# Other generative AI settings

This section walks you through the advanced settings that let you customize your Agent's behavior and optimize its responses for your specific needs.

### AI settings

Open your Agent's  Settings and click  **AI** to access the dedicated section. From here, you can:

1. [**Enable or disable generative AI**](/en/generative-ai/activate-generative-ai-and-change-model.md#how-to-enable-generative-ai-for-your-agent);
2. Enable or disable [**Deep Thinking**](/en/extensions/deep-thinking.md);
3. Edit the [**instructions**](/en/generative-ai/ai-instructions.md);
4. [**Change the AI models**](/en/generative-ai/activate-generative-ai-and-change-model.md#how-to-change-your-agents-model) your Agent uses.

Below the instructions field, you'll also find two options:

* [Always respond with generative AI](#always-respond-with-generative-ai);
* [Automatic media extraction](#disable-automatic-image-extraction-from-contents).

Inside the "Provider settings" section, you'll also find:

* [Maximum token count](#maximum-token-count);
* [Temperature](#temperature).

Let's look at each one in detail 👇

#### Always respond with generative AI

When this setting is enabled, **the Agent will always and only respond using generative AI**: your Agent will no longer generate answers on its own — instead, it will have the chosen model formulate them (using the "Q\&A configuration").

**The Agent will share the most relevant contents with the chosen language model**, and the model will generate the response for the user. This lets you keep responses relevant and accurate without sacrificing personalization!

{% hint style="info" %} <mark style="color:blue;">**Generative AI may provide incorrect responses**</mark><mark style="color:blue;">. Make sure to test every use case scenario for your Agent to confirm it continues to respond correctly.</mark>
{% endhint %}

#### Disable automatic image extraction from contents

As covered in the [contents](/en/agent-training/contents.md) section, you can add media and links to specific Q\&A pairs.

By default, the AI Agent can read the titles of images and links associated with the most relevant contents and may show them to the user even in AI-generated responses.

If you experience issues, you can check this option to disable that behavior.

#### Maximum token count

This parameter controls the **maximum length of each AI-generated response**.

On the platform, you can select between 256 and 4,000 maximum tokens per AI-generated response (the default is 512). If you're using a model that supports a higher output token count, you can adjust this value via the API.

To roughly convert tokens to characters: **1 token ≈ 4 characters** in English.

{% hint style="info" %} <mark style="color:blue;">**Not all models support a high output token count**</mark><mark style="color:blue;">. If you're using your own API key, make sure to check the model's maximum output token limit.</mark>
{% endhint %}

#### Temperature

Temperature controls how creative your Agent's responses are. It's a number from 0 (more predictable and consistent responses) to 1 (less predictable and more creative responses).

**When creating an Agent, we recommend setting temperature to 0**: you'll get more consistent responses. If you're looking for more creative answers, you can gradually increase the temperature.

### Optimizing settings

Finding the perfect configuration for your Agent may take some trial and error. Don't hesitate to experiment with these settings to find the right balance between accuracy, creativity, and performance.

Every project has unique needs. Use these powerful options to shape your Agent and make it perfect for your specific use case!


---

# 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.aisuru.com/en/generative-ai/advanced/other-generative-ai-settings.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.