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 = ">=3.13"
# dependencies = [
# "fsspec==2026.2.0",
# "marimo>=0.19.9",
# "obstore==0.8.2",
# "python-dotenv==1.2.1",
# "requests==2.32.5",
# "s3fs==2026.2.0",
# ]
# ///
import marimo
__generated_with = "0.25.2"
app = marimo.App(width="medium")
@app.cell
def _():
import marimo as mo
from fsspec.implementations.local import LocalFileSystem
from fsspec.implementations.github import GithubFileSystem
import s3fs
import obstore
import os
from dotenv import load_dotenv
load_dotenv()
return GithubFileSystem, LocalFileSystem, os, s3fs
@app.cell
def _(LocalFileSystem):
local = LocalFileSystem(asynchronous=True)
@app.cell
def _(os):
from obstore.store import S3Store
access_key_id = os.environ.get("CLOUDFLARE_ACCESS_KEY_ID")
secret_access_key = os.environ.get("CLOUDFLARE_SECRET_ACCESS_KEY")
url = os.environ.get("CLOUDFLARE_MARIMO_URL")
store = S3Store.from_url(
f"{url}/marimo-artifacts",
access_key_id=access_key_id,
secret_access_key=secret_access_key,
)
return (url,)
@app.cell
def _(GithubFileSystem):
marimo_repo = GithubFileSystem(org="marimo-team", repo="marimo")
return (marimo_repo,)
@app.cell
def _(marimo_repo):
print(marimo_repo.read_text("github://AGENTS.md"))
@app.cell
def _(os, s3fs, url: None | str):
s3_client = s3fs.S3FileSystem(
endpoint_url=url,
key=os.getenv("CLOUDFLARE_ACCESS_KEY_ID"),
secret=os.getenv("CLOUDFLARE_SECRET_ACCESS_KEY"),
)
if __name__ == "__main__":
app.run()