A reactive notebook for Python — run reproducible experiments, query with SQL, execute as a script, deploy as an app, and version with git. Stored as pure Python. All in a modern, AI-native editor.
Python artificial-intelligencedagdata-sciencedata-visualizationdataflowdeveloper-toolsmachine-learningnotebookspipelinepythonreactivesqlweb-app
# /// script
# requires-python = ">=3.9"
# dependencies = [
# "db-dtypes==1.3.0",
# "google-cloud-bigquery==3.25.0",
# "marimo",
# "protobuf==5.28.2",
# ]
# ///
import marimo
__generated_with = "0.25.2"
app = marimo.App(width="medium")
@app.cell
def _():
import marimo as mo
import os
from google.cloud import bigquery
return bigquery, mo, os
@app.cell(hide_code=True)
def _(mo):
mo.md("""
# Google Cloud BigQuery
""")
@app.cell(hide_code=True)
def _(mo):
# Configuration
credentials = mo.ui.text(placeholder="path/to/creds.json")
mo.md(
f"""
## **⚙ Configuration**
This app requires a Google Cloud Platform account and a bucket to access.
Authenticate with `gcloud auth login`, or provide a path to a credentials
file: {credentials}
"""
)
return (credentials,)
@app.cell(hide_code=True)
def _(mo):
project = mo.ui.text(label="gcloud project")
project
return (project,)
@app.cell(hide_code=True)
def _(bigquery, credentials, mo, os, project):
# Set up client
if credentials.value:
os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = credentials.value
else:
os.environ.pop("GOOGLE_APPLICATION_CREDENTIALS", None)
if project.value is not None:
_project = project.value
else:
_project = os.env.get("GCLOUD_PROJECT")
mo.stop(not _project, mo.md("☝️ Provide a gcloud project."))
client = bigquery.Client(project=_project)
datasets = list(client.list_datasets())
return client, datasets
@app.cell
def _(datasets: list, mo):
# Dataset selection
selected_dataset = mo.ui.dropdown(
label="Select dataset", options=[d.dataset_id for d in datasets]
)
selected_dataset
return (selected_dataset,)
@app.cell
def _(client, mo, selected_dataset):
mo.stop(not selected_dataset.value)
dataset = client.dataset(selected_dataset.value)
return (dataset,)
@app.cell
def _(client, dataset, mo):
# Table selection
tables = list(client.list_tables(dataset))
selected_table = mo.ui.dropdown(
label="Select table", options=[t.table_id for t in tables]
)
selected_table
return (selected_table,)
@app.cell
def _(client, dataset, mo, selected_table):
results = client.list_rows(dataset.table(selected_table.value), max_results=10)
mo.ui.table(results.to_dataframe(), selection=None)
if __name__ == "__main__":
app.run()