Yes it’s possible to not use Jupyter, but you’ll still need to use a REST API so that the front-end (Azle) can communicate with the back-end (Python/TensorFlow).

You can use Flask instead if you like. Azle can call any back-end code as long as the endpoint is made available. This tutorial used Jupyter’s Kernel Gateway to expose those endpoints, but Flask can also be used.

All the code for this tutorial is on GitHub so you shouldn’t be missing anything.

See if you can get Flask up-and-running and let me know.