Skip to contents

Saving workflows

saving_workflows.Rmd

How to save of workflow

Since workflows can contain quite a few parameter declarations, you may need to save them into specific objects for portability. This is exactly what you can do with save_workflow() :


myflow_pitch <- ai_workflow() |>
  set_connector(connector = "ollama") |>
  set_model(model_name = "llama3.2:latest") |>
  set_audience(audience = "Marketing professionals") |>
  set_temperature(temperature = 0.6) |>
  set_processing_skill(processing_skill = "write_pitch") |>
  set_n_predict(3000)
#> → Default IP address has been set to 127.0.0.1.
#> → Default port has been set to 11434.

res <- myflow_pitch |> save_workflow(filepath = "wflow_pitch.json")
#> → The workflow has been saved to wflow_pitch.json.

This will save the configuration of your workflow in a JSON file.

How to reload a workflow

The reverse operation uses load_workflow() :


myflow_pitch_reloaded <- load_workflow(filepath = "wflow_pitch.json")
#> → The workflow has been successfully reloaded from wflow_pitch.json.

Note that JSON does not ensure the loaded workflow is identical to the one you saved before. For example, figures stored as double may be reloaded as integers. This should not lead to issues when running the workflow.

Let’s clean-up afterwards:


unlink("wflow_pitch.json")