Note: For upgrade instructions please refer to this page.
Sep 20, 2018
- PIO-161: Spark 2.3 support.
- PIO-158: More officially deprecate support for Scala 2.10 and Spark 1.x.
- PIO-152: DOAP syntax error.
- PIO-155: Fix 'Topic Labelling with Wikipedia' Template Link.
- PIO-156: Stale release on download page.
- PIO-160: Array out of bound exception in JDBCUtils when --env is not supplied to CreateWorkflow.
The following contributors have spent a great deal of effort to bring to you this release:
Donald Szeto, Takako Shimamoto
Mar 11, 2018
- PIO-126: Update install.sh to use binary release.
- PIO-137: Create a connection object at a worker to delete events.
- PIO-101: Document usage of plug-in of event server and engine server.
- PIO-127: Update PMC documentation for release process.
- PIO-129: Move CLI document in side menu.
- PIO-131: Fix Apache licensing issues for doc site.
- PIO-133: Make sure project web site meets all requirements in Apache Project Website Branding Policy.
- PIO-135: Remove all incubating status.
- PIO-139: Update release process doc to include closing all resolved stories within the new release.
- PIO-146: Change TM to (R) on text marks.
- PIO-147: Fix broken Scala API documentation.
- PIO-150: Update Ruby gem dependency versions for security improvement.
- PIO-151: Add S3 storage provider docs.
The following contributors have spent a great deal of effort to bring to you this release:
Chan Lee, Donald Szeto, Helene Brashear, James Ward, Jeffrey Cafferata, Mars Hall, Naoki Takezoe, Shinsuke Sugaya, Steven Yan, Takahiro Hagino, Takako Shimamoto
Sep 27, 2017
- PIO-61: S3 support for model data
- PIO-69, PIO-91: Binary distribution of PredictionIO
- PIO-105, PIO-110, PIO-111: Batch predictions
- PIO-95: Raise request timeout for REST API to 35-seconds
- PIO-114: Basic HTTP authentication for Elasticsearch 5.x StorageClient
- PIO-116: PySpark support
- PIO-106: Elasticsearch 5.x StorageClient should reuse RestClient (see the pull request)
- PIO-59:
pio app new
uses /dev/urandom/ to generate entropy. - PIO-72:
pio-shell
properly loads storage dependencies. - PIO-83, PIO-119: Default environment changed to Spark 2.1.1, Scala 2.11.8, and Elasticsearch 5.5.2.
- PIO-99:
pio-build
checks for compilation errors before proceeding to build engine. - PIO-100:
pio
commands no longer display SLF4J warning messages.
- PIO-56: Core unit tests no longer require meta data setup.
- PIO-60, PIO-62: Minor fixes in authorship information and license checking.
- PIO-63: Apache incubator logo and disclaimer is displayed on the website.
- PIO-65: Integration tests on Travis caches downloaded jars.
- PIO-66: More detailed documentation regarding release process and adding JIRA tickets.
- PIO-90: Improved performance for /batch/events.json API call.
- PIO-94: More detailed stack trace for REST API errors.
- PIO-97: Update examples in official templates.
- PIO-102, PIO-117, PIO-118, PIO-120: Bug fixes, refactoring, and improved performance on Elasticsearch behavior.
- PIO-104: Bug fix regarding plugins.
- PIO-107: Obsolete experimental examples are removed.
The following contributors have spent a great deal of effort to bring to you this release:
Aayush Kumar, Chan Lee, Donald Szeto, Hugo Duksis, Juha Syrjälä, Lucas Bonatto, Marius Rabenarivo, Mars Hall, Naoki Takezoe, Nilmax Moura, Shinsuke Sugaya, Tomasz Stęczniewski, Vaghawan Ojha
Apr 25, 2017
- PIO-30: Scala 2.11 support
- PIO-30: Spark 2 support
- PIO-49: Elasticsearch 5 support
- PIO-30, PIO-49: Flexible build system
- PIO-47, PIO-51: Removal of engine manifests
- PIO-49: Modularized storage backend modules
- PIO-45: Self cleaning data source
- PIO-25:
pio-start-all
will no longer start PostgreSQL if it is not being used. - PIO-47, PIO-51:
pio build
no longer requires access to the metadata repository.pio
commands will now accept an optional--engine-dir
parameter if you want to runpio build
,pio train
, orpio deploy
outside of an engine directory. This is an interim solution before an engine registry feature becomes available in the future. - PIO-49: PostgreSQL JDBC driver is no longer bundled with the core assembly. If you are using PostgreSQL, you must download the JDBC driver and update your configuration to point to the correct JDBC driver file.
- PIO-54: New generated access keys will no longer start with a
-
character.
- PIO-28: Code refactoring of the command line interface. It is now possible to develop new interfaces that perform the same functionalities provided by the CLI.
- PIO-53: Integration tests can now be tied to every single Git commit, without the need to update the official test Docker image.
- The meta data and model data access object methods are now public and marked as Core Developer API.
The following contributors have spent a great deal of effort to bring to you this release:
Ahmet DAL, Alexander Merritt, Amy Lin, Bansari Shah, Chan Lee, Chris Woodford, Daniel Gabrieli, Dennis Jung, Donald Szeto, Emily Rose, Hari Charan Ayada, infoquestsolutions, Jonny Daenen, Kenneth Chan, Laertis Pappas, Marcin Ziemiński, Naoki Takezoe, Rajdeep Dua, Shinsuke Sugaya, Pat Ferrel, scorpiovn, Suneel Marthi, Steven Yan, Takahiro Hagino, Takako Shimamoto
Oct 7, 2016
- Make SSL optional
- Merge ActionML fork
- First Apache PredictionIO release
Aug 5, 2016
- changed version id so artifacts don't conflict with naming in the Salesforce sponsored project.
- bug fix in memory use during moving window event trim and compaction EventStore data.
- update install.sh script for single line installs with options that support various combinations required by some templates.
April 11, 2015
For a detailed list of commits check this page
- Upgrade components for install/runtime to Hbase 1, Spark 1.5.2 PIO still runs on older HBase and Spark back to 1.3.1, upgrading install of Elaticsearch to 1.5.2 since pio run well on it but also runs on older versions.
- Support for maintaining a moving window of events by discarding old events from the EventStore
- Support for doing a deploy without creating a Spark Context
March 26, 2016
- Upgrade components for install/runtime to Hbase 1.X, Spark 1.5.2 PIO still runs on older HBase and Spark back to 1.3.1, upgrading install of Elasticsearch to 1.5.2 since pio run well on it but also runs on older versions.
- Support for maintaining a moving window of events by discarding old events from the EventStore
- Support for doing a deploy without creating a Spark Context
October 14th, 2015
Release Notes have been moved to Github and you are reading them. For a detailed list of commits check this page
- Support batches of events sent to the EventServer as json arrays
- Support creating an Elasticsearch StorageClient created for an Elasticsearch cluster from variables in pio-env.sh
- Fixed some errors installing PredictionIO through install.sh when SBT was not correctly downloaded
July 16th, 2015
Release Notes
- Support event permissions with different access keys at the Event Server interface
- Support detection of 3rd party Apache Spark distributions
- Support running
pio eval
withoutengine.json
- Fix an issue where
--verbose
is not handled properly bypio train
May 20th, 2015
Release Notes
- Add support of developing prediction engines in Java
- Add support of PostgreSQL and MySQL
- Spark 1.3.1 compatibility fix
- Creation of specific app access keys
- Prevent a case where
pio build
accidentally removes PredictionIO core library
April 14th, 2015
Release Notes
- Channels in the Event Server
- Spark 1.3+ support (upgrade to Spark 1.3+ required)
- Webhook Connector support
- Engine and Event Servers now by default bind to 0.0.0.0
- Many documentation improvements
March 17th, 2015
Release Notes
- Improved
pio-start-all
- Fixed a bug where
pio build
failed to set PredictionIO dependency version for engine templates
March 4th, 2015
Release Notes
- E-Commerce Recommendation Template which includes 1) out-of-stock items support 2) new user recommendation 3) unseen items only
- Complementary Purchase Template for shopping cart recommendation
- Lead Scoring Template predicts the probability of an user will convert in the current session
pio-start-all
,pio-stop-all
commands to start and stop all PredictionIO related services
Feb 10th, 2015
Release Notes
- New engine template - Product Ranking for personalized product listing
- CloudFormation deployment available