[SOLVED] Hey! Can I inject a custom dataset when using kedro boot? @Takieddine Kadiri

Hi, you can inject any object into any dataset. What are trying to achieve ? you want to inject data into kedro project or kedro package ? You need to embed your app inside kedro project or want to keep it external ? Here is an example of injecting data from external application into kedro project :

from kedro_boot.app.booter import boot_project
from kedro_boot.framework.compilation.specs import CompilationSpec
session = boot_project(project_path="<your_project_path>", compilation_specs=[CompilationSpec(inputs=["your_dataset_name"])], kedro_args={"pipeline": "your_pipeline", conf_source="<your_conf_source>", ....})
run_results = session.run(inputs={"your_dataset_name": your_injected_data})

For now try installing kedro boot directly from the main branch. We’ll release the new version in the next days

Hi! Thank you for the response. I love the plugin! I am running kedro boot in embedded mode like the spaceflights_fastapi example. I want to be able to send a request a request that specifies the path that defines the path of a

PartitionedDataset

#catalog.yml
documents:
  type: PartitionedDataset
  path: data/01_raw/ <-want to specify this in request
  dataset:
    type: embeddings.extras.datasets.document_dataset.DocumentDataSet <- this is a custom dataset
    load_args: 
...

Is this possible?

You’re using kedro-boot 0.1.0 with kedro < 0.19 ?

Yeah using

kedro==0.18.14
kedro-boot==0.1.0

You need to inject your data through

template_params

Session.run(template_params={path_param: <your_path>})

In your catalog.yml try to parametrize your path attribute:

Path: ‘[[path_param]]’

aaah okey okey that makes sense. Im not "injecting a dataset" rather im just changing the catalog based on request. Going to try this

Thank you for the feedback ! Enjoy 🥳