Product / UX / Telemetry & observability

Laminar

Observability platform built for AI agents, providing tracing, evaluation and analytics for agentic workflows with an emphasis on debugging multi-step runs. It has a Rust backend and a TypeScript UI, deploys from the repository's docker-compose files, and makes traces, spans, metrics and events queryable with SQL.

The lmnr PyPI download count runs far above what a project of this size plausibly serves and is treated as mirror-inflated, so the adoption band is discounted rather than read straight off it. Verified 2026-08-13 via the lmnr-ai/lmnr repository and the lmnr PyPI project page.

Openness

5 medium confidence

5.0

license: Apache-2.0(OSI)
source: public(Rust core + TS UI)
SDKs: Apache-2.0
core-gated: ungated

Repo and SDKs Apache-2.0 with full source; a hosted cloud exists but no enterprise-gated module split was visible in the repo (less explicit open-core boundary than Langfuse/LangWatch).

https://github.com/lmnr-ai/lmnr recorded 2026-08-13
Repository page. The license sidebar resolves LICENSE.md to Apache-2.0 (spdxId "Apache-2.0"), there is no ee/ or enterprise directory in the root listing, and the root carries docker-compose.yml plus four variants (full, local-build, local-dev, local-dev-full), which is the self-host path.
https://pypi.org/project/lmnr/ recorded 2026-08-13
lmnr 0.7.59, "Python SDK for Laminar", uploaded 11 Aug 2026, with the License expression field reading Apache-2.0.

Adoption

3 low confidence

3.0

pypistats reports 34,062,990 downloads in the last month for lmnr (9,228,203 in the last week, 1,293,104 in the last day). That is a level-5 figure on the download scale and it is not believed: at roughly 1.3M downloads a day lmnr would outrun Langfuse, a project an order of magnitude larger by stars, contributors and customers, which is the tell for transitive or mirror pulls of an OpenTelemetry auto-instrumentation SDK. That is why the band rests on reported traction rather than a download count - a download band claims a figure somebody else can recompute, and the only figure available here is one this record disputes. What the level does rest on is the standing of an Apache-2.0 tracing platform of this size: a real but small user base, around 3,000 GitHub stars.

https://pypistats.org/api/packages/lmnr/recent recorded 2026-08-14
last_month = 34,084,386 downloads for lmnr (last_week 9,210,651, last_day 1,617,970); recorded as a measurement of the channel, not as the basis of the band.
https://pypistats.org/api/packages/lmnr/recent recorded 2026-08-14
pypistats for lmnr: `{"data":{"last_day":1293104,"last_month":34062990,"last_week":9228203},"package":"lmnr"}`. Recorded as a measurement of the PyPI channel, explicitly NOT as the basis of the band.
https://pypistats.org/packages/lmnr recorded 2026-06-04
~17.06M lmnr downloads in last month (flagged as likely mirror-inflated)

Capability

3 medium confidence

3.0

Solid coverage of OpenTelemetry-native tracing, evaluations run from the SDK, CLI and CI, datasets with an annotation UI, and a differentiated SQL-over-traces and dashboard layer. Prompt versioning and explicit cost tracking are not surfaced in the repository, which leaves it below Langfuse and MLflow on breadth, one rung under the anchor at 4.

https://github.com/lmnr-ai/lmnr recorded 2026-08-13
README feature checklist - "Evals. Unopinionated, extensible SDK and CLI for running evals locally or in CI/CD pipeline", "UI for visualizing evals and comparing results", and "Query traces, spans, metrics, and events with SQL". No prompt-versioning or prompt-management item appears.

Verified 2026-08-13