Agent Prompt Node

Overview

The agent prompt node serves as the central logic of the agent in many cases. It allows users to configure the model and prompts that the agent will use to process inputs effectively.

Model List

In this node section, you select the model that best suits the task you want the agent to accomplish. Different models may have varying capabilities and performance characteristics.

Prompt Place

Here, you write the prompt for the selected model, specifying what the language model (LLM) needs to do when it receives input from the user or a tool.

Tuning Settings

The LLM tuning settings include options for adjusting the maximum token limit, randomness of the AI responses, and whether you prefer more focused or diverse outputs (temperature).

Max Token Setting

You can choose a number from 1 to 8000 (it may depend on the model!) tokens to set the maximum length of the output generated by the model.

Randomness Setting

This setting determines whether the LLM's responses are more random or more predictable. Randomness refers to the variability in the model's outputs; a higher randomness setting may lead to more unexpected responses, while a lower setting results in more consistent and predictable outcomes. You can choose a value from 1 to 10, where 1 is low randomness and 10 is high randomness.

Focused or Diverse Setting

This setting is commonly known as "temperature" in LLM terminology. Temperature controls the creativity of the model's responses: a lower temperature (closer to 1) results in more focused and deterministic outputs, while a higher temperature (closer to 10) encourages more diverse and creative responses. You can choose a value from 1 to 10, where 1 is low temperature and 10 is high temperature.

Tool Integration

This section allows for the integration of various tools, specifically those that have a compatible blue connection dot. When a tool is added, the model will utilize that tool to enhance its functionality.

While it is not mandatory to add tool, it is highly recommended to use the appropriate tool, as this will enable the model to function more effectively.