The Search & Recommendations REST API implements the REST API for Guru Cue Search & Recommendations Engine for:

• providing data about products and consumers,
• providing data about events,
• retrieving recommendations and search results.

The minimum required JDK version to build the API with is 1.8.

Before you start building the API put into the libs directory these Guru Cue Search & Recommendations libraries:

• database,
• data-provider-jdbc,
• data-provider-postgresql.

Perform the build using gradle. If you don't have it installed, you can use the gradle wrapper script gradlew (Linux and similar) or gradlew.bat (Windows).

Before running the API for the first time a PostgreSQL database must be created and configured. Please refer to the documentation of the data-provider-postgresql library on how to create a database.

Database connection parameters are configured using environment variables:

• PROVIDER_POSTGRESQL_URL in the form jdbc:postgresql://<hostname>/<database>,
• PROVIDER_POSTGRESQL_USERNAME with the username to connect to the database,
• PROVIDER_POSTGRESQL_PASSWORD with the password to connect to the database.

The API uses run-time compilation for the blenders subsystem, so Java SE Development Kit, version 8 or more, must be installed and used to run the API. Put the blenders' files in source form into the /opt/GuruCue/blender directory. First-level subdirectories must use the username field of the partner database table for their names. These directories in turn contain filter definitions for all blenders of respective partners (blenders for recommenders stored in the recommenders subdirectory, and blenders for searching stored in the searchers subdirectory).

The API must be deployed in a Java EE container such as Apache Tomcat. Simply copy the war file that is the result of the build process, into the Java EE deployment directory.

To be fully functional the API expects recommender engines to be configured and available. API communicates with recommender engines via RMI. The hostnames of recommender engines (needed by the RMI subsystem) must be configured in the database table recommender. Each recommender engine can support many recommenders, which are exposed as a type of built-in filters to the blenders subsystem. These are configured in the partner_recommender database table.

Please refer to the API documentation for details.