# 워크플로우 구축

{% hint style="info" %}
이 챕터의 핵심 개념

* **모듈 및 그 설정**
  {% endhint %}

### A Simple Agent Example

In Pro Config, a workflow is a cascade of multiple Modules that can perform a series of tasks. In this chapter, we will build a simple agent that involves the cascade of two Modules. 

```json
{
  "type": "automata",
  "id": "chat_demo",
  "initial": "chat_page_state",
  "inputs": {},
  "outputs": {},
  "transitions": {},
  "states": {
    "chat_page_state": {
      "inputs": {
        "user_message": {
          "type": "IM",
          "user_input": true
        }
      },
      "tasks": [
        {
          "name": "generate_reply",
          "module_type": "AnyWidgetModule",
          "module_config": {
            "widget_id": "1744214024104448000",
            "system_prompt": "You are a teacher teaching Pro Config.",
            "user_prompt": "{{user_message}}",
            "output_name": "reply"
          }
        },
        {
          "name": "generate_voice",
          "module_type": "AnyWidgetModule",
          "module_config": {
            "content": "{{reply}}",
            "widget_id": "1743159010695057408",
            "output_name": "reply_voice"
          }
        }
      ],
      "render": {
        "text": "{{reply}}",
        "audio": "{{reply_voice}}"
      },
      "transitions": {
        "CHAT": "chat_page_state"
      }
    }
  }
}
```

In the above config, we have created an agent that generates a reply to the user's input. Please note that this is a very simple agent without any memory (we will implement an agent with memory in [Expressions and Variables](/ko/create/pro-config-mode/tutorial/expressions-and-variables.md)). Now let's delve into the details of the above example:

In the agent above, we define the variable `user_message` of type `IM` (Instant Messaging). This allows the agent to take input in the form of messages sent to the agent. In the above example, we have set a transition on event `CHAT` to point to `chat_page_state`. So, whenever a user sends a message in the chat, the state is reloaded and the message is taken as in `IM` input.

This `user_message` is then passed into the tasks.

### Configuration of Each Module

Tasks contain multiple modules that execute sequentially. For each module, we need to specify the `module_type` and `module_config`.  Rather, `name` is optional (for readability). In the example above, we include a GPT-3.5 LLM widget and a TTS widget provided on MyShell. 

For demonstration purposes, we only used a simple system prompt `"You are a teacher teaching Pro Config."`  In real-world applications,  it is required to put some effort into prompt engineering and optimizing this system prompt for better performance. \
\
As for the TTS Widgets, you can choose any of your favorite voices from <https://app.myshell.ai/robot-workshop> and paste the widget ID into the config.&#x20;

Specifically, after clicking the "workshop" button on the left menu bar, you will see this widget center. On the very top selection bar, select "TTS".  Then pick your favorite voice model, use the "more" button to copy the widget ID, and paste into the config file.

&#x20;![](/files/60ubKlfpW4GPaMtY7oZN)![](/files/lLdk5KFOsKfJChZLbqwl)

{% hint style="info" %}
If you would like to add more modules or widgets, please refer to\
[모든 위젯과의 통합](/ko/create/pro-config-mode/tutorial/integration-with-any-widget.md) and [모듈](/ko/create/pro-config-mode/basic/modules.md)
{% endhint %}

### Render and Transitions

After the execution of the tasks, two variables called `reply` and `reply_voice` are obtained and ready to be rendered in a message that would appear in the User Interface. We have also handled a special event called `CHAT` in the transitions, which means when a user sends a message, the automata will jump into  `chat_page_state` and execute that state. In the next section, we will describe how to handle and use different types of transitions.


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.myshell.ai/ko/create/pro-config-mode/tutorial/building-workflow.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.