tangram VS Metabase

In this post, we will go over how we optimized our Gradient Boosted Decision Tree library. This is based on a talk that we gave at RustConf 2021: Writing the Fastest Gradient Boosted Decision Tree Library in Rust. The code is available on GitHub.

We are using Rust for our web application and website at Tangram. You can view the source here: https://github.com/tangramdotdev/tangram/tree/main/crates/www. The website is at https://www.tangram.dev. We decided to write our own web framework because we needed server rendering and we wanted to use the builder pattern in creating components. Here is an example component for our Logo to get a sense of what this looks like: https://github.com/tangramdotdev/tangram/blob/main/crates/www/ui/logo.rs

Check out https://github.com/tangramdotdev/tangram.

Tangram | Senior Rust Programmer | Remote | https://www.tangram.dev

Tangram is an all in one machine learning framework designed for programmers. With Tangram, developers can train models and make predictions on the command line or with libraries for languages including Elixir, Go, JS, Python, Ruby, and Rust, and learn about their models and monitor them in production from a web application. To learn more about what the product does, watch the demo on the homepage at https://www.tangram.dev or check it out on GitHub at https://www.github.com/tangramdotdev/tangram.

We are looking to grow our engineering team with senior Rust programmers. We are currently based in Boston, MA but are looking to build a remote team. At Tangram, you'll get to work on everything from our core machine learning algorithms to writing front-end code in Rust! We are looking for developers with experience in Rust and familiarity with or willingness to learn machine learning concepts. If this sounds exciting, email me (Isabella, cofounder) at [email protected].

We chose to use Rust instead of TypeScript for the front end of https://github.com/tangramdotdev/tangram.

This allows us to:

* Share code with our server written in Rust.

I'm working on machine learning in Rust at Tangram. We currently only provide an implementation of linear models and gradient boosted decision trees but will move into exposing training of deep models in the future. You can check out Tangram here: https://github.com/tangramdotdev/tangram. You may also be interested in checking out Linfa https://github.com/rust-ml/linfa. If you're interested in the future of machine learning in Rust, check out Luca Palmieri's blog post: https://www.lpalmieri.com/posts/2019-12-01-taking-ml-to-production-with-rust-a-25x-speedup/

Great idea! We definitely want to do this. We have an open issue to track it: https://github.com/tangramdotdev/tangram/issues/28.

Hi HN! We are Isabella and David, and we're excited to share Tangram, our attempt to make it easy for programmers who are not machine learning experts to train, deploy, and monitor machine learning models. With Tangram, developers train a model from a CSV file on the command line, make predictions with libraries for Elixir, Go, JavaScript, Python, Ruby, and Rust, and learn about their models and monitor them in production with a convenient web app. Watch the video on the homepage: https://www.tangram.dev, or check it out on GitHub: https://github.com/tangramdotdev/tangram.

Over the past few months we have been working closely with a handful of early users. A team at a small company had a TensorFlow model deployed as a Flask service consumed by their Elixir app. They replaced it with a Tangram model because they didn't want to maintain a server separate from their monolith. A team of front end engineers at a large company was looking for a way to to train and deploy models on their own, without the overhead of involving their data scientists, machine learning engineers, or backend engineers. They trained a model on their own and embedded it directly in their React front-end with the Tangram JavaScript library that makes predictions with WebAssembly.

Tangram is written entirely in Rust, from the core machine learning algorithms, to the bindings for each language, to the front and back end of the web application. We have benefited from Rust's fast performance, strong typing, convenient tooling, and high quality libraries (serde, tokio, hyper, sqlx, and more).

We hope to make Tangram a sustainable business with the "open core" business model. The CLI and language libraries are MIT licensed, while the web application is source available, free to use for testing, but requires a paid license to use in production.

We would love to hear your feedback. Give it a try and let us know what you think!

We've used it for about a year - Blazer is okay if you need a quick SQL query console, but we found it lacking as an actual business intelligence tool. The support for graphs and dashboards is limited, for graphs it requires you to structure the query in an exact way as you can see in the Blazer readme.

After some research on available alternatives that don't break the bank, we decided to deploy a self-hosted instance of Metabase[0]. This took only a few minutes to set up using their Docker image[1] and it has much better graphing capabilities and you can easily put a custom layout together for dashboards. Upgrading is similarly easy (just redeploy). Also easy to configure: data sources, hiding or changing the data type of a column, G Suite sign-in for our domain. Highly recommend it if you need anything more than Blazer's table output.

[0]: https://github.com/metabase/metabase

I've never used Tableau, but heard a lot of hate about it. However, in my previous role, we were big fans of Metabase (https://metabase.com). You can also self-host it, which was a huge win for us.

It seems folks want a working example. Here's one in prod:

Metabase is a BI tool, backend written mostly in Clojure. Like basically all BI tools they have this intermediate representation language thing so you write the same thing in "MBQL (metabase query language)" and it theoretically becomes same query in like, Postgres and Mongo and whatever. End user does not usually write MBQL, it's a service for the frontend querybuilding UI thing and lots of other frontend UI stuff mainly in usage.

Whole processing from MBQL -> your SQL or whatever is done via a buncha big-ass transducers. Metabase is not materially faster than other BI tools (because all the other BI tools do something vaguely similar in their langs) but it's pretty comparable speed and the whole thing was materially written by like 5 peeps

https://github.com/metabase/metabase/blob/master/src/metabas...

(nb: I used to work for Metabase but currently do not. but open core is open core)

They backported it to v0.45x and those changes don't seem to be included: https://github.com/metabase/metabase/compare/v0.45.4...v0.45...

aka, It isn't checked in to source control publicly yet. Interesting.

I tried to "decompile" the jars and loop over the files but it didn't yield much/wasn't clean enough to be of help.

They haven't released the source, and the compiled versions are non-trivial to diff (e.g. there are nondeterministic numbers from the clojure compiler that seem to have changed from one to the other, and .clj files have been removed from the jar).

The old version has `hash=1bb88f5`, which is a public commit: https://github.com/metabase/metabase/commit/1bb88f5

We are unsure about the right license to use, so this is a great feedback. We had a MIT license one week ago that we know that we cannot hold on long term and we felt we were lying to the community by keeping an MIT license and changing it in one year.

By using AGPL, we feel it's the right level of restriction. It's the license used by Metabase for example (https://github.com/metabase/metabase) that many companies use internally.

The solution really depends on what sort of problems you are trying to solve and who your customers are.

There are a fair few low-code solutions out there for reporting and data visualisation that are great for finance and marketing teams for example. e.g. https://metabase.com/ , https://evidence.dev/

For multipurpose SMB workflows and organisational processes, I have used n8n in the recent past and found it was quite good and incredibly easy to maintain. https://n8n.io/engineering-resources/

For enterprise processes I'd go with Camunda (solely based on recommendations and not first hand experience). Although only parts of their platform are OSS https://github.com/camunda

Bear in mind that some of these are not suitable if you want to build something that competes with them while taking their OSS code. But are perfectly fine otherwise.

Anyway, I have a collection of scripts that merge ActivityWatch data from all my machines and WakaTime exports to a PostgreSQL database which I then query with a project called Metabase. If you're curious, the scripts are in a repository called sqrt-data. I've been playing with this for ~4-5 years already I think.

Metabase | https://metabase.com | REMOTE | Full-time | Backend, Frontend, Full Stack, and DevOps engineers

Metabase is open source analytics software that lets anyone in your company rummage around in the databases you have. It connects to a number of databases / data warehouses (BigQuery, Redshift, Snowflake, Postgres, MySQL, etc).