Generate a workflow outline from a BPMN diagram

[josephjclark]

This is a high level project proposal for the gen-ai / apollo server. You should contact @josephjclark before getting started.

Overview

This project is to generate a workflow.json (or maybe a project.yaml??) which captures the flow specified in a BPMN diagram: https://en.wikipedia.org/wiki/Business_Process_Model_and_Notation

TODO: find a better example image

Details

This is a different sort of challenge because it requires the parsing of an image, rather than text.

We must identify the steps in a BPNM diagram and how they are connected, and generate a workflow as best we can.

Maybe this can be done in stages - like using one module to pull out the names and connections of steps in the BPMN, and another to generate a workflow.json based on this (this may even not be AI).

We don't need the actual steps of the workflow to be generated. We just need the structure. Comment stubs would be nice - they could be later be passed downstream to a job code generator.

I don't know if we need a workflow.json or a project.yaml. At the time of writing the CLI only supports workflow.json so that's a much easier thing to test. But to import into Lightning and visualise it, we need a project.yaml. tricky one.

Inputs

We (openfn) should provide a set of sample BPM diagrams and example workflows to help the training process and validate the model.