Skip to content
Nappai Documentation

CrewAI Agent

This component, CrewAI Agent, is a powerful tool in Nappai that allows you to create a custom AI assistant to automate tasks and processes. You define its role, goals, and capabilities, and it gets to work!

Relationship with CrewAI

This component uses the CrewAI library to create and manage intelligent agents. CrewAI allows these agents to interact with language models, use various tools, and complete complex tasks based on your instructions. For more information on CrewAI, visit https://github.com/crewai/crewai.

Inputs

  • Role: Describe the agent’s job. For example: “Data Analyst,” “Customer Service Representative,” or “Report Generator.” Use multiple lines if needed.
  • Goal: Specify the task you want the agent to accomplish. For example: “Analyze sales data and identify top-performing products,” or “Respond to customer inquiries about order status.” Use multiple lines if needed.
  • Backstory: Provide background information to help the agent understand the context of its task. This is optional but can be helpful for more complex tasks. Use multiple lines if needed.
  • Tools: Select the tools the agent can use to complete its task. This will be a selection from a list of available tools within Nappai.
  • Language Model: Choose the AI model the agent will use to process information and make decisions. This will be a selection from a list of available models within Nappai.
  • Memory: (On/Off) Enable this to allow the agent to remember previous interactions and use that information to improve its performance. This is ON by default.
  • Verbose: (On/Off) Enable this to get detailed information about the agent’s actions and reasoning. This is OFF by default.
  • Allow Delegation: (On/Off) Enable this to allow the agent to delegate sub-tasks to other agents if needed. This is ON by default.
  • Allow Code Execution: (On/Off) Enable this to allow the agent to execute code. This is OFF by default. Use caution with this setting.
  • kwargs: (Advanced) Additional parameters for fine-tuning the agent’s behavior. Leave this blank unless you have specific advanced requirements.

Outputs

The component produces a configured AI agent ready to perform the task you’ve defined. This agent will be used by other Nappai components to execute your automation workflow.

Usage Example

Let’s say you want to create an agent to summarize customer feedback from emails.

  1. Role: Customer Feedback Summarizer
  2. Goal: Summarize customer feedback from emails into concise bullet points.
  3. Backstory: Customer emails are stored in a database accessible via the “Email Database” tool.
  4. Tools: Select the “Email Database” tool.
  5. Language Model: Choose your preferred language model.
  6. Memory: Leave this ON.
  7. Verbose: Leave this OFF.
  8. Allow Delegation: Leave this OFF.
  9. Allow Code Execution: Leave this OFF.
  10. kwargs: Leave this blank.

After configuring these inputs, the component will generate an agent ready to process customer emails and provide summaries.

Templates

This component is used in the following Nappai templates:

  • Complex Agent
  • Hierarchical Tasks Agent
  • Runnable With Message History: Executes a runnable task and keeps a record of its interactions.
  • Runnable Executor: Executes a runnable task.
  • Hierarchical Crew: Manages a group of agents working together on a task. More details: https://github.com/crewai/crewai
  • Sequential Task: Executes tasks sequentially, assigning each to a specific agent. More details: https://github.com/crewai/crewai

Tips and Best Practices

  • Clearly define the agent’s role and goal for optimal performance.
  • Select appropriate tools and language models based on the task’s complexity.
  • Start with simpler configurations and gradually increase complexity as needed.
  • Carefully consider the implications of enabling “Allow Code Execution.”

Security Considerations

When using the “Allow Code Execution” option, ensure that the code executed by the agent is secure and does not pose a risk to your system. Only enable this option if absolutely necessary and after careful review of the code. Nappai is not responsible for any security breaches resulting from the execution of user-provided code.