Website • Docs • Cube.js Models docs • Docker Hub • Slack community

MLCraft is a low-code metrics layer and a modern open-source alternative to Looker.

MLCraft (Metrics Layer Craft) originally was designed to extract and transform the data from several data warehouses and run machine learning experiments. Cube.js is used as a primary query layer and makes it suitable for handling trillions of data points.

We offer a new way of working with data – it's called metrics store.

• Access insights with confidence, keep teams up-to-date on metrics changes, build institutional knowledge with stakeholders to empower your decision-making

• Take advantage of an enterprise-level metrics platform that allows you to visualize key insights via an easy GraphQL API or JDBC client

• Create individual dashboards for different roles and teams or stakeholder groups, and track changes across multiple data sources with a unified dashboard

Features

• Data modeling: a Cube.js Data Schema is used to model raw data into meaningful business metrics, transform and pre-aggregate data for optimal results.
• Version control: manage schema changes with version control, rollback fast to the valid one.
• Auto testing and documentation: good documentation and data testing will help downstream consumers curate the datasets.
• Metrics Governance: keep an eye on metrics changes, know data owners and contributors, understand the context rapidly.
• Multitenancy and row-level security: collaborate on the metrics and warehouses together with your team.
• Query orchestration and multi-level caching: any production-ready analytics solution requires key components such as analytic SQL generation, query results caching and execution orchestration.

note: not all features are deployed yet. Please, sync with our roadmap. Leave your email in the newsletter form on mlcraft.io if you would know about new features.

Use cases

• Team Business Intelligence
• Events collection & Event analytics
• Machine Learning
• Reverse ETL (as a single source of truth)
• Business Metrics governance
• Metrics metadata management

Demo

Demo: app.mlcraft.org

Login: demo@mlcraft.io

Password: demodemo

Quick start Prerequisites

• Docker
• Docker Compose

Step 1: Get the docker-compose file

The mlcraft-io/mlcraft/install-manifests repo contains all installation manifests required to deploy MLCraft anywhere. Get the docker compose file from there:

# in a new directory run
wget https://raw.githubusercontent.com/mlcraft-io/mlcraft/main/install-manifests/docker-compose/docker-compose.yml


or run

curl https://raw.githubusercontent.com/mlcraft-io/mlcraft/main/install-manifests/docker-compose/docker-compose.yml -o docker-compose.yml

The following command will run MLCraft along with a Postgres database to store its data.

$ docker-compose up -d

Check if the containers are running:

$ docker ps


CONTAINER ID IMAGE                 ... CREATED STATUS PORTS          ...
c8f342d086f3 mlcraft/mlcraft-stack ... 1m ago  Up 1m  80->8888/tcp ...
30ea14ddaa5e postgres:12           ... 1m ago  Up 1m  5432/tcp

It will take about 2-5 minutes to install all dependencies, wait until MLCraft Stack is ready message. Read logs by docker-compose logs -f to understand if it's finished.

Step 3: Try out MLCraft

MLCraft itself will be available on http://localhost/

GraphQL endpoint will be http://localhost/v1/graphql

Admin Console (Hasura Console) will be available on http://localhost/console

NOTE: check out HASURA_GRAPHQL_ADMIN_SECRET in the docker-compose file. You'll need it to enter the Admin Console.

Local development Prerequisites

• Docker
• Docker Compose
• Python3 (with pip3)
• Node.js 8.9+ with yarn installed

git clone https://github.com/mlcraft-io/mlcraft.git
cd mlcraft
bash init.sh

To start UI in dev mode:

python3 cli.py ui

Then visit http://localhost:3000

To start containers manually:

python3 cli.py services up

To destroy all services (volumes will still present):

python3 cli.py services destroy

Also check:

python3 cli.py --help

Ngrok is a tool for allow you to instantly open access to remote systems. We use it for proxying requests from dockerized apps to frontend part in development mode.

After installing the tool you need to obtain your authtoken in ngrok dashboard and update the variable in ngrok.yml file. Then run bash ngrok.sh command. It opens tunnels for backend, frontend and object storage services. Don't forget to change default routes by ngrok-generated ones.

It's necessary to test such features as scheduled reports and alerts locally.

Community support

For general help using MLCraft, please refer to the official MLCraft documentation. For additional help, you can use one of these channels to ask a question:

• Slack (For live discussion with the Community and MLCraft team)
• GitHub (Bug reports, Contributions)

Roadmap

Check out our roadmap to get informed on what we are currently working on, and what we have in mind for the next weeks, months and years.

License

The core MLCraft is available under the Apache License 2.0 (Apache-2.0).

All other contents are available under the MIT License.

Authors

@ifokeev

@ilyozzz

@Libertonius