Hey, We've been using Looker (Google) as a BI & semantic layer tool, but it's not that good and seems to have no roadmap. Thinking about moving to something more universal for serving analytics data to some future BI tool, and also to some of our back-office tools.

What do you guys think of CubeJS ? is it mature enough ? Thenks !

We actively use the open source version self hosted to act as an api layer for us.

It works pretty well for that, there are some gotchas if you want to run multi-tenant / enforce security. It also does quite a bit of caching which can cause issues when running multi-tenant / many dbs. Depending on the db you use underneath it, the sql it generates could be better optimized (like all ORMs). We do quite a bit of pre-modeling in the db and expose more optimized tables to cube as the query layer.

Ultimately it works best as an internal single tenant semantic layer. A lot of the work that's being done has been for the cloud service so some parts of the OSS version are starting to fall behind

Docs aren't amazing for parts of it, some query functions aren't documented at all but can be found in the source. Some updates introduce new un-documented features like row level security that significantly slow performance so like all libs have to test well on upgrade.