How to use Workflow

1. Introduction

Welcome to Yala – an extension that connects AI to Google Sheets for automating tasks and workflows. In this tutorial, you’ll learn how to set up and use the columns and functionalities in the tab where Yala was installed, so that each row represents a task to be executed when you click Run Workflow.

Overview

Fill out the columns for each task.

Set the Status to Pending to process.

Click Run Workflow in the Yala menu.

Yala will read each row, interpreting Memory, Agent, Parsing, Data, and Action until the task is completed (or fails).

2. Main Sheet Overview

In the main tab where Yala is installed:

Each row = a task in your workflow.

Each column = a Yala feature that controls how AI interacts with and transforms your data (e.g., Memory, Agent, Parsing, Data, Action).

Status = the control column indicating whether a task will be processed or not.

At the end of processing:

If everything goes well, Status becomes Success.

In case of an error, it becomes Failure, which stops the global workflow and shows where it failed (for example, in Memory Output, Assistant Output, or Parsing Output).

Why is each row a “task”?

Imagine you want to generate a nutritional table for a product. You fill in variables like Task, Category, Product, and Composition, describe in Memory Prompt what should be “remembered,” ask the Agent to format the information, refine the output with Parsing, record data in another tab via Data, and if needed, perform extra steps with Action. All of that is done in the same row, constituting one complete task.

3. Yala Features (Columns)

Yala organizes columns into specific functionalities. Below is an explanation of each:

3.1. Variables

Columns you create to store free-form information needed by the workflow. For example, Task, Category, Product, Composition.

How to use

Add columns with names that make sense for your automation.

Fill these fields with relevant data.

You can reference them in Memory, Agent, Parsing, etc.

3.2. Status

A column to define whether a row will be processed and to track its state.

Options

Pending: task ready to be executed the next time you click Run Workflow.

Processing: indicates this row is currently in progress.

Success: task completed without errors.

Failure: an error occurred, stopping the workflow.

Ignore: this row will not be processed.

Continue: resumes the workflow starting from this row.

To actually process the row, simply set Status = Pending and click Run Workflow in the menu.

3.3. Memory

Used to provide context or memory to Yala, helping the AI produce coherent responses.

Columns

Memory Type: defines the scope:

Spreadsheet: considers the entire spreadsheet.

Sheet: only the current tab.

Row: just the current row.

Custom: free text.

Memory Prompt: instruction telling Yala what to memorize or which context to consider.

Memory Output: automatically generated by Yala, combining what it found in the chosen scope (Memory Type) with what you described in Memory Prompt.

3.4. Agent

Defines who (which agent/persona) and which AI model will respond to your main request.

Columns

Agent Name: choose one of the available agents.

Model Name: pick a language model best suited to that agent.

User Prompt: the final request (e.g., “Create the nutritional table for Product, considering Composition.”).

Assistant Output: the response generated after Yala combines Memory Output with your User Prompt.

3.5. Parsing

A way to extract or filter specific parts of the text produced in Assistant Output.

Columns

Parsing Prompt: specify what you want to analyze or extract (e.g., “List only the vitamins and minerals”).

Parsing Output: generated by Yala, containing only the requested portion.

3.6. Data

Enables data manipulation in other Google Sheets tabs, such as inserting or updating rows.

Columns

Data Prompt: describe the data actions to perform (e.g., “Insert the columns Category, Product, Serving Size, and Nutrients into the Nutritional Table tab.”).

3.7. Action

Defines “extra” actions outside of data manipulation – for example, skipping to another row or finding a similar item.

Columns

Action Prompt: describe complementary actions (e.g., “Update line 3 to Ignore and go to line 4.”).

4. Practical Example: Generating a Nutritional Table

Suppose your sheet has four Variables columns: Task, Category, Product, and Composition. Below is how you’d fill out each functionality:

Memory

Memory Type: Row.

Memory Prompt: “Obtain the Category, Product, and its Composition information.”

Agent

Agent Name: “Nutrition Specialist.”

Model Name: “GPT-4o.”

User Prompt: “Create a basic nutritional information table for the Product, taking into account its Composition. Include columns for: Category, Product, Serving Size (g), Proteins (g), Carbohydrates (g), Fats (g), Vitamins, and Minerals.”

Parsing

Parsing Prompt: “Transform into JSON with the structure: ["Category":"X","Product":"Y","Serving Size (g)":100,…].”

Data

Data Prompt: “Insert the data into the Nutritional Table tab using the Assistant Output and Parsing Output columns.”

Action

Action Prompt: “Update line 3 to Ignore and go to line 4.”

Finally, set Status to Pending and click Run Workflow.

What happens then?

Yala reads the Memory Prompt and generates a Memory Output.

It combines Memory Output with the User Prompt, creating an Assistant Output.

It applies the Parsing Prompt to Assistant Output, producing a Parsing Output.

It runs the Data Prompt, inserting the appropriate data into the Nutritional Table tab.

It handles the Action Prompt, updating one row to Ignore and moving to the next line.

If no error occurs, the row’s Status is Success.

5. Final Tips

Create as many Variables as needed to contextualize your workflow (e.g., Task, Category, Product, Composition).

Always set Status = Pending for tasks you want processed during execution.

Be concise in your prompts. The clearer the instruction, the better the AI’s response.

Check the output columns (Memory Output, Assistant Output, Parsing Output) if something fails.

Combine Data and Action for more complex workflows (managing data and jumping to specific rows, for example).

6. Conclusion

In this tutorial, you’ve seen how each column in Yala’s main sheet plays a role in your workflow:

Variables – store free-form data for later use.

Status – manage the row’s state (Pending, Success, Failure, etc.).

Memory – create a context or memory for the AI.

Agent – ask the AI agent for something and receive a response.

Parsing – extract specific parts of the generated text.

Data – insert/update records in another tab.

Action – execute additional steps (move to another row, etc.).

By following this model, you can automate your tasks easily and at scale in Google Sheets. Just:

Configure the prompts in each functionality,

Adjust the variables,

And click Run Workflow.

Yala takes care of the rest, ensuring each row (each task) is processed step by step according to your instructions. Happy automating!