Skip to main content

Agent Node

The Agent Node is a core component of the Sequential Agent architecture. It acts as a decision-maker and orchestrator within our workflow.

Understanding the Agent Node​

Upon receiving input from preceding nodes, which always includes the full conversation history state.messages and any custom State at that point in the execution, the Agent Node uses its defined "persona", established by the System Prompt, to determine if external tools are necessary to fulfill the user's request.

  • If tools are required, the Agent Node autonomously selects and executes the appropriate tool. This execution can be automatic or, for sensitive tasks, require human approval before proceeding (Human-in-the-Loop, HITL).
  • The Agent Node also maintains an ongoing dialogue with users, providing relevant responses based on the conversation's context and any tool-derived information.

Inputs​

RequiredDescription
System PromptYesA text prompt that defines the Agent's personality, role, and constraints. It guides the Agent's decision-making and response generation.
ToolsNoThe Tool Nodes that the Agent Node can access and execute, each offering specific functionality to retrieve information or perform actions.
Chat ModelNoA custom Chat Model (LLM) to use instead of the default one defined in the Start Node.
JSON SchemaNoA schema defining the structure of the Agent Node's response when JSON Structured Output is enabled.
Enable JSON Structured OutputNoA configuration to force the Agent Node's response to follow a predefined JSON schema.
Update StateNoA JSON object that defines how the Agent Node should update the custom State before passing it to the next node.

Outputs​

The Agent Node can connect to the following nodes as outputs:

  • Agent Node: Connects to another Agent Node to continue the conversation with a different agent or persona.
  • LLM Node: Routes the conversation flow to an LLM Node for processing and response generation.
  • Condition Agent Node: Connects to a Condition Agent Node to implement branching logic based on the agent's evaluation of the conversation.
  • Condition Node: Connects to a Condition Node to implement branching logic based on predefined conditions.
  • Loop Node: Connects to a Loop Node to implement repetitive processes based on specific conditions.
  • End Node: Connects to an End Node to conclude the conversational flow.

Features​

Tool Usage​

When connected to Tool Nodes, the Agent Node can analyze the user's request and determine if tools are necessary to fulfill it. If multiple Tool Nodes are available, the Agent Node selects the most appropriate one(s) for the task at hand.

For example, if a user asks a recipe-related question, the Agent Node might execute a "Recipe Search" Tool Node to retrieve relevant recipes, or if a user queries about weather data, the Agent Node might execute a "Weather API" Tool Node.

Human-in-the-Loop (HITL)​

The Agent Node supports Human-in-the-Loop (HITL) functionality, allowing human intervention in the agent's decision-making process, particularly for sensitive or critical operations. When HITL is enabled, the Agent Node might request human approval before executing a tool, making it a critical feature for workflows where accuracy, security, or alignment are paramount.

State Management​

In addition to responding to the user and executing tools, the Agent Node can modify the custom State to reflect progress or changes in the conversation. This enables the Agent Node to share information with subsequent nodes in the workflow.

For example, an Agent Node might update the custom State to mark a client's ID verification as complete, enabling other nodes in the workflow to proceed with sensitive operations.

Best Practices​

Clear system prompt

Craft a concise and unambiguous System Prompt that accurately reflects the agent's role and capabilities. This guides the agent's decision-making and ensures it acts within its defined scope.

Strategic tool selection

Choose and configure a set of tools that align with the agent's purpose. Avoid providing too many tools, which can lead to confusion and suboptimal selections. Instead, select specific, purpose-driven tools that integrate seamlessly with the agent's role.