Installation

Select the installation mode that best suits your needs

Usage Scenario	Installation Mode
Run FeatureByte tutorials	Tutorials
Prototype using data warehouse without deployment	Standalone
Deploy features	Hosted Service
Collaborate with other users	Hosted Service
Scalable production	High Availability

Installation modes

TutorialsStandaloneHosted ServiceHigh Availability

Get started quickly with the FeatureByte tutorials.

Installation Steps

Requirements

Software

• Python 3.8 or higher

Step 1: Obtain an API Token

1. Visit the FeatureByte Hosted Tutorials page
2. Follow the instructions to sign up for an API token.

This token will grant you access to the hosted tutorials server running on a Snowflake data warehouse, where the tutorial data is already populated.

Step 2: Set up a virtual environment.

Using a virtual environment helps avoid potential conflicts with other packages. Here's how you can set one up:

virtualenvconda

# Create a new virtual environment
python -m venv featurebyte-env

# Activate the virtual environment
## On Windows, run:
featurebyte-env\Scripts\activate.bat
## On Unix or MacOS, run:
source featurebyte-env/bin/activate

Further reading: venv documentation.

# Create a new virtual environment
conda create --name featurebyte-env python=3.9 -y
# Activate the virtual environment
conda activate featurebyte-env

Further reading: conda documentation.

Step 3: Download the tutorial notebooks.

Download the end-to-end notebooks archive to a location on your computer.

Step 4: Load the tutorials in your preferred notebook environment.

Install your preferred notebook environment, like JupyterLab and start with the 00_Setup_Environment.ipynb notebook.

virtualenvconda

(featurebyte-env) ~/featurebyte_tutorials> pip install jupyterlab
(featurebyte-env) ~/featurebyte_tutorials> jupyter-lab 00_Setup_Environment.ipynb

(featurebyte-env) ~/featurebyte_tutorials> conda install -c conda-forge jupyterlab
(featurebyte-env) ~/featurebyte_tutorials> jupyter-lab 00_Setup_Environment.ipynb

Step 5: Run 00_Setup_Environment.ipynb

1. Replace <api_token> with the API token you received after registering

   

2. Run the notebook.

You are now all set to immerse yourself in the tutorials and the end-to-end workflow!

Local installation of the FeatureByte service is the easiest way to get started with the FeatureByte SDK. It is a single-user installation that can be used to prototype features locally with your data warehouse.

Installation Steps

Requirements

Hardware

• Intel, AMD or Apple Silicon processor with 4 cpu cores
• 8GB of RAM

Software

• Python 3.8 or higher
• Docker service

Step 1: Activate a virtual environment.

Using a virtual environment will help avoid potential conflicts with other packages.

virtualenvconda

# Create a new virtual environment
python -m venv featurebyte-env

# Activate the virtual environment
## On Windows, run:
featurebyte-env\Scripts\activate.bat
## On Unix or MacOS, run:
source featurebyte-env/bin/activate

Check out the venv docs for more details.

# Create a new virtual environment
conda create --name featurebyte-env python=3.9 -y
# Activate the virtual environment
conda activate featurebyte-env

Check out the conda docs for more details.

Step 2: Install the featurebyte python package using pip.

pip install featurebyte

Step 3: Launch the environment with python, or from the command shell.

Docker service required

Ensure that your docker service is running.

PythonShell

import featurebyte as fb
fb.start()

featurebyte start

Follow instructions to create a FeatureStore that connect to your data warehouse.

Cleaning Up

Stop all services when you are done using it.

Warning

Deployed feature jobs will be stopped when the service is stopped.

PythonShell

fb.stop()

featurebyte stop

FeatureByte service hosted on a single server provides a light-weight option to support collaboration and job scheduling with limited scalability and availability. Multiple users can connect to the service using the FeatureByte SDK, and deploy features for production.

Installation Steps for Service Host

Service Host Requirements

Hardware

• Intel, AMD or Apple Silicon processor with 4 cpu cores
• 8GB of RAM
• Up all the time, so that scheduled tasks can be executed on time
• Has access to data warehouses to perform computation queries

Software

• Python 3.8 or higher
• Docker service

Step 1: Activate a virtual environment.

Using a virtual environment will help avoid potential conflicts with other packages.

virtualenvconda

# Create a new virtual environment
python -m venv featurebyte-env

# Activate the virtual environment
## On Windows, run:
featurebyte-env\Scripts\activate.bat
## On Unix or MacOS, run:
source featurebyte-env/bin/activate

Check out the venv docs for more details.

# Create a new virtual environment
conda create --name featurebyte-env python=3.9 -y
# Activate the virtual environment
conda activate featurebyte-env

Check out the conda docs for more details.

Step 2: Install the featurebyte python package using pip.

pip install featurebyte

Step 3: Start the service with python, or from the command shell.

Docker service required

Ensure that your docker service is running.

PythonShell

import featurebyte as fb
fb.start()

featurebyte

This will start the services as docker containers in daemon mode.

Installation Steps for Python SDK (Individual User)

Python SDK Requirements

Hardware

• Intel, AMD or Apple Silicon processor with 2 cpu cores
• 4GB of RAM

Software

• Python 3.8 or higher

Step 1: Activate a virtual environment.

Using a virtual environment will help avoid potential conflicts with other packages.

virtualenvconda

# Create a new virtual environment
python -m venv featurebyte-env

# Activate the virtual environment
## On Windows, run:
featurebyte-env\Scripts\activate.bat
## On Unix or MacOS, run:
source featurebyte-env/bin/activate

Check out the venv docs for more details.

# Create a new virtual environment
conda create --name featurebyte-env python=3.9 -y
# Activate the virtual environment
conda activate featurebyte-env

Check out the conda docs for more details.

Step 2: Install the FeatureByte SDK for each user that needs to connect to the service.

pip install featurebyte

Update the FeatureByte configuration file to specify how to connect to the FeatureByte service.

~/.featurebyte/config.yaml

profile:

 - name: local
   api_url: http://localhost:8088
 - name: featurebyte-svc
   api_url: http://<featurebyte-svc-hostname>:8088

Modify the existing entry or create a new one with the correct api_url value.

If you have more than one profile, specify the profile to use when connecting to the service.

import featurebyte as fb
fb.use_profile("featurebyte-svc")

Follow instructions to create a FeatureStore that connect to your data warehouse.

High availability installation of the FeatureByte service is the recommended way to run the service in production. Scale to a large number of users and deployed features, and provide highly available services.

Hardware Requirements

• ⁣> 16 vCPUs
• ⁣> 32GB of RAM
• ⁣> 200GB of storage from specified storageclass

Software Requirements

• Kubernetes Cluster
• kubectl
• helm

Installation Steps

Visit the Featurebyte Helm charts repository and follow the instructions to install the FeatureByte service on your Kubernetes cluster.