> For the complete documentation index, see [llms.txt](https://help.botpenguin.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.botpenguin.com/bots/sms-bot/ai-settings/ai-settings-latest.md).

# AI Settings (Latest)

### What is AI Settings?

AI Settings is where you define how your chatbot thinks, speaks, and operates. It gives you control over the instructions your bot follows, how it communicates with users, what data it extracts from conversations, and which AI model powers it.

AI Settings is divided into four tabs:

| Tab                     | Purpose                                                               |
| ----------------------- | --------------------------------------------------------------------- |
| **Prompt**              | Define the bot's core instructions and conversation flow logic        |
| **Communication Style** | Control tone, language, response length, and formatting               |
| **Meta Data**           | Configure which user attributes the bot extracts during conversations |
| **Advance Setting**     | Set up your AI provider, model selection, and fallback behaviour      |

***

### When Should You Use This?

Use this section when:

* You are setting up your chatbot for the first time and need to define its behaviour
* You want to update the bot's instructions, tone, or language
* You need to connect your own AI API key or switch to a different model
* You want to configure fallback behaviour when AI credits run out

***

### How to Access AI Settings

1. From the left sidebar, click on **"Chatbot"** and select the bot you want to configure.
2. From the bot's left menu, click on **"AI Settings"**.

<figure><img src="/files/8JMFfnNo2yPn9S4UhfIg" alt=""><figcaption></figcaption></figure>

You will see four tabs at the top: **Prompt**, **Communication Style**, **Meta Data**, and **Advance Setting**.

***

### Tab 1 — Prompt

{% hint style="info" icon="bullseye-arrow" %}
**Goal**: The Prompt tab is where you define what your bot is supposed to do and how it should handle conversations. There are two fields to configure.
{% endhint %}

3. Click on the **"Prompt"** tab.

<figure><img src="/files/QwjxDoSBxcmC3xbXU1Y6" alt=""><figcaption></figcaption></figure>

4. In the **Core Instructions** field, enter the main directive that defines the bot's purpose, role, and capabilities. This is the primary instruction the bot follows in every conversation. Use `/` to reference custom attributes or automation tools inline.

> **Example:** *"Act as a customer support assistant for an e-commerce store. Help users track orders, process returns, and answer product questions."*

<figure><img src="/files/3sjZUtDppNdOJ4ASZNMK" alt=""><figcaption></figcaption></figure>

{% hint style="info" %}
To learn the best way to make a prompt, please refer to: [Setting Up Your AI Agent: Guidelines and Best Practices](/ai-agents/creating-ai-agents-in-botpenguin/setting-up-your-ai-agent/setting-up-your-ai-agent-guidelines-and-best-practices.md)
{% endhint %}

5. In the **Conversation Flow Instructions** field, define the step-by-step logic for how the bot should drive the conversation. Use this to set conditions, guide the flow, or trigger automations. Use `/` to reference attributes or conditions dynamically.

> **Example:** *"Start with a greeting. Ask what the user needs help with. If the query includes 'return', guide them through the returns process."*

<figure><img src="/files/CxDzDl8P0ONC8dXbb42w" alt=""><figcaption></figcaption></figure>

{% hint style="info" %}
To learn the best way to make a prompt, please refer to: [Conversation Flow Setup & Best Practices](/ai-agents/creating-ai-agents-in-botpenguin/setting-up-your-ai-agent/conversation-flow-setup-and-best-practices.md)
{% endhint %}

6. Click **"Update"** or **"Update & Next"** to save.

<figure><img src="/files/ouA5T3a5GyBqwJWGZGBi" alt=""><figcaption></figcaption></figure>

***

### Tab 2 — Communication Style

{% hint style="info" icon="bullseye-arrow" %}
**Goal**: The Communication Style tab controls how your bot speaks — its tone, language, response length, and formatting.
{% endhint %}

7. Click on the **"Communication Style"** tab.

<figure><img src="/files/Kt9aOdGuVPSS4EA0NCos" alt=""><figcaption></figcaption></figure>

8. Configure each of the following options:

| Field                            | Options                                                     |
| -------------------------------- | ----------------------------------------------------------- |
| **Answer Length**                | Very Short, Short, Medium, Long, Let Agent Decide           |
| **Answer As**                    | First Person, Second Person, Third Person, Let Agent Decide |
| **Tone**                         | Formal, Informal, Joyful, Sincere, Let Agent Decide         |
| **Language**                     | English, Hindi, Spanish, French, German, Let Agent Decide   |
| **Answer Formatting**            | Bullet List, Numbered List, Paragraph, Let Agent Decide     |
| **Include Sources for Response** | Toggle ON to show the source of the bot's answers           |
| **Ice Breaker**                  | Add ice breakers for your AI Agent component.               |

<figure><img src="/files/CkhR5f0YS8DxmNvZK09O" alt=""><figcaption></figcaption></figure>

9. Click **"Update"** or **"Update & Next"** to save.

***

### Tab 3 — Meta Data

{% hint style="info" icon="bullseye-arrow" %}
**Goal**: The Meta Data tab lets you define which user attributes the bot should automatically extract and store during a conversation — such as names, dates, locations, or preferences.
{% endhint %}

10. Click on the **"Meta Data"** tab.

<figure><img src="/files/vugPBPK3ZhfBF72H3u4D" alt=""><figcaption></figcaption></figure>

11. You will see a list of attributes. Each row shows:
    1. **Attribute Name** — the label identifying the data point
    2. **Category** — System or Custom
    3. **Extract Meta Data** — toggle to enable or disable extraction for that attribute
    4. **Actions** — delete button to remove the attribute from scope
12. To create new attributes or modify existing ones, click **"Manage Attribute"**. This opens the Custom Attributes. Refer to the Managing Custom Attributes guide for full configuration details.

<figure><img src="/files/flguM1iiXNIFZcfTys1R" alt=""><figcaption></figcaption></figure>

{% content-ref url="/pages/6ttpiKsmzt3DBoquI8zO" %}
[Managing Custom Attribute](/settings/managing-custom-attribute.md)
{% endcontent-ref %}

***

### Tab 4 — Advance Setting

{% hint style="info" icon="bullseye-arrow" %}
**Goal**: The Advance Setting tab is where you configure which AI provider and model powers your bot, and what happens when your credits or token limit runs out.
{% endhint %}

There are two sections inside: **LLM Provider** and **Advance**.

***

#### LLM Provider

13. Click on **"Advance Setting"**, then expand the **"LLM Provider"** section.

<figure><img src="/files/q5dshIulKw8rOqcVqo2g" alt=""><figcaption></figcaption></figure>

14. Choose how you want to power your bot's AI:
    1. **Add LLM Credentials** — Connect your own API key from OpenAI or OpenRouter.
    2. **Purchase Add-ons** — Uses BotPenguin's built-in AI credits. No API key required.
15. If you select **"Integrate Your Own Key"**, choose your provider:
    1. **OpenAI** — Paste your OpenAI API key directly. See the [OpenAI Integration guide](/settings/integrations/chatgpt.md) for steps to generate your key.
    2. **OpenRouter** — Select your saved OpenRouter configuration from the dropdown, then choose the specific model you want to use. See the [OpenRouter Integration guide](/settings/integrations/openrouter.md) for setup steps.
    3. Select the Credential configured in the "**LLM Provider Credential**" dropdown, and the LLM Model to use.

<figure><img src="/files/Gvfa3DMVlW4zNp9oLpWl" alt=""><figcaption></figcaption></figure>

16. Click **"Update"** to save your LLM provider settings.

<figure><img src="/files/q6I3fihTJkmldwrsKnRA" alt=""><figcaption></figcaption></figure>

***

#### Advance

17. Expand the **"Advance"** section to configure fallback and response behaviour.

<figure><img src="/files/ptLWBhWowq9LzH0yjxuv" alt=""><figcaption></figcaption></figure>

18. Configure the following fields:

| Field                                        | Description                                                                                                    |
| -------------------------------------------- | -------------------------------------------------------------------------------------------------------------- |
| **Switch model after exceeding token count** | Toggle ON to automatically switch to a fallback model once the max token count is reached                      |
| **Model after switching**                    | The fallback model the bot uses once the primary model's token limit is exceeded                               |
| **Search Mode**                              | Controls how the bot searches its knowledge base. **Fast Search** prioritises speed.                           |
| **Max Token Count**                          | Maximum tokens the AI can use per response. Default is **4096**.                                               |
| **Temperature**                              | Controls response creativity. **0** = precise and consistent. **1** = varied and creative. Default is **0.5**. |

**Available models for fallback:**

* gpt-5.4
* gpt-5.1
* gpt-5.1-chat-latest
* gpt-5.2
* gpt-5.2-chat-latest
* gpt-4.1
* gpt-4.1-mini

19. Click **"Update"** to save.

<figure><img src="/files/rct3oustgU56RwROGtFQ" alt=""><figcaption></figcaption></figure>

***

### Best Practices

> 💡 Write Core Instructions in plain language — clear, role-specific prompts produce better and more consistent bot responses.

> ⚠️ Always configure a fallback model under Advance Setting so your bot continues working if primary credits run out.

> 🎯 Set Temperature to **0.3–0.5** for support or FAQ bots. Use higher values only for creative or open-ended conversation bots.

> 🔄 After making any changes to AI Settings, test your bot using the Bot Preview panel on the right before publishing.

***

### Troubleshooting

<details>

<summary>Bot is not following the instructions set in the Prompt tab</summary>

Ensure you clicked **Update** after saving.

Check that the instructions are clear and do not contradict each other across Core Instructions and Conversation Flow Instructions.

</details>

<details>

<summary>Bot stops responding after some time</summary>

Your AI credits or token limit may be exhausted.

Check your API account balance and ensure a fallback model is configured under the Advance section.

</details>

<details>

<summary>OpenRouter configuration is not appearing in the LLM Provider dropdown</summary>

Ensure you have successfully connected OpenRouter and saved at least one configuration under **More → Integrations → OpenRouter**.

</details>

<details>

<summary>Responses feel too random or off-topic</summary>

Lower the Temperature value (try **0.3 or below**) and click **Update**.

</details>

<details>

<summary>Meta Data attributes are not being extracted during conversations</summary>

Confirm the **Extract Meta Data** toggle is enabled for the relevant attributes in the Meta Data tab.

</details>

<details>

<summary>Communication style settings are not reflected in bot responses</summary>

Click **Update** after making changes.

If the issue persists, try refreshing the page and reapplying your settings.

</details>

***

### FAQs

<details>

<summary>Can I use my own API key instead of platform credits?</summary>

Yes. Under the Advance Setting tab, select **Integrate Your Own Key** under LLM Provider and connect either an OpenAI or OpenRouter key.

</details>

<details>

<summary>What happens if my API credits run out?</summary>

If no fallback model is configured, the bot will stop generating AI responses.

Configure a fallback model under the Advance section to prevent this.

</details>

<details>

<summary>Can I set the bot to respond in a language other than English?</summary>

Yes. Under the Communication Style tab, use the Language dropdown to select your preferred language or set it to **Let Agent Decide** for automatic detection.

</details>

<details>

<summary>What is the difference between Core Instructions and Conversation Flow Instructions?</summary>

Core Instructions define the bot's overall role and purpose.

Conversation Flow Instructions define the step-by-step logic for how the bot drives individual conversations — conditions, routing, and interaction sequences.

</details>

<details>

<summary>What does Temperature control?</summary>

Temperature controls how varied the bot's responses are.

A low value (0.1–0.3) produces focused, predictable answers. A high value (0.7–1.0) produces more creative, varied responses.

For most business bots, 0.3–0.5 is recommended.

</details>

<details>

<summary>What is Max Token Count?</summary>

Tokens are units of text the AI processes.

Max Token Count limits how long each AI response can be.

The default of 4096 covers most use cases. Lower it to reduce API costs; raise it if responses are being cut off mid-sentence.

</details>

<details>

<summary>Can I create custom attributes for Meta Data extraction?</summary>

Yes. Click **Manage Attribute** inside the Meta Data tab.

This takes you to the Custom Attributes section where you can create, edit, and configure attributes.

Refer to the Managing Custom Attributes guide for details.

</details>

<details>

<summary>Can I switch AI models without disconnecting my API key?</summary>

Yes. Update the model selection under LLM Provider or the fallback model under Advance at any time and click **Update** — no need to reconnect your key.

</details>

***

### Related Articles

* [OpenAI Integration](https://claude.ai/chat/be7c1a8a-b08d-46cf-9b0b-0cb87d243a8a#)
* [OpenRouter Integration](https://claude.ai/chat/be7c1a8a-b08d-46cf-9b0b-0cb87d243a8a#)
* [Managing Custom Attributes](https://claude.ai/chat/be7c1a8a-b08d-46cf-9b0b-0cb87d243a8a#)
* [Bot Training](https://claude.ai/chat/be7c1a8a-b08d-46cf-9b0b-0cb87d243a8a#)

{% content-ref url="/pages/vyRdBTzfhsij31WXcff8" %}
[ChatGPT](/settings/integrations/chatgpt.md)
{% endcontent-ref %}

{% content-ref url="/pages/iA8HJvWgMhEJazf1r6bP" %}
[OpenRouter](/settings/integrations/openrouter.md)
{% endcontent-ref %}

{% content-ref url="/pages/6ttpiKsmzt3DBoquI8zO" %}
[Managing Custom Attribute](/settings/managing-custom-attribute.md)
{% endcontent-ref %}

{% content-ref url="/pages/HOgnTrcXuPrLpu3vZMkS" %}
[Best Practices for Training Your AI](/bots/website-mobile-app-bot/train-your-ai-chatbot/best-practices-for-training-your-ai.md)
{% endcontent-ref %}

***

### Support

If you still have questions for our team, write to us at **<support@botpenguin.com>**. We'll get back to you within 48 hours.

***

🎉 You have successfully configured AI Settings for your SMS Chatbot. Your bot is now ready to engage users with the right intelligence, tone, and behaviour.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## 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://help.botpenguin.com/bots/sms-bot/ai-settings/ai-settings-latest.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.