What we have so far:

all custom GT is downloaded and processed by NextFlow scripts based on the workflows we recommend on the website
the first relevant information (CER, processing time) is extracted and stored for all workflows in one JSON file over at https://github.com/mweidling/quiver-back-end/blob/implement-first-complete-workflow/data/workflows.json
the evaluation workspaces are stored to https://github.com/mweidling/quiver-back-end/tree/implement-first-complete-workflow/workflows/results
the whole process can be triggered either by make benchmarks or quiver-ocrd benchmarks

To be done in this sprint:

[x] decouple OCR workflow from eval workflow (they are currently processed in one step)
[x] refactor the code and merge it to the main repo
[x] add all the recommended workflows (we currently one offer the "minimal" workflow with ocrd-tesserocr)