confidence

4.0

Broad feature matrix covering all six recipe dimensions (tracing, prompt mgmt, evals, datasets/annotation-adjacent, cost, OTel+integrations) plus GPU observability; comparable to Langfuse C4. Held at 4 (not 5) as no observability product is a clear frontier-definer over MLflow/Langfuse anchors.

https://github.com/openlit/openlit recorded 2026-06-04
feature list: tracing/metrics, 11 eval types, Prompt Hub, cost tracking, guardrails, vault, 50+ integrations

Unchanged since 2026-07-30 (last edited, not re-checked)