This repository has been archived by the owner. It is now read-only.
Fetching contributors…
Cannot retrieve contributors at this time
executable file 199 lines (128 sloc) 5.72 KB

⚠️ This project is no longer active. No further updates are planned.

Local Development

Use predictionio-buildpack to setup your local PredictionIO environment.

To do any real development work with PredictionIO, you'll need to run it locally. We'll use the bin/local/ scripts to simplify that setup procedure and ensure parity between the dev (local) & production (Heroku).


This local dev technique sets up a complete installation of PredictionIO inside each engine you wish to work on. Each engine may have different configuration and dependencies, so the entire environment is contained within an engine directory.

Supported Platforms

This workflow augments the Heroku/Linux-based deployment, and so only supports similar platforms:


  • macOS ⭐️ primary, best experience
  • Debian/Ubuntu Linux

Should Work

  • Linux via Docker or virtualization
  • Windows 10 w/ Linux subsystem

Not Working

  • Windows MS/DOS or PowerShell
  • mobile OSs


0. Remove previously installed pio

If you previously used PredictionIO, then you might have added the pio command to the PATH of the shell. Setup will abort if pio already exists. Please remove any existing PredictionIO entries from the PATH. It may be set in ~/.profile, ~/.bash_profile, or ~/.bashrc.

1. Install Dependencies

⚠️ This step is only required once for your computer.

  1. Install Java/JDK 8
  2. Install PostgreSQL 9.6 and start it

2. The Buildpack

⚠️ This step is only required once for your computer.

# First, change directory up to your top-level projects.
cd ~/my/projects/

git clone
cd predictionio-buildpack/

# Capture this directory's path for use in next steps.
export PIO_BUILDPACK_DIR="$(pwd)"

3. The Engine

⚠️ This step is required for each local engine directory.

With a few commands, we'll install PredictionIO & its dependencies into ./PredictionIO-dist/ and configure it all via env vars & rendered config files.

# First, change directory to the target engine:
cd ~/my/projects/engine-dir/

# Depending on the engine, an environment template may be available.
# Copy & edit it:
cp .env.local .env
# …or create a new one:
echo 'PIO_EVENTSERVER_APP_NAME=my-engine' >> .env
echo 'PIO_POSTGRES_OPTIONAL_SSL=true'     >> .env

# Ignore the local dev artifacts
echo                      >> .gitignore
echo 'bin/pio'            >> .gitignore
echo 'bin/dotenv'         >> .gitignore
echo '.env'               >> .gitignore
echo 'PredictionIO-dist/' >> .gitignore
echo 'repo/'              >> .gitignore

# Setup this working directory:

Refreshing the setup

♻️ Run bin/local/setup whenever:

  • the buildpack is updated/moved
  • an environment variable (including the .env file) is changed that effects dependencies, like:
    • PIO_S3_* or

Here's how:

# Capture the path to the buildpack on your machine (from Step 2.)
export PIO_BUILDPACK_DIR=~/my/projects/predictionio-buildpack

# Then, inside the engine to refresh:

# Finally, verify the new setup is working; Postgres & optionally Elasticsearch must be running:
bin/pio status

If you encounter errors, it may be necessary to reset the local development installation.

4. Postgres

  1. Start Postgres

    • for, use the 🐘 menubar item to start the server
    • other installation methods have their own start-up process
  2. Configure each engine to use its own database. The .env file should define the unique database connection like this:

    ✏️ Replace my_database_name with a name for your engine's database.

  3. Create the Postgres user & database to match that configuration:

    createuser pio
    createdb my_database_name

    👓 These binary commands are included with Postgres. You may need to reference them directly from the database installation.

5. Elasticsearch (optional)

⚠️ Only available if PIO_ELASTICSEARCH_URL is set during bin/local/setup.

Configure Elasticsearch

  1. In the engine, open .env file and add the default local address for ES:

  2. Refresh the setup

Run Elasticsearch

In a new terminal, from the engine's directory:

cd PredictionIO-dist/vendors/elasticsearch/

6. Finally, use bin/pio

bin/pio status
bin/pio app new my-engine-name
bin/pio build --verbose
# Importing data is required before training will succeed
bin/pio train -- --driver-memory 8G
bin/pio deploy

👓 the bin/pio command reads the local environment (config vars) from the local .env file everytime it's invoked.

Run Eventserver

In a new terminal, from the engine's directory:

bin/pio eventserver


▶️ How to deploy to Heroku


If local dev seems broken, try clearing out the local install, and then reinstall.

🚨 This will result in destruction of local state contained in these directories.

rm -rf PredictionIO-dist/ repo/

Then, refresh the setup.