> ## Documentation Index
> Fetch the complete documentation index at: https://messenger-docs.cogfy.com/llms.txt
> Use this file to discover all available pages before exploring further.

# OpenAI Assistant

> Chat completion models by OpenAI

Use chat completion powered by OpenAI to its full potential, including token control and tool usage.

## Input

<ResponseField name="Input" type="string" required>
  The user's prompt — the actual input or question from the **user**.

  In most scenarios, you will want to interpolate this field with messages coming from the user (e.g., when using a Message Received trigger).
</ResponseField>

<ResponseField name="Model" type="string" required>
  The model is the AI “brain” your users will interact with.

  Currently, Cogfy Workflows supports the following OpenAI models:

  `GPT-5.4`, `GPT-5.4 mini`, `GPT-5.4 nano`, `GPT-5.2`, `GPT-5.1`, `GPT-5`, `GPT-5 mini`, `GPT-5 nano`, `GPT-4.1`, `GPT-4.1 mini`, `GPT-4.1 nano`, `GPT-4o`, `GPT-4o mini`, `GPT-4`.

  <Info>
    For more information about these models, please refer to the [official OpenAI documentation](https://developers.openai.com/api/docs).
  </Info>
</ResponseField>

<ResponseField name="Instructions" type="string">
  Also known as the *system prompt*, this field is used to guide the AI’s overall behavior.

  Use it to define personality, boundaries, verbosity, and other behavioral aspects.

  This field is very important because it affects **everything the AI generates afterward**, so use it carefully.
</ResponseField>

<ResponseField name="Max Output Tokens" type="number" required>
  The upper limit on how many tokens the model can generate in its response.

  * If the limit is low → shorter, more concise responses
  * If the limit is high → longer, more detailed responses

  If the AI reaches this limit:

  The response stops immediately. It may be cut off mid-sentence or appear incomplete.

  #### Example:

  <Prompt actions={[]} description="The main reasons for this are economic, political, and histor—" />

  <Info>
    A **token** is a fundamental unit of data — such as a word, part of a word, or character — that AI models use to process language.

    Large Language Models (LLMs) break text into tokens instead of reading whole words. Tokens serve as both input (for understanding) and output (for generating responses), and are also used as the basis for pricing and usage limits.

    For pricing, please refer to the [official OpenAI documentation](https://developers.openai.com/api/docs).
  </Info>
</ResponseField>

## Output Schema

The output schema defines the properties available to subsequent nodes.

<ResponseField name="messages" type="message_object[]">
  An array of messages generated by the model, including assistant messages and tool calls. Does not include the user message.
</ResponseField>

<ResponseField name="lastMessage" type="message_object">
  The last message generated by the model, including assistant or tool call messages. Does not include the user message.
</ResponseField>

### `message_object`

Represents a single message.

<ResponseField name="role" type="string">
  The role of the message sender.

  Can be:

  * `user` — sent by the user
  * `assistant` — generated by the AI model
  * `tool` — generated by a tool call
</ResponseField>

<ResponseField name="content" type="string | null">
  The content of the message. Can be null if no content is present.
</ResponseField>

<ResponseField name="toolCalls" type="tool_call[]">
  A list of tool calls made by the assistant in this message, if any.

  <Expandable title="tool_call">
    <ResponseField name="id" type="string">
      Unique tool call identifier.
    </ResponseField>

    <ResponseField name="name" type="string">
      The name of the tool being called.
    </ResponseField>

    <ResponseField name="arguments" type="string | null">
      Arguments provided for the tool call. Can be null if no arguments are provided.
    </ResponseField>
  </Expandable>
</ResponseField>

<ResponseField name="toolCallId" type="string | null">
  The ID of the tool call associated with this message, if applicable (e.g., when the role is `tool`).
</ResponseField>