Create a pipeline

A pipeline can be created from a solution template in the app or through the API.

A pipeline have following properties:

a logical key identifier of maximum 32 char (required)
a link to a template (required)
a set of variables (optional)
a list of steps (required)

Using Python

To create or update a pipeline you can use a JSON definition or define it from code directly. Below here is a sample script to create a pipeline and its equivalent JSON definition

import wizata_dsapi

pipeline = wizata_dsapi.Pipeline(
    key="my_pipeline",
    template_id=uuid.UUID('639e7537-0d80-4056-9670-06a63870d386'),
    variables={
        "var1": "string",
        "from": "relative",
        "to": "relative"
    }
)
pipeline.add_query(
    request=wizata_dsapi.Request(
        datapoints=['dp1','dp2','dp3'],
        interval=60000,
        agg_method='mean',
        start="@from",
        end="@to"
    ),
    df_name="query_output"
)
pipeline.add_transformation(
    script="my_script",
    input_df_names=['query_output'],
    output_df_names=['transform_data']
)
pipeline.add_model(
    config=wizata_dsapi.MLModelConfig(
        model_key='my_model_key',
        train_script='linear_reg_sc',
        features=['dp1', 'dp2'],
        target_feat='dp3',
    ),
    input_df='transform_data',
    output_df='predicted_data'
)
pipeline.add_writer(
    config=wizata_dsapi.WriteConfig(
        datapoints={
            'predicted_dp': 'dp_output'
        }
    ),
    input_df='predicted_data'
)
pipeline.add_plot(
    df_name='predicted_data',
    script='describe_df'
)
wizata_dsapi.api().upsert_pipeline(pipeline)

{
    "key": "my_pipeline",
    "templateId": "639e7537-0d80-4056-9670-06a63870d386",
    "steps": [
      {
        "type": "query",
        "config": {
          "equipments_list": [
            {
              "id": null,
              "datapoints": [
                "dp1",
                "dp2",
                "dp3"
              ],
              "shift": "0s"
            }
          ],
          "timeframe": {
            "start": "@from",
            "end": "@to"
          },
          "aggregations": {
            "agg_method": "mean",
            "interval": 60000
          },
          "filters": {},
          "options": {}
        },
        "inputs": [],
        "outputs": [
          {
            "dataframe": "query_output"
          }
        ]
      },
      {
        "type": "script",
        "config": {
          "function": "my_script"
        },
        "inputs": [
          {
            "dataframe": "query_output"
          }
        ],
        "outputs": [
          {
            "dataframe": "transform_data"
          }
        ]
      },
      {
        "type": "model",
        "config": {
          "train_script": "linear_reg_sc",
          "train_test_split_pct": 1,
          "train_test_split_type": "ignore",
          "target_feat": "dp3",
          "features": [
            "dp1",
            "dp2"
          ],
          "model_key": "my_model_key",
          "output_property": "result",
          "function": "predict"
        },
        "inputs": [
          {
            "dataframe": "transform_data"
          }
        ],
        "outputs": [
          {
            "dataframe": "predicted_data"
          }
        ]
      },
      {
        "type": "write",
        "config": {
          "datapoints": {
            "predicted_dp": "dp_output"
          }
        },
        "inputs": [
          {
            "dataframe": "predicted_data"
          }
        ],
        "outputs": []
      }
    ],
    "variables": {
      "var1": "string",
      "from": "relative",
      "to": "relative"
    }
}

You can therefore alternatively create or update your pipeline directly from a JSON.

import wizata_dsapi
import json

with open('your_pipeline.json', 'r') as f:
    pipeline_dict = json.load(f)
pipeline = wizata_dsapi.Pipeline()
pipeline.from_json(pipeline_dict)
wizata_dsapi.api().upsert_pipeline(pipeline)

Using UI

Navigate to AI Lab > Pipelines and click on 'Add' to create a new pipeline.