Store

The Store is Exosphere's graph-level key-value storage system that persists data across the entire workflow execution. It provides a way to share state between nodes and maintain persistent data throughout the workflow lifecycle.

Overview

The Store provides persistent storage that survives across node executions, enabling complex workflows that need to maintain state, track progress, or share data between different parts of the workflow.

graph TB
    A[Graph Trigger] --> B[Store Initialization]
    B --> C[Node 1: Read/Write Store]
    C --> D[Node 2: Read/Write Store]
    D --> E[Node 3: Read/Write Store]
    E --> F[Workflow Complete]

    G[Store Data] -.->|Persists| C
    G -.->|Persists| D
    G -.->|Persists| E

    style G fill:#e8f5e8

How Store Works

Store Lifecycle

1. Initialization: Store is created when the graph is triggered
2. Persistence: Data persists across all node executions in the workflow
3. Cleanup: Store data is automatically cleaned up when the workflow completes

Store Access

• All nodes can read from and write to the store
• Key-value pairs are stored as strings
• Automatic persistence ensures data survives node restarts and failures

Implementation

Store Configuration

Define store requirements in your graph template:

{
  "store_config": {
    "required_keys": ["cursor", "batch_id"],
    "default_values": {
      "cursor": "0",
      "batch_size": "100"
    }
  },
  "nodes": [
    {
      "node_name": "DataProcessorNode",
      "identifier": "processor",
      "inputs": {
        "cursor": "${{ store.cursor }}",
        "batch_size": "${{ store.batch_size }}"
      },
      "next_nodes": []
    }
  ]
}

Python SDK Example

from exospherehost import StoreConfigModel

store_config = StoreConfigModel(
    required_keys=["cursor", "batch_id"],
    default_values={
        "cursor": "0",
        "batch_size": "100"
    }
)

result = await state_manager.upsert_graph(
    graph_name="my-workflow",
    graph_nodes=graph_nodes,
    store_config=store_config
)

Triggering with Store Data

# Trigger graph with initial store values
result = await state_manager.trigger(
    "my-workflow",
    inputs={"user_id": "123"},
    store={
        "cursor": "0",
        "batch_id": "batch_001"
    }
)

Store Operations

Reading from Store

Use ${{ store.key }} syntax to access store values:

{
  "inputs": {
    "current_cursor": "${{ store.cursor }}",
    "batch_size": "${{ store.batch_size }}",
    "user_id": "${{ store.user_id }}"
  }
}

Writing to Store

Store values are updated when nodes complete successfully. The store is automatically updated with any new values that nodes produce.

Store Validation

• Required keys: Must be present when the graph is triggered
• Default values: Automatically provided if not specified
• Key constraints: Keys cannot contain dots (.) or be empty

Next Steps

• Fanout - Learn how to create parallel execution paths
• Unite - Synchronize parallel execution paths
• Signals - Control workflow execution flow
• Retry Policy - Build resilient workflows