Record Recommender

The Record Recommender creates recommendations for Invenio. By reading page views and downloads from Elasticsearch calculating the recommendations and storing them in Redis to be retrieve and displayed by Invenio.

Usage

The workflow to generate the recommendations is:

1. recommender fetch 24 to cache the last 24 weeks of page views and downloads from Elasticsearch.
2. recommender profiles 24 generate the user profiles from the page views and downloads. For the not logged in users profiles based on the ip-address and user agent are created.
3. recommender build 50 calculates the recommendations using 50 processes and stores them in the specified Redis server.

Alternative the recommendations can be automatically be fetched, the profiles generated and the recommendations calculated all this with one command:

recommender update_recommender 24 50 for the last 24 weeks and using 50 processes.

Configuration

The configuration file is expected to be in /etc/record_recommender.yml otherwise the path to the config file can be defined by using the command line option --config_path.

# Record-Recommender configuration.

elasticsearch:
  es_index: ['index-2014', 'index-2015', 'index-2016']
  es_user: user
  es_password:
  es_host: localhost
  es_port: 443

recommendation_version: 2

# Sentry connection string
sentry:

redis:
  host: localhost
  port: 6379
  db: 0
  prefix: 'Reco_1::'

cache:
  base_path: cache/
  cache_file_prefix: ''

logging:
  version: 1
  disable_existing_loggers: False
  formatters:
    simple:
      format: '%(asctime)s - %(name)s - %(levelname)s - %(message)s'
  handlers:
    console:
      class: logging.StreamHandler
      level: INFO
      formatter: simple
      stream: ext://sys.stdout
    sentry:
      class: raven.handlers.logging.SentryHandler
      level: WARN
      dsn:
    file:
      class : logging.FileHandler
      formatter: simple
      level: DEBUG
      filename: log_recommender.log
  loggers:
    record_recommender:
      level: DEBUG
      handlers: [console, file, sentry]
      propagate: no
    # elasticsearch:
    #   level: WARN
    #   handlers: [console]
    #   propagate: no
  root:
    level: ERROR
    handlers: [console, file, sentry]

Additional to the configuration options found in the config file environment variables, which overwrite the ones from the config file can be set.

• RECOMMENDER_ES_PASSWORD to set the Elasticsearch password.
• RECOMMENDER_SENTRY to set the Sentry connection string.

Command line

Usage: recommender [OPTIONS] COMMAND [ARGS]...

  Record-Recommender command line version.

Options:
  -v, --verbose           Enables verbose mode.
  -c, --config_path TEXT  Path to the configuration file.
  --help                  Show this message and exit.

Commands:
  debug               Debug the application and recommender.
  fetch               Fetch newest PageViews and Downloads.
  build               Calculate all recommendations.
  profiles            Number of weeks to build.
  update_recommender  Download and build the recommendations.

Debugging the Recommendations

As first step look into the created user profiles in the defined cache folder.

To explore the graph with all loaded data use the recommender debug command to get a interactive python shell.