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
import secrets
import subprocess
import time
import modal
app = modal.App(
image=modal.Image.debian_slim()
.pip_install("marimo>=0.12.8", "modal>=0.73.166")
.add_local_dir("nbs", remote_path="/root/nbs")
)
TOKEN = secrets.token_urlsafe(16)
PORT = 2718
@app.function(max_containers=1, timeout=1_500, gpu="t4")
def run_marimo(timeout: int):
with modal.forward(PORT) as tunnel:
marimo_process = subprocess.Popen(
[
"marimo",
"edit",
"--headless",
"--host",
"0.0.0.0",
"--port",
str(PORT),
f"--token-password",
TOKEN,
"nbs/notebook.py",
],
)
print(f"Marimo available at => {tunnel.url}?access_token={TOKEN}")
try:
end_time = time.time() + timeout
while time.time() < end_time:
time.sleep(5)
print(
f"Reached end of {timeout} second timeout period. Exiting..."
)
except KeyboardInterrupt:
print("Exiting...")
finally:
marimo_process.kill()
@app.local_entrypoint()
def main():
run_marimo.remote(1000)