Immutable vs Mutable State

LangGraph usually treats state as Immutable snapshots.

Why Immutability Matters

# BAD: Modifying global state in place
def bad_node(state):
    state["user"]["age"] = 26 # MUTATION!
    # If other nodes are reading 'state' right now, they might see partial updates.

# GOOD: Returning an update
def good_node(state):
    # Retrieve current value
    current_age = state.get("user", {}).get("age", 25)
    # Return structure to be merged
    return {"user": {"age": current_age + 1}} 

Concurrency Safety

If Node A and Node B run in parallel, they check out the same version of the state (Snapshot 1). They compute their updates independently. The runtime then merges them safely into Snapshot 2. Direct mutation breaks this safety guarantee.