Argo Workflow: DAG

In a previous article, we described steps that compose a set of workflow templates into a sequential execution by grouping them. Here, we will take a look at DAG (Directed Acyclic Graph), a more general execution of the set where the order of workflow template executions is specified by dependencies.

We will reuse the working example in the steps article. We extend it such that the extracted text will be translated into other languages.

Here is the workflow template that we will use.

---
apiVersion: argoproj.io/v1alpha1
kind: WorkflowTemplate
metadata:
  name: translate-ai
spec:
  templates:
  - name: pil-image-checker
    inputs:
      parameters:
      - name: filename
        value: "demo_image.jpg"
      artifacts:
        - name: inputfile
          path: "/tmp/{{inputs.parameters.filename}}"
          s3:
            key: "input/{{inputs.parameters.filename}}"
    outputs:
      parameters:
      - name: inputfile
        valueFrom:
          default: "demo_image"
          path: "/tmp/inputname" 
    script:
      image: [container-repo]/translation-ai:latest
      command: [bash]
      source: |
        python3 verify_image_extension.py -i "/tmp/{{inputs.parameters.filename}}"
        echo "{{inputs.parameters.filename}}" | cut -f1 -d"." > /tmp/inputname
  - name: ocr-florence2
    inputs:
      parameters:
      - name: filename
        value: "demo_image.jpg"
      artifacts:
        - name: inputfile
          path: "/tmp/{{inputs.parameters.filename}}"
          s3:
            key: "input/{{inputs.parameters.filename}}"
    outputs:
      artifacts:
        - name: output-json
          path: "/tmp/ocr-output.json"
        - name: output-txt
          path: "/tmp/ocr-output.txt"
    script:
      image: [container-repo]/translation-ai:latest
      command: [bash]
      source: |
        python3 ocr_florence2.py -i "/tmp/{{inputs.parameters.filename}}" -o "/tmp/ocr-output"
  - name: translate-text
    inputs:
      parameters:
      - name: inputname
        value: "demo_image"
      - name: translator
        value: "google"
      - name: target-language
        value: "fr"
      artifacts:
        - name: inputfile
          path: "/tmp/input.txt"
    outputs:
      artifacts:
        - name: output_txt
          path: "/tmp/output.txt"
          s3:
            key: "output/{{inputs.parameters.inputname}}-{{inputs.parameters.translator}}-{{inputs.parameters.target-language}}.txt.tgz"
    script:
      image: [container-repo]/translation-ai:latest
      command: [bash]
      source: |…