How to use Agent

1. Introduction

In this tutorial, we will explore the Agent section of Yala in detail, a powerful tool that integrates artificial intelligence into Google Sheets to automate and optimize your workflows. The Agent section allows you to personalize the system interaction by selecting different Agent Name (profiles or personalities) and Model Name (language models) to process your prompts. Understanding how to use this feature is essential to fully leverage Yala’s capabilities.

In your sheet (for example, the Nutritionist tab), you might have columns like Task, Category, Product, Composition, Status, and also Agent Name, Model Name, User Prompt, and Assistant Output.

2. How the Agent Works

The Agent section is responsible for determining how Yala interprets and responds to your User Prompt. By choosing an Agent Name and a Model Name, you define both the “personality” and the “capacity” of the AI assistant that will handle the processing.

Operational Flow

Selecting the Agent Name
You choose the agent best suited for the task (for instance, "Nutrition Specialist" or "Data Analyst").

Selecting the Model Name
You pick the language model that aligns with the chosen agent (e.g., "GPT-4o").

Providing the User Prompt
You enter the question or instruction you want the agent to handle (for example: “Based on the Memory Output, classify these Products…”).

Processing
Yala combines the chosen Agent Name and Model Name to generate the Assistant Output.

Interactions with Other Components

Memory
The Assistant Output can be influenced by Memory Output, if present, adding more context to the response.

Parsing and Action
The Assistant Output can be used in subsequent workflow steps for analysis or to trigger specific actions.

3. Required Parameters and Inputs

Input Data

Agent Name
The name of the agent that will process the User Prompt.

Model Name
The language model to be used for the interaction.

User Prompt
The question or instruction you want the agent to process.

Prerequisites

Defined Agents and Models
Make sure the agents and models you intend to use (e.g., "Nutrition Specialist" or "GPT-4o") are correctly configured in [Data] Agents or [Data] Models.

Proper Configuration
If you are using the Memory section, ensure that the Memory Output has already been generated before calling the Agent.

4. Detailed Look at the Agent Columns

Agent Name

Description
Defines which persona will process the User Prompt.

How to Use

Select an agent (e.g., "Nutrition Specialist") from the dropdown menu.

The chosen agent determines the assistant’s style, domain knowledge, and behavior.

Model Name

Description
Specifies which language model will interpret and respond to the User Prompt.

How to Use

Choose a model (e.g., "GPT-4o") from the dropdown.

The model determines the assistant’s capacity for understanding and generating text.

User Prompt

Description
This is where you enter the question or instruction you want the agent to fulfill.

How to Use

Clearly describe what you want to obtain (e.g., “Create a basic nutritional information table for the Category = Cake, considering Composition data.”).

Reference any relevant columns or Memory Output to add context (for instance, “Use the memory data with the Product details.”).

Assistant Output

Description
Shows the response generated by the agent based on the settings and the User Prompt.

How to Use

This field is automatically filled once processing is complete.

Check the content to see if the response meets your requirements.

If you need to refine the response, update your User Prompt or consider adjusting the agent/model.

5. Practical Examples

Example 1: Generating a Customized Summary

Scenario
You want to get a brief summary about certain Products using a specialized agent.

Steps

Set Agent Name to "Nutrition Specialist".

Choose Model Name as "GPT-4o".

In User Prompt, enter: “Based on Composition data for each Product, summarize the key nutritional facts in one paragraph.”

Set Status to Pending.

Result
The Assistant Output will provide a concise summary about the nutritional facts for your Products.

Example 2: Using the Memory Output in the User Prompt

Scenario
You have already generated Memory Output that consolidates Category and Composition data, and you want the agent to analyze it.

Steps

Confirm that Memory Output is ready (e.g., “It now contains combined information about gluten-free products.”).

Set Agent Name to "Data Analyst".

Choose Model Name as "GPT-4o".

In User Prompt, enter: “Using the data from Memory Output, list any potential dietary restrictions or allergens found.”

Set Status to Pending.

Result
The Assistant Output will analyze the data from the memory and highlight relevant dietary restrictions or allergens.

6. Best Practices

Choosing the Right Agent Name

Select an agent aligned with the task. For example, "Nutrition Specialist" to handle diet-related topics, or "Copywriter" for writing marketing copy.

Model Selection

Opt for advanced models like "GPT-4o" if the task demands detailed analysis or broader language understanding.

Clarity in the User Prompt

Be specific when writing your request. The more context you provide, the more accurately the AI can respond.

Referencing Variables and Memory Output

Use columns like Category, Product, Composition, and any generated Memory Output to give the agent a strong context.

7. Troubleshooting

Empty or Unexpected Assistant Output

Check that Agent Name and Model Name exist and are spelled correctly.

Review the User Prompt for clarity and intent.

Processing Error

If Status changes to Failure, look at the Assistant Output or logs for potential error messages.

Ensure all required columns (Agent Name, Model Name, User Prompt) are filled.

Agent or Model Not Found

Verify that the agent or model name matches what was set up in [Data] Agents or [Data] Models.

Update or create new entries if necessary.

8. Relationship with the Status

Pending
After configuring Agent Name, Model Name, and User Prompt, set Status to Pending. This tells Yala that the row should be processed.

Processing
Indicates that the Agent is working on the User Prompt with the chosen model.

Success
The Assistant Output was created successfully and is ready for further steps (like Parsing, Data, or Action).

Failure
An error occurred during processing. Check the response or logs to fix the inputs.

9. References and Additional Resources

Complementary Guides

See the Memory tutorial for adding context to the User Prompt.

Review Parsing to extract information from the Assistant Output.

Personalizing Agents

You can add new agents in the Yala menu: Builders > Agent Builder.

Support

If you have questions or difficulties, contact the Yala support team.

By mastering the Agent section, you can tailor how Yala interacts with your prompts, choosing agents and models that best meet your needs. Experiment with different combinations (e.g., "Nutrition Specialist" + "GPT-4o") to discover how they can help you achieve your specific project goals. Make the most of this functionality to optimize your workflows and achieve even better results!