Txt2BPMN

This is a Tool to generate bpmn-diagrams from business process descriptions with the use of the OpenAI LLM gpt-4o-mini.

Features:

1. Generate BPMN Diagrams
2. Edit Diagrams with bpmnjs Editor
3. Export Diagrams as SVG or as bpmn-file

How to use the app with Docker:

(You will need an open ai api key and docker installed)

1. Pull the image from GitHub Container Registry:

docker pull ghcr.io/lkmnch/text2bpmn:latest

2. Run the container and provide your OpenAI API key as an environment variable:

docker run -e OPENAI_API_KEY={your-api-key} -p 3000:3000 ghcr.io/lkmnch/text2bpmn:latest

3. Access the app in your browser at http://localhost:3000.

4. Enter your business process description and generate your BPMN diagram.

How it works:

1. Put in the Description of your Process

   

2. The Description is send to the LLM as part of a Prompt which returns a JSON including the Events, Tasks, Gateways and Sequence Flows of the Process.

   jsonProcess: {
  tasks: [
    { id: 'task_1', name: 'Clerk decides on shipment type' },
    { id: 'task_2', name: 'Warehouse worker packages goods' },
    { id: 'task_3', name: 'Request quotes from carriers' },
    ...
  ],
  gateways: [
    {
      id: 'gateway_1',
      name: 'Parallel Gateway',
      type: 'parallelGateway'
    },
    {
      id: 'gateway_2',
      name: 'Exclusive Gateway (mode of delivery)',
      type: 'exclusiveGateway'
    },
    ...
  ],
  start_events: [ { id: 'start_1', name: 'Goods to ship' } ],
  end_events: [ { id: 'end_1', name: 'Goods available for pick' } ],
  sequence_flows: [
    {
      id: 'flow_1',
      sourceRef: 'start_1',
      targetRef: 'gateway_1',
      name: ''
    },
    {
      id: 'flow_2',
      sourceRef: 'gateway_1',
      targetRef: 'task_1',
      name: ''
    },
    {
      id: 'flow_3',
      sourceRef: 'gateway_1',
      targetRef: 'task_2',
      name: ''
    },
   ...
  ]
}

5. The Process is then showen in the bpmnjs Editor, where you can refine the Diagram.

   

6. After editing the diagram you can export it.

   

Problems:

• Sometimes some Sequenceflows will not be generated from the LLM