Skip to content

Getting started with Kili CLI

What is Kili CLI

Kili CLI has been designed to run key actions on your projects with powerful commands. For the actions it supports, the CLI offers a more compact way to manage your projects than the Python SDK. Note that the Python SDK offers more options and may still be used for more complex project management tasks.

Authentication

  • Create and copy a Kili API key
  • Add the KILI_API_KEY variable in your bash environment (or in the settings of your favorite IDE) by pasting the API key value that you copied earlier:
export KILI_API_KEY='<you api key value here>'

Info

While launching commands, you can also provide you API key through the --api-key option. If you set your api key in the KILI_API_KEY environment variable and provide it once again through the --api-key option, Kili CLI will use the api key value provided in command options.

Usage

The main command is kili. It currently has only one subcommand project that entails all the commands for project management :

kili project [COMMAND]

Workflow example

Let's take an example where you want to start a project from scratch:

Create a project

To create an IMAGE project:

kili project create \\
            --interface path/to/interface.json \\
            --input-type IMAGE \\
            --title "Defect detection" \\
            --description "Bottle defects on production line"

Ouput:

Project successfully created. Id: cl4ljd3awc5gj0lpbb89nbcqg

List your projects

kili project list --max 10

Ouput:

title                                     id                           progress  description
----------------------------------------  -------------------------  ----------  -------------------------------
Defect detection                          cl4ljd3awc5gj0lpbb89nbcqg        0.0%  Bottle defects on production...
invoice NER                               cl3d43bzb0rl71mx4580mpbt1       92.5%  For intelligent document pro...

Import data to your project

To import data, provide a list of files or folders

kili project import \\
    reference_image.png datasets/defect_detection/ \\
    --project-id cl4ljd3awc5gj0lpbb89nbcqg \\
    --verbose

Ouput:

datasets/defect_detection/visit1.mp4    SKIPPED
datasets/defect_detection/visit2.mp4    SKIPPED
Paths either do not exist, are filtered out or point towards wrong data type for the project

567 files have been successfully imported

Import labels to your project

To import labels, provide a CSV file with two columns, separated by a semi-column:

  • external_id: external id for which you want to import labels.
  • json_response_path: paths to the json files containing the json_response to upload.

CSV file template

external_id;json_response_path 
asset1;./labels/label_asset1.json 
asset2;./labels/label_asset2.json

kili project label \\
    datasets/defect_detection/labels.csv \\
    --project-id cl4ljd3awc5gj0lpbb89nbcqg

Outputs:

82 labels have been successfully imported

If you have run a pre-annotation model, you can also import labels as predictions. These labels will be seen as pre-annotation in the labeling interface.

kili project label \\
    datasets/defect_detection/predictions.csv \\
    --project-id cl4ljd3awc5gj0lpbb89nbcqg \\
    --prediction \\
    --model-name YOLO-run-3

Outputs:

82 labels have been successfully imported

Get metrics of your project

kili project describe \\
    --project-id cl4ljd3awc5gj0lpbb89nbcqg

Ouput:

Title        Defect detection
Description  Bottle defects on production line

Dataset KPIs
------------
Total number of assets      567
Number of remaining assets  485
Skipped assets              0
Progress                    14.5%

Quality KPIs
------------
Project consensus           79
Project honeypot            22
Number of reviewed assets   76
Number of open issues       5
Number of solved issues     12
Number of open questions    2
Number of solved questions  9