Multi Step Workflow

Learn how to handle complex execution flow orchestrating the execution of multiple nodes

Multiple steps are created by defining custom events that can be emitted by nodes and trigger other nodes. Let's define a simple 3-step workflow.

Custom Events

We define two custom events, FirstEvent and SecondEvent. These classes can have any names and properties, but must implement Event:

Unix

vendor/bin/neuron make:event App\\Neuron\\FirstEvent

Windows

.\vendor\bin\neuron make:event App\Neuron\FirstEvent

namespace App\Neuron;

class FirstEvent implements Event 
{
    public function __construct(protected string $firstMsg){}
}

class SecondEvent implements Event 
{
    public function __construct(protected string $secondMsg){}
}

Defining the workflow

Now we define the workflow itself. We do this by defining the input and output types on each node. Here is the minimal implementation of the nodes for the purpose of this demo.

InitialNode

Unix

vendor/bin/neuron make:node App\\Neuron\\InitialNode

Windows

.\vendor\bin\neuron make:node App\Neuron\InitialNode

namespace App\Neuron;

use NeuronAI\Workflow\Node;
use NeuronAI\Workflow\StartEvent;
use NeuronAI\Workflow\WorkflowResources;
use App\Neuron\FirstEvent;

class InitialNode extends Node
{
    /**
     * Gets the "StartEvent" and returns "FirstEvent"
     */
    public function __invoke(
        StartEvent $event, 
        WorkflowState $state,
        WorkflowResources $resources
    ): FirstEvent {
        echo "\n- Handling StartEvent";
        
        return new FirstEvent("InitialNode complete");
    }
}

NodeOne

Unix

vendor/bin/neuron make:node App\\Neuron\\NodeOne

Windows

.\vendor\bin\neuron make:node App\Neuron\NodeOne

class NodeOne extends Node
{
    /**
     * Takes "FirstEvent" as input and returns "SecondEvent"
     */
    public function __invoke(
        FirstEvent $event, 
        WorkflowState $state,
        WorkflowResources $resources
    ): SecondEvent {
        echo "\n- ".$event->firstMsg;
        
        return new SecondEvent("NodeOne complete");
    }
}

NodeTwo

Unix

vendor/bin/neuron make:node App\\Neuron\\NodeTwo

Windows

.\vendor\bin\neuron make:node App\Neuron\NodeTwo

class NodeTwo extends Node
{
    /**
     * Takes "SecondEvent" as input and returns "StopEvent"
     */
    public function __invoke(
        SecondEvent $event, 
        WorkflowState $state,
        WorkflowResources $resources
    ): StopEvent {
        echo "\n- ".$event->secondMsg;
        
        echo "\n- NodeTwo complete";
        
        return new StopEvent();
    }
}

Define the Workflow attaching the nodes:

use NeuronAI\Workflow\Workflow;

$workflow = Workflow::make()
    ->setWorkflowId('wk_id')
    ->addNodes([
        new InitialNode(),
        new NodeOne(),
        new NodeTwo(),
    ]);

/*
 * Run the workflow
 */
$workflow->run();

The full output will be:

- Handling StartEvent
- InitialNode complete
- NodeOne complete
- NodeTwo complete

Of course there is still not much point to a workflow if you just run through it from beginning to end! Let's do some branching and looping.

Monitoring & Debugging

Many of the applications you build with Neuron will contain multiple steps with multiple invocations of LLM calls. As these applications get more and more complex, it becomes crucial to be able to inspect what exactly is going on inside your agentic system. The best way to do this is with Inspector.

Neuron AI | Inspectordocs.inspector.dev