​

Node Inputs

​

Required Fields

Instructions:
Custom instructions provided to the LLM for task execution. Use dynamic variables ({{myVariable}}) to personalize input.
Example: "Write a story about F1 driver loosing the championship for this driver {{driver}} with the team {{team}}"

​

Optional Fields

Tools:
Defines the tools or functions the agent can call during its execution.
Example: "Scrapper"

Datasources:
Specifies the data sources the agent can access or query.
Example:
  Datasource: "Internal company knowledge"
  Instructions block:

{Your instruction}  

Here is the data source you will be working with: "Internal company knowledge"

Advanced Settings:
Allows configuration of advanced LLM parameters such as temperature, token limits, and output formats. (see “Node AI Settings” for more information)

​

Node AI Settings

Model:
Select the LLM model to be used.
Example: GOOGLE_VERTEXAI - gemini-1.5-pro-002

Temperature:
Controls creativity and randomness in responses.

• Lower values (e.g., 0.2) make outputs more deterministic.
• Higher values (e.g., 0.8) increase creativity.

Max Output Tokens:
Sets the maximum number of tokens in the response.
Example: 1000

Top K:
Restricts the response to the top K most likely tokens.
Example: 40

Top P:
Adjusts the probability distribution for token selection.
Example: 1 (uses the full distribution).

Output JSON Schema:
Define the format of the response in JSON schema if structured output is required.
Example:

{
  "summary": "string",
  "keywords": ["string"]
}

​

Node Output

Output:
The response generated by the LLM based on the provided instructions, tools, and datasources.
Example Output:

{
  "summary": "Coffee beans are essential for brewing high-quality espresso.",
  "keywords": ["coffee beans", "espresso", "grinding"]
}

​

Node Functionality

The Agent node:

• Leverages LLMs for dynamic task processing and natural language interactions.
• Integrates tools and datasources to expand functionality and enrich outputs.
• Offers customization through advanced LLM settings for tailored responses.
• Supports structured outputs via JSON schema for compatibility with other systems.