state_schema

State Schema Example

This example demonstrates how to use State Schema and Reducers in LangGraphGo to manage complex state updates.

1. Background

In LangGraph, state is not just a simple variable that gets overwritten at each step. It is a structured object where different fields can have different update behaviors. This concept corresponds to TypedDict and Annotated in the Python library.

For example:

• A list of messages should typically be appended to.
• A counter should be incremented.
• A status flag should be overwritten.

2. Key Concepts

• StateSchema: Defines the structure of the state and how updates are merged. graph.MapSchema is the most common implementation.
• Reducer: A function that takes the current value and a new value, and returns the merged value.
  • AppendReducer: Appends new items to a list.
  • OverwriteReducer: Replaces the old value with the new one (default).
  • Custom Reducer: You can define your own logic (e.g., SumReducer).

3. How It Works

1. Define Schema: We create a MapSchema and register reducers for specific keys.
   • count: Uses SumReducer (custom) to add values.
   • logs: Uses AppendReducer to accumulate strings.
   • status: Uses default overwrite behavior.
2. Nodes: Each node returns a partial state update (a map with some keys).
3. Execution: When a node finishes, the runtime uses the Schema to merge the returned partial state into the global state.

4. Code Highlights

Defining Custom Reducer

func SumReducer(current, new interface{}) (interface{}, error) {
    // ... logic to add integers ...
    return c + n, nil
}

Configuring Schema

schema := graph.NewMapSchema()
schema.RegisterReducer("count", SumReducer)
schema.RegisterReducer("logs", graph.AppendReducer)
g.SetSchema(schema)

5. Running the Example

go run main.go

Expected Output:

--- Final State ---
Count (Sum): 6
Logs (Append): [Start Processed by A Processed by B Processed by C]
Status (Overwrite): Completed