Kubeflow Slackbot on AWS lambda

Template for creating modals to trigger kubeflow pipelines. To create a modal you simply need to define the blocks in yaml modals/my_modal.yaml; see modals/ for examples. You'd also have to create the corresponding slash command in your slack bot settings.

An example of a modal that attaches to a /kfp-predict slash command and triggers a kubeflow pipeline called predict.

# modals/predict-modal.yaml
name: predict
title: Predict
slash_command: /kfp-predict
channel:  mlops # if empty, it would send the message to the user's slackbot
validate_args_func: validate_predict_args
kfp:
  pipeline_name: predict
  experiment_name: Evaluation
blocks:
  - name: model_version
    type: str
  - name: dataset-id
    display_name: Dataset ID
    type: int
  - name: output_format
    choices: [json, yaml]
  - name: output_name
    optional: true

You can optionally write a validation function in validation.py for checking the input arguments for errors and refer to the name of the function in validate_args_func:. Once deployed, this should create a modal that is called with /kfp-predict in the mlops channel where the bot is deployed.

Install serverless and edit serverless.yaml. The main things you need to edit are:

    SLACK_BOT_TOKEN: ${file(config.${opt:stage, 'dev'}.json):SLACK_BOT_TOKEN}
    SLACK_SIGNING_SECRET: ${file(config.${opt:stage, 'dev'}.json):SLACK_SIGNING_SECRET}
    KUBECONFIG: /tmp/kubeconfig
    CLUSTER_NAME: my-k8s-cluster
    REGION: eu-west-2
    BASE_URL: "https://kfp.mydomain.com/"

see serverless variables for more information about defining environment variables.

# apply roles to cluster
kubectl apply -f roles.yaml

# install plugin
serverless plugin install -n serverless-python-requirements

# deploy
serverless deploy

Copy the url generated and use it as a webhook for your slash command.

Serverless Usage

Deployment

In order to deploy the example, you need to run the following command:

$ serverless deploy

After running deploy, you should see output similar to:

Serverless: Packaging service...
Serverless: Excluding development dependencies...
Serverless: Creating Stack...
Serverless: Checking Stack create progress...
........
Serverless: Stack create finished...
Serverless: Uploading CloudFormation file to S3...
Serverless: Uploading artifacts...
Serverless: Uploading service aws-python.zip file to S3 (711.23 KB)...
Serverless: Validating template...
Serverless: Updating Stack...
Serverless: Checking Stack update progress...
.................................
Serverless: Stack update finished...
Service Information
service: aws-python
stage: dev
region: us-east-1
stack: aws-python-dev
resources: 6
functions:
  api: aws-python-dev-hello
layers:
  None

Invocation

After successful deployment, you can invoke the deployed function by using the following command:

serverless invoke --function hello

Which should result in response similar to the following:

{
    "statusCode": 200,
    "body": "{\"message\": \"Go Serverless v2.0! Your function executed successfully!\", \"input\": {}}"
}

Local development

You can invoke your function locally by using the following command:

serverless invoke local --function hello

Which should result in response similar to the following:

{
    "statusCode": 200,
    "body": "{\"message\": \"Go Serverless v2.0! Your function executed successfully!\", \"input\": {}}"
}

Bundling dependencies

In case you would like to include third-party dependencies, you will need to use a plugin called serverless-python-requirements. You can set it up by running the following command:

serverless plugin install -n serverless-python-requirements

Running the above will automatically add serverless-python-requirements to plugins section in your serverless.yml file and add it as a devDependency to package.json file. The package.json file will be automatically created if it doesn't exist beforehand. Now you will be able to add your dependencies to requirements.txt file (Pipfile and pyproject.toml is also supported but requires additional configuration) and they will be automatically injected to Lambda package during build process. For more details about the plugin's configuration, please refer to official documentation.