Skip to content

pyiron/pyironFlow

Repository files navigation

pyironFlow

Visual Programming Interface

The visual programmming interface pyironflow or PyironFlow is a gui skin based on ReactFlow that currently works on top of pyiron_workflow. Theoretically, one could currently pack pyiron_base jobs into nodes for execution. The gui could also be extended to pack the workflow graph (extracted from the gui using get_workflow()) into a pyiron_base job for execution. An existing code-based workflow graph can be packed into the gui using PyironFlow([wf]) where wf is the existing graph.

Example of a multiscale simulation

Problem definition:

Gui1

Constructing the workflow:

Gui2

Execution:

Gui3

Viewing source code:

Gui4

Visualization of FEM meshes using PyVista:

Gui5

Atomistic sub-workflow:

Gui6

Continuum sub-workflow:

Gui7

Planned/desired/in-development features

  • Dynamically add nodes from the notebook to the node library (currently being polished by Julius Mittler)
  • Dynamically make macros from the gui using a box selection and add them to the node library (currently in development by Julius Mittler)
  • Automatic positioning of nodes in a nice way on (re-)instantiation (current positioning is passable, but room for improvement exists)
  • Caching based on a hash-based storage. Currently caching does not work at all in the gui since the workflow graph is reinstantiated everytime there's a change in the gui. This could be theoretically fixed by comparing the dictionary of the existing workflow to the new dictionary from the gui. But this would be a painstaking process and easily broken. A better way would be to work with a hash-based storage which would reload outputs when the inputs of a node have not changed. However this would have to be implemented in pyiron_workflow and not in the gui side of things.
  • Edit the source code of nodes from within the gui. Maybe something that uses the FileReader api. Need to look into this further.

See the demo.ipynb jupyter notebook for brief discussion of the key ideas, the link to pyiron_workflows and a few toy application of the xyflow project.

Binder