Using the ORM

pybana provides a simple ORM for manipulating kibana saved objects.

The ORM was implemented to ease the automatic creation/update of kibana objects. For instance:

  • If you’ve added an access-control layer on top of kibana to handle multi-tenancy, you may want to automate the creation of kibana indexes and the default index-pattern.
  • If an index-pattern correspond to a table defined somewhere else (like a sql table), you may want to automate the creation of index-pattern.
  • If a dashboard is defined in another database (like a sql db), you may want to delete the kibana object if the sql object is deleted.

Initializing kibana

A kibana server instance performs several checks when it starts:

  1. Create if it does not exists a .kibana index on elasticsearch. pybana does mimic this behaviour.
  2. Create a Config document.
    • This document has the following id: config:${ELASTICSEARCH_VERSION} (example: config:6.7.1)
    • It contains a config field which stores:
      • defaultIndex. The identifier of the default index
      • All the settings you can configure in the “Advanced settings” menu. The official documentation provide a full list of available options.
    • You may create programmatically this document using the Kibana.init_config api.
  3. Configure the default index-pattern. To do it programmatically, you can use the Kibana.update_or_create_default_index_pattern api.


For now, four models have been implemented:

  • IndexPattern
  • Search
  • Visualization
  • Dashboard


from elasticsearch import Elasticsearch
from elasticsearch_dsl import connections
from pybana import Kibana, Visualization, Dashboard

# Instantiate a connection to a Elasticsearch cluster
elastic = Elasticsearch()

# Add this connection as the default connection for elasticsearch_dsl
connections.add_connection("default", elastic)

# Instantiate a kibana client to the default index
kibana = Kibana()

# Instantiate a kibana client to a custom index
kibana = Kibana(".kibana_tenant1")

# Init config (does nothing if the config is already here)

# Init config (does nothing if the config is already here)

# Search for dashboards
dashboards = kibana.dashboards()
# Here dashboards is only have a `elasticsearch_dsl.Search`

# Iterate over the first 10 dashboards
for dashboard in dashboards:

# Iterate over all the dashboards
for dashboard in dashboards.scan():

# Get one dashboard
dashboard = kibana.dashboard("7b12e580-dae6-11e9-94be-2b2f7d5f3e45")

# Fetch all the associated visualization to this dashboard
visualizations = dashboard.visualizations()

# Get one visualization
visualization = visualizations[0]

# Deserialize the visState

# Get the search associated to the visualization (raise an error if ther's not)
search = visualization.related_search()

# Get the index pattern associated to a visualization (go through the search if there's one)
index_pattern = visualization.index()

# Get the index pattern associated to a search
index_pattern = search.index()

# Some objects fields are serialized in json like Dashboard.panelsJSON.
# Some of those fields can be directly accessed like the example below.
# The deserialization is cached in order to keep good performances

# [{'gridData': {'x': 0, 'y': 0, 'w': 24, 'h': 15, 'i': '1'},...

# {'darkTheme': False, 'useMargins': True, 'hidePanelTitles': ...}

# {'title': 'My viz', 'type': ...}

# {}

# [{'name': '_id', 'type': 'string', 'count': 0, 'scripted': Fa...}...]

# {'foo': {'id': 'number', 'params': {'pattern': '0.0'}}}