PyVisGraph

Visual connecting Python classes in Directed Acyclic Graphs using Litegraph interface.

Introduction

Start requires some work

1. Add to your class variables INPUT and OUTPUT in format ("arbitray node name","extension")

   from typing import ClassVar
   class Myclass:
       INPUT: ClassVar =  ("Data", ".csv")
       OUTPUT: ClassVar = ("Model", ".pickle")
   
       # your init arguments will be properties of model
       def __init__(self, my_awesome_arg: int):
           # type will be converted. Use planar input int, string, float and etc
           ...
   
   class InferenceClass:
       # several inputs and ouputs are allowed
       INPUT: list[ClassVar] =  [("Data", ".csv"),("Model", ".csv")]
       OUTPUT: ClassVar = [("Data", ".csv")]

   LiteGraph will make sure, that you won't connect input and output with unequal extensions.

2. Install everything you need

   git clone https://github.com/NMashalov/PyVisGraph# clone repo with git 
   poetry shell # creates virtual environment for project
   poetry install # install all dependencies
   pyvisgraph start # load example 

3. (Optional) Customize .py-vis-graph-preset.yaml to your needs

   import_paths:
   - module_name: scoring
     import_paths: ./test/nodes.py
   - module_name: scoring
     import_paths: ./test/nodes.py
   preset:
       output_settings:
           format: yaml
           mode: groupped
           dependency_name: needs
       graph_info:
           name: MyDag 
           group_name: "group"
           task_name: "task"
           properties:
           schedule: '0 * * * *'    

Demo via link

Quick start 🎈

You can add your own model by just simple dropping of .py file with models.

Functionality

• flexible organization of nodes defined via pydantic models
• transform LiteGraph linkage to Gitlab style aka depends_on

Features 🧰

• Use buttons validate and download for checking your workflow
• New nodes can be added via dropping .py file to canvas.
• Docstring and Pydantic fields description forms hint

Adding custom model

For defining models you need only pydantic.

Upload can be performed in several convenient way.

• explicitly define path

Configurate yaml of Graph

Can be configured with graph.conf.yaml.

Modifications

Under the hood of module is client-server with FastApi. This helps to couple Javascript and Python.

All modifications can be performed as simple as

References

• FastApi
• Litegraph
• Pydantic