[BUG] The WorkflowStepFactory is not actually a factory

[owaiskazi19]

What is the bug?

{
                    "id": "workflow_step_3",
                    "type": "register_agent",
                    "previous_node_inputs": {
                        "workflow_step_1": "tools",
                        "workflow_step_2": "tools"
                    },
                    "user_inputs": {
                        "name": "TOOLS-TOOLS-TOOLS",
                        "type": "cot",
                        "description": "this is a test agent",
                        "parameters": {"hello": "world"},
                        "llm.model_id": "ldzS04kBxRPZ5cnWrqpd",
                        "llm.parameters": {

For a case like the above, where a workflow step depends on 2 different workflow steps. It requires the outputs of

public CompletableFuture<WorkflowData> execute(
        String currentNodeId,
        WorkflowData currentNodeInputs,
        Map<String, WorkflowData> outputs,
        Map<String, String> previousNodeInputs
    )

to consist of the responses received from workflow_step_1 and workflow_step_2. Currently, the outputs just have the response of only one previous nodes.

This issue can also hamper when a WorkflowSteps requires multiple modelIds from different DeployModelSteps.

How can one reproduce the bug?

Use the below sample template to create an agent which requires multiple tools

{
    "name": "tool-register-agent",
    "description": "test case",
    "use_case": "REGISTER_AGENT",
    "version": {
        "template": "1.0.0",
        "compatibility": [
            "2.12.0",
            "3.0.0"
        ]
    },
    "workflows": {
        "provision": {
            "nodes": [
                {
                    "id": "workflow_step_1",
                    "type": "create_tool",
                    "user_inputs": {
                        "name": "MLModelTool",
                        "type": "MLModelTool",
                        "alias": "language_model_tool",
                        "description": "A general tool to answer any question. But it can't handle math problem",
                        "parameters": {
                            "model_id": "ldzS04kBxRPZ5cnWrqpd",
                            "prompt": "Answer the question as best you can.",
                            "response_filter": "choices[0].message.content"
                        }
                    }
                },
                {
                    "id": "workflow_step_2",
                    "type": "create_tool",
                    "user_inputs": {
                        "name": "MathTool",
                        "type": "MathTool",
                        "description": "A general tool to calculate any math problem. The action input must be valid math expression, like 2+3",
                        "parameters": {
                            "max_iteration": 5
                        }
                    }
                },
                {
                    "id": "workflow_step_3",
                    "type": "register_agent",
                    "previous_node_inputs": {
                        "workflow_step_1": "tools",
                        "workflow_step_2": "tools"
                    },
                    "user_inputs": {
                        "name": "TOOLS-TOOLS-TOOLS",
                        "type": "cot",
                        "description": "this is a test agent",
                        "parameters": {"hello": "world"},
                        "llm.model_id": "ldzS04kBxRPZ5cnWrqpd",
                        "llm.parameters": {
                                "max_iteration": "5",
                                "stop_when_no_tool_found": "true"
                        },
                        
                        "memory": {
                            "type": "conversation_buffer_window"
                        },
                        "app_type": "chatbot",
                        "created_time": 1689793598499,
                        "last_updated_time": 1689793598530
                    }
                }
            ],
            "edges": [
                {
                    "source": "workflow_step_1",
                    "dest": "workflow_step_3"
                },
                {
                    "source": "workflow_step_2",
                    "dest": "workflow_step_3"
                }
            ]
        }
    }
}

Response after logging the output:

[2023-12-02T23:14:34,901][INFO ][o.o.f.w.RegisterAgentStep] [ip-172-31-56-214] TOOL ADDED MLModelTool
[2023-12-02T23:14:34,901][INFO ][o.o.f.w.RegisterAgentStep] [ip-172-31-56-214] TOOLS LIST [org.opensearch.ml.common.agent.MLToolSpec@46b3a260]

What is the expected behavior?

The outputs field should have all the responses of the previous workflow steps.

What is your host/environment?

Ubuntu

Do you have any screenshots?

If applicable, add screenshots to help explain your problem.

Do you have any additional context?

Add any other context about the problem.

[dbwiddis]

The problem is that the WorkflowStepFactory isn't actually a factory. It's just storing a map of instances, so if the same workflow type is called a second time, it returns the same instance. If the future is already completed, then future.complete() becomes a no-op and the second "instance" retains the value of the completed future from the first instance.

I'll fix this.