Graph Development

Build workflow graphs using SpoonOS StateGraph.

Quick Start

from spoon_ai.graph import StateGraph, END
from typing import TypedDict

class MyState(TypedDict):
    counter: int
    result: str

async def process(state: MyState) -> dict:
    return {"counter": state["counter"] + 1}

graph = StateGraph(MyState)
graph.add_node("process", process)
graph.set_entry_point("process")
graph.add_edge("process", END)

app = graph.compile()
result = await app.invoke({"counter": 0, "result": ""})

Scripts

References

Node Types

• Function nodes: async def node(state) -> dict
• Agent nodes: SpoonReactMCP instances
• Subgraph nodes: Nested StateGraph

Edge Types

• Direct: graph.add_edge("a", "b")
• Conditional: graph.add_conditional_edges("a", router, {...})

Best Practices

1. Use TypedDict for state schemas
2. Return only changed fields from nodes
3. Use checkpointing for long workflows
4. Implement timeout for external calls