write_jupyter_notebook.md

Example 9: Use dbt from a Jupyter Notebook

We also offer fal as an importable package to load in a Python environment to reference and use your dbt models and sources. You may want to do this to run some ad-hoc analysis on the data.

We start by importing fal into your project

from fal import FalDbt

Then instantiate a new FalDbt project with the dbt project information:

faldbt = FalDbt(profiles_dir="~/.dbt", project_dir="../my_project")

faldbt.list_sources()
# [['results', 'ticket_data_sentiment_analysis']]

faldbt.list_models()
# {
#   'zendesk_ticket_metrics': <RunStatus.Success: 'success'>, 
#   'stg_o3values': <RunStatus.Success: 'success'>, 
#   'stg_zendesk_ticket_data': <RunStatus.Success: 'success'>, 
#   'stg_counties': <RunStatus.Success: 'success'>
# }

Reference these objects as you would in a regular fal script, from the faldbt object:

sentiments = faldbt.source('results', 'ticket_data_sentiment_analysis')
# pandas.DataFrame
tickets = faldbt.ref('stg_zendesk_ticket_data')
# pandas.DataFrame

NOTE: You can use any other function available in the fal script runtime through the faldbt object.

Then you can just use the data in these Pandas DataFrames to analyze and plot an interesting graph:

def calc_weighted_label(row):
  val = 1 if row['label'] == 'POSITIVE' else -1
  return val * row['score']

joined = sentiments.merge(tickets, on='id')
joined['weighted_label'] = joined.apply(calc_weighted_label, axis=1).astype(float)

from matplotlib import pyplot as plt

joined.plot(y=['weighted_label'], x='created_at')
plt.show()

You can find the full code example here.