Create or update a workflow

Creating a workflow adds the content of a workflow template to the flow framework system index. You can provide workflows in JSON format (by specifying Content-Type: application/json) or YAML format (by specifying Content-Type: application/yaml). By default, the workflow is validated to help identify invalid configurations, including:

• Workflow steps requiring an OpenSearch plugin that is not installed.
• Workflow steps relying on previous node input that is provided by those steps.
• Workflow step fields with invalid values.
• Workflow graph (node/edge) configurations containing cycles or with duplicate IDs.

To obtain the validation template for workflow steps, call the Get Workflow Steps API.

You can include placeholder expressions in the value of workflow step fields. For example, you can specify a credential field in a template as openAI_key: '${{ openai_key }}'. The expression will be substituted with the user-provided value during provisioning, using the format ${{ <value> }}. You can pass the actual key as a parameter by using the Provision Workflow API or by using this API with the provision parameter set to true.

Once a workflow is created, provide its workflow_id to other APIs.

The POST method creates a new workflow. The PUT method updates an existing workflow. You can specify the update_fields parameter to update specific fields.

You can only update a complete workflow if it has not yet been provisioned.

Path and HTTP methods

POST /_plugins/_flow_framework/workflow
PUT /_plugins/_flow_framework/workflow/<workflow_id>

Path parameters

The following table lists the available path parameters.

Query parameters

Workflows are normally created and provisioned in separate steps. However, once you have thoroughly tested the workflow, you can combine the create and provision steps by including the provision query parameter:

POST /_plugins/_flow_framework/workflow?provision=true

When set to true, the Provision Workflow API is executed immediately following creation.

By default, workflows are validated when they are created to ensure that the syntax is valid and that the graph does not contain cycles. This behavior can be controlled with the validation query parameter. If validation is set to all, OpenSearch performs a complete template validation. Any other value of the validation parameter suppresses validation, allowing an incomplete/work-in-progress template to be saved. To disable template validation, set validation to none:

POST /_plugins/_flow_framework/workflow?validation=none

In a workflow that has not been provisioned, you can update fields other than the workflows field. For example, you can update the name and description fields as follows:

PUT /_plugins/_flow_framework/workflow/<workflow_id>?update_fields=true
{
  "name": "new-template-name",
  "description": "A new description for the existing template"
}

You cannot specify both the provision and update_fields parameters at the same time.

If a workflow has been provisioned, you can update and reprovision the full template:

PUT /_plugins/_flow_framework/workflow/<workflow_id>?reprovision=true
{
  <updated complete template>
}

You can add new steps to the workflow but cannot delete them. Only index setting, search pipeline, and ingest pipeline steps can currently be updated.

You can create and provision a workflow using a workflow template as follows:

POST /_plugins/_flow_framework/workflow?use_case=<use_case>&provision=true
{
    "create_connector.credential.key" : "<YOUR API KEY>"
}

The following table lists the available query parameters. All query parameters are optional. User-provided parameters are only allowed if the provision parameter is set to true.

Request body fields

The following table lists the available request fields.

Example request: Register and deploy an externally hosted model (YAML)

To provide a template in YAML format, specify Content-Type: application/yaml in the request header:

curl -XPOST "http://localhost:9200/_plugins/_flow_framework/workflow" -H 'Content-Type: application/yaml'

YAML templates permit comments.

The following is an example YAML template for registering and deploying an externally hosted model:

# This name is required
name: createconnector-registerremotemodel-deploymodel
# Other fields are optional but useful
description: This template creates a connector to a remote model, registers it, and
  deploys that model
# Other templates with a similar use case can be searched
use_case: REMOTE_MODEL_DEPLOYMENT
version:
  # Templates may be versioned by their authors
  template: 1.0.0
  # Compatibility with OpenSearch 2.12.0 and higher and 3.0.0 and higher
  compatibility:
  - 2.12.0
  - 3.0.0
# One or more workflows can be included, presently only provision is supported
workflows:
  provision:
    # These nodes are the workflow steps corresponding to ML Commons APIs
    nodes:
    # This ID must be unique to this workflow
    - id: create_connector_1
      # There may be multiple steps with the same type
      type: create_connector
      # These inputs match the Create Connector API body
      user_inputs:
        name: OpenAI Chat Connector
        description: The connector to public OpenAI model service for GPT 3.5
        version: '1'
        protocol: http
        parameters:
          endpoint: api.openai.com
          model: gpt-3.5-turbo
        credential:
          openAI_key: '12345'
        actions:
        - action_type: predict
          method: POST
          url: https://${parameters.endpoint}/v1/chat/completions
    # This ID must be unique to this workflow
    - id: register_model_2
      type: register_remote_model
      # This step needs the connector_id produced as an output of the previous step
      previous_node_inputs:
        create_connector_1: connector_id
      # These inputs match the Register Model API body
      user_inputs:
        name: openAI-gpt-3.5-turbo
        function_name: remote
        description: test model
    # This ID must be unique to this workflow
    - id: deploy_model_3
      type: deploy_model
      # This step needs the model_id produced as an output of the previous step
      previous_node_inputs:
        register_model_2: model_id
    # Since the nodes include previous_node_inputs these are optional to define
    # They will be added automatically and included in the stored template
    # Additional edges may also be added here if required for sequencing
    edges:
    - source: create_connector_1
      dest: register_model_2
    - source: register_model_2
      dest: deploy_model_3

Example request: Register and deploy a remote model (JSON)

To provide a template in JSON format, specify Content-Type: application/json in the request header:

curl -XPOST "http://localhost:9200/_plugins/_flow_framework/workflow" -H 'Content-Type: application/json'

The following JSON template is equivalent to the YAML template provided in the previous section:

{
  "name": "createconnector-registerremotemodel-deploymodel",
  "description": "This template creates a connector to a remote model, registers it, and deploys that model",
  "use_case": "REMOTE_MODEL_DEPLOYMENT",
  "version": {
    "template": "1.0.0",
    "compatibility": [
      "2.12.0",
      "3.0.0"
    ]
  },
  "workflows": {
    "provision": {
      "nodes": [
        {
          "id": "create_connector_1",
          "type": "create_connector",
          "user_inputs": {
            "name": "OpenAI Chat Connector",
            "description": "The connector to public OpenAI model service for GPT 3.5",
            "version": "1",
            "protocol": "http",
            "parameters": {
              "endpoint": "api.openai.com",
              "model": "gpt-3.5-turbo"
            },
            "credential": {
              "openAI_key": "12345"
            },
            "actions": [
              {
                "action_type": "predict",
                "method": "POST",
                "url": "https://${parameters.endpoint}/v1/chat/completions"
              }
            ]
          }
        },
        {
          "id": "register_model_2",
          "type": "register_remote_model",
          "previous_node_inputs": {
            "create_connector_1": "connector_id"
          },
          "user_inputs": {
            "name": "openAI-gpt-3.5-turbo",
            "function_name": "remote",
            "description": "test model"
          }
        },
        {
          "id": "deploy_model_3",
          "type": "deploy_model",
          "previous_node_inputs": {
            "register_model_2": "model_id"
          }
        }
      ],
      "edges": [
        {
          "source": "create_connector_1",
          "dest": "register_model_2"
        },
        {
          "source": "register_model_2",
          "dest": "deploy_model_3"
        }
      ]
    }
  }
}

Example response

OpenSearch responds with the workflow_id:

{
  "workflow_id" : "8xL8bowB8y25Tqfenm50"
}

Once you have created a workflow, you can use other workflow APIs with the workflow_id.