Persistence
Persist the Workflow State across executions.
When we talk about persistence in Neuron, we're talking about the system's ability to capture and preserve the complete state of a running workflow at any moment. This includes:
All variables and their current values
The exact execution position – which node is active, which have completed, which are waiting
Context and metadata – timestamps, user information, decision history
Error states and retry counters – so failures can be handled gracefully
Think of it like a sophisticated "save game" feature, but for business processes. At any point, when an interruption is asked from a node, Neuron create a snapshot of your workflow's state and store it in the persistence layer. Later – whether that's seconds, hours, or weeks – the workflow can be restored to exactly that moment and continue as if nothing happened.
As usual in Neuron the Workflow persistence layer is built on top of a common interface so it's extensible and interchangeable. Below the supported persistence layer.
When to use Persistence
Persistence comes into play when you intend to use interruption. The persistence component requires to decalre also a workflow ID.
InMemoryPersistence
It keep data in memory only for the current execution cycle.
use NeuronAI\Workflow\Persistence\InMemoryPersistence;
$workflow = new WorkflowAgent(
new InMemoryPersistence(),
'CUSTOM_ID'
);FilePersistence
It will store the Workflow data and state into a local file.
use NeuronAI\Workflow\Persistence\FilePersistence;
$workflow = new WorkflowAgent(
new FilePersistence(__DIR__),
'CUSTOM_ID'
);DatabasePersistence
To persist the workflow interruption in the database you need to pass a PDO instance. If you are working on top of a framework you can get it from the ORM in the same way of the SQLChatHistory.
use NeuronAI\Workflow\Persistence\DatabasePersistence;
$workflow = new WorkflowAgent(
new DatabasePersistence(
pdo: new \PDO(...),
table: 'workflow_interrupts'
),
'CUSTOM_ID'
);Here are the SQL scripts to create the table:
CREATE TABLE IF NOT EXISTS workflow_interrupts (
workflow_id VARCHAR(255) PRIMARY KEY,
data LONGBLOB NOT NULL,
created_at DATETIME NOT NULL,
updated_at DATETIME NOT NULL,
INDEX idx_workflow_id (workflow_id),
INDEX idx_updated_at (updated_at)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;CREATE TABLE workflow_interrupts (
workflow_id VARCHAR(255) PRIMARY KEY,
data BYTEA NOT NULL,
created_at TIMESTAMP NOT NULL,
updated_at TIMESTAMP NOT NULL
);
CREATE INDEX idx_workflow_id ON workflow_interrupts(workflow_id);
CREATE INDEX idx_updated_at ON workflow_interrupts(updated_at);CREATE TABLE workflow_interrupts (
workflow_id TEXT PRIMARY KEY,
data BLOB NOT NULL,
created_at TEXT NOT NULL,
updated_at TEXT NOT NULL
);
CREATE INDEX idx_workflow_id ON workflow_interrupts(workflow_id);
CREATE INDEX idx_updated_at ON workflow_interrupts(updated_at);Last updated