udata-recommendations

This plugin acts as a bridge between uData and a recommendation system.

In our case (data.gouv.fr), it's a set of scripts living here https://github.com/etalab/piwik-covisits.

Recommendations are stored on datasets. Recommendations can come from various sources and are stored in a descending order, according to the provided score (from 1 to 100). The top recommendations are displayed at the bottom on the dataset page.

Compatibility

udata-recommendations requires Python 3.7+ and uData.

Installation

Install uData.

Remain in the same virtual environment (for Python).

Install udata-recommendations:

pip install udata-recommendations

Modify your local configuration file of udata (typically, udata.cfg) as following:

PLUGINS = ['recommendations']
RECOMMENDATIONS_SOURCES = {
    'source-name': 'https://path/to/recommendations.json',
    'other-source': 'https://path/to/other/recommendations.json',
}
RECOMMENDATIONS_NB_RECOMMENDATIONS = 4

• RECOMMENDATIONS_SOURCES: A key-value dictionary of recommendation sources and URLs to fetch. Default: {}
• RECOMMENDATIONS_NB_RECOMMENDATIONS: The maximum number of recommendations to display on the dataset page. Default: 4

Usage

Adding recommendations

You can fetch and store recommendations as a task, using your configuration in RECOMMENDATIONS_SOURCES, on a schedule if needed. By default, previous recommendations are cleaned before the importing new ones, but you're in control.

udata job run recommendations-add
# Don't clean each source before importing new recommendations
udata job run recommendations-add should_clean=false

Deleting recommendations

To clean all recommendations, you can run the following task.

udata job run recommendations-clean

Expectations

This plugin expects the following format to provide datasets recommendations:

[
  {
    "id": "dataset-id",
    "recommendations": [
      {
        "id": "dataset-slug-1",
        "score": 100
      },
      {
        "id": "5ef1fe80f50446b8f41ba691",
        "score": 1
      }
    ]
  },
  {
    "id": "dataset-id2",
    "recommendations": [
      {
        "id": "5ef1fe80f50446b8f41ba691",
        "score": 50
      }
    ]
  }
]

Dataset IDs can be IDs or slugs. Scores should be between 1 and 100, inclusive. You can validate your JSON using a JSON Schema.