Kedro, running inference on user input

Question

I have a pipeline with the model I want to use. Outside of the project, I have an app.py file where I'm going to create the UI/UX for my users to run my model. Right now I'm just using a sample string but later on, you can imagine that there will be a textbox for users to type.

How can I pass the user input as an input to the pipeline? I though I would be able to do so with the kedro.framework.session.session.KedroSession as seen in the code below, but doing so results in the error ValueError: Pipeline input(s) {'user-input'} not found in the DataCatalog

from kedro.framework.session import KedroSession
from kedro.framework.startup import bootstrap_project
from kedro.io import MemoryDataSet
import os

bootstrap_project("<project path>")

user_input = "this is a sample text"
user_input = MemoryDataSet(user_input)


with KedroSession.create("project") as session:
    output = session.run(
        "nlp-pipeline", 
        from_inputs={
            "user-input": user_input
        }
    )

print(output)