[FEATURE] Add an API to return validation parameters

[dbwiddis]

Is your feature request related to a problem?

Users do not have an easy way to validate workflow steps, particularly in understanding what parameters are part of the output (e.g., model_id) and which inputs are required vs. optional (other than by laboriously visiting original API docs). We should allow users to easily identify these requirements via API.

In addition, the planned front-end will need access to this same information in order to enable real-time validation in the drag-and-drop interface.

What solution would you like?

Create a new REST API to return content equivalent to the workflow-steps.json file.

Allow an optional parameter for the workflow step type which returns only the content of that portion of the file.

What alternatives have you considered?

1. Force users to navigate documentation.
2. Point users to the JSON file on GitHub.
3. Document every single step on opensearch.org.

Do you have any additional context?

We already parse the contents of this file into a WorkflowValidator class, which internally contains a Map<String, WorkflowStepValidator>.

1. The API can simply return the contents of these objects, which simply requires toXContent() methods.
2. We could also avoid the potential issues with parsing this file (e.g., users experiencing runtime validation checks rather than compile time checks) by simply hardcoding the WorkflowStepValidator map. We can then remove the parse methods and all the complexity with exceptions that they bring.

[owaiskazi19]

Request: GET /_plugins/_flow_framework/workflow/_steps

Response:

{
  "undeploy_model": {
    "inputs": [
      "model_id"
    ],
    "outputs": [
      "success"
    ],
    "required_plugins": [
      "opensearch-ml"
    ]
  },
  "register_model_group": {
    "inputs": [
      "name"
    ],
    "outputs": [
      "model_group_id",
      "model_group_status"
    ],
    "required_plugins": [
      "opensearch-ml"
    ]
  },
  "register_remote_model": {
    "inputs": [
      "name",
      "connector_id"
    ],
    "outputs": [
      "model_id",
      "register_model_status"
    ],
    "required_plugins": [
      "opensearch-ml"
    ]
  },
  "register_local_sparse_encoding_model": {
    "inputs": [
      "name",
      "version",
      "model_format",
      "function_name",
      "model_content_hash_value",
      "url"
    ],
    "outputs": [
      "model_id",
      "register_model_status"
    ],
    "required_plugins": [
      "opensearch-ml"
    ],
    "timeout": "1m"
  },
  "create_index": {
    "inputs": [
      "index_name",
      "default_mapping_option"
    ],
    "outputs": [
      "index_name"
    ],
    "required_plugins": []
  },
  "delete_agent": {
    "inputs": [
      "agent_id"
    ],
    "outputs": [
      "agent_id"
    ],
    "required_plugins": [
      "opensearch-ml"
    ]
  },
  "create_ingest_pipeline": {
    "inputs": [
      "id",
      "description",
      "type",
      "model_id",
      "input_field_name",
      "output_field_name"
    ],
    "outputs": [
      "pipeline_id"
    ],
    "required_plugins": []
  },
  "register_local_pretrained_model": {
    "inputs": [
      "name",
      "version",
      "model_format"
    ],
    "outputs": [
      "model_id",
      "register_model_status"
    ],
    "required_plugins": [
      "opensearch-ml"
    ],
    "timeout": "1m"
  },
  "create_tool": {
    "inputs": [
      "type"
    ],
    "outputs": [
      "tools"
    ],
    "required_plugins": [
      "opensearch-ml"
    ]
  },
  "noop": {
    "inputs": [],
    "outputs": [],
    "required_plugins": []
  },
  "create_connector": {
    "inputs": [
      "name",
      "description",
      "version",
      "protocol",
      "parameters",
      "credential",
      "actions"
    ],
    "outputs": [
      "connector_id"
    ],
    "required_plugins": [
      "opensearch-ml"
    ]
  },
  "delete_model_group": {
    "inputs": [],
    "outputs": [],
    "required_plugins": []
  },
  "register_agent": {
    "inputs": [
      "name",
      "type"
    ],
    "outputs": [
      "agent_id"
    ],
    "required_plugins": [
      "opensearch-ml"
    ]
  },
  "deploy_model": {
    "inputs": [
      "model_id"
    ],
    "outputs": [
      "deploy_model_status"
    ],
    "required_plugins": [
      "opensearch-ml"
    ],
    "timeout": "15s"
  },
  "delete_connector": {
    "inputs": [
      "connector_id"
    ],
    "outputs": [
      "connector_id"
    ],
    "required_plugins": [
      "opensearch-ml"
    ]
  },
  "register_local_custom_model": {
    "inputs": [
      "name",
      "version",
      "model_format",
      "function_name",
      "model_content_hash_value",
      "url",
      "model_type",
      "embedding_dimension",
      "framework_type"
    ],
    "outputs": [
      "model_id",
      "register_model_status"
    ],
    "required_plugins": [
      "opensearch-ml"
    ],
    "timeout": "1m"
  },
  "delete_model": {
    "inputs": [
      "model_id"
    ],
    "outputs": [
      "model_id"
    ],
    "required_plugins": [
      "opensearch-ml"
    ]
  }
}

[dbwiddis]

Should we remove the create index and create ingest pipeline from the JSON?

[owaiskazi19]

Should we remove the create index and create ingest pipeline from the JSON?

We should. Let me push the change.