create_agent

Create Agent Example

This example demonstrates how to use the prebuilt.CreateAgent function in LangGraphGo to easily create a ReAct agent with functional options.

1. Background

Creating an agent often involves setting up a graph with an agent node, a tool execution node, and conditional routing logic. While CreateReactAgent provides a basic implementation, CreateAgent offers a more flexible and extensible way to construct agents, inspired by LangChain 1.0's design. It supports functional options for easier configuration of system messages, state modifiers, and future extensions like checkpointing.

2. Key Concepts

• CreateAgent: A factory function that builds a StateGraph for an agent.
• Functional Options: A pattern to pass optional parameters (e.g., WithSystemMessage) to the factory function, making the API clean and extensible.
• System Message: A predefined instruction that sets the behavior or persona of the agent.
• State Modifier: A function that intercepts and modifies the message history before it is sent to the LLM. This is useful for dynamic prompt engineering or filtering.

3. How It Works

1. Initialize LLM: Create an instance of a language model (e.g., OpenAI).
2. Define Tools: Create a list of tools the agent can use.
3. Create Agent: Call prebuilt.CreateAgent with the model, tools, and options.
   • WithSystemMessage: Sets the system prompt.
   • WithStateModifier: Allows custom logic to modify messages.
4. Invoke: Run the agent with an initial user message. The agent will loop through reasoning and tool execution until a final answer is reached.

4. Code Highlights

Creating the Agent with Options

agent, err := prebuilt.CreateAgent(model, inputTools,
    // Set a system message
    prebuilt.WithSystemMessage("You are a helpful weather assistant. Always be polite."),
    
    // Add a state modifier to log or alter messages
    prebuilt.WithStateModifier(func(msgs []llms.MessageContent) []llms.MessageContent {
        log.Printf("Current message count: %d", len(msgs))
        return msgs
    }),
)

Running the Agent

inputs := map[string]interface{}{
    "messages": []llms.MessageContent{
        llms.TextParts(llms.ChatMessageTypeHuman, "What is the weather in San Francisco?"),
    },
}

result, err := agent.Invoke(ctx, inputs)

5. Running the Example

export OPENAI_API_KEY=your_api_key
go run main.go

Expected Output:

2024/12/01 22:30:00 Starting agent...
2024/12/01 22:30:00 Current message count: 2
Agent Response: The weather in San Francisco is sunny and 25°C.