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

# Configure AI Agent Tone, Persona, and Behavior Settings

> Customize how your Beenos Solutions AI agent responds — set its persona, response style, citation format, conversation limits, and confidence threshold.

Once your agent is created, you can fine-tune how it communicates with your users. Beenos Solutions gives you controls for shaping your agent's personality, adjusting the length and format of its responses, and setting boundaries on how conversations unfold. All behavior settings are found on the **Behavior** tab of your agent's detail page.

## Persona & tone

The persona settings define your agent's voice and the constraints it operates within.

### System prompt

The **System Prompt** is a free-text field where you write instructions that shape your agent's personality and scope before any user interaction begins. It supports Markdown formatting for readability. Use it to define who the agent is, how it should communicate, and any topics it should avoid.

```text theme={null}
You are a helpful support assistant for Acme Corp. Your job is to answer
questions about Acme's products and policies using only the provided
knowledge base. Always respond in a professional yet approachable tone.
If you cannot find a relevant answer, tell the user clearly and suggest
they contact support@acme.com.
```

```text theme={null}
You are Aria, a friendly onboarding assistant. Welcome new users warmly,
use simple language, and avoid technical jargon. Never speculate about
pricing — direct pricing questions to the sales team.
```

### Tone presets

If you prefer not to write a full system prompt, choose one of the built-in **Tone Presets** to apply a baseline communication style:

* **Professional** — Formal, precise language suitable for enterprise or legal contexts.
* **Friendly** — Warm and conversational, ideal for consumer-facing products.
* **Concise** — Short, direct answers that minimize prose; well-suited for technical audiences.

<Info>
  Tone presets and a custom system prompt can be used together. The preset sets a baseline, and your system prompt adds specific instructions on top of it.
</Info>

## Response settings

Response settings control the structure and length of what your agent returns.

### Response length

Choose how much detail the agent includes in each answer:

| Option       | Best for                                                                         |
| ------------ | -------------------------------------------------------------------------------- |
| **Short**    | Quick factual lookups, FAQs, or mobile interfaces.                               |
| **Medium**   | General-purpose support and onboarding flows.                                    |
| **Detailed** | Technical documentation, troubleshooting guides, or users who need full context. |

### Citation style

Select how the agent references the source content it used:

* **Inline** — Citations appear as linked phrases within the answer text.
* **Footnote** — Sources are listed in a numbered reference block at the end of the response.
* **None** — Responses contain no visible citations. Use this only if your users do not need to verify sources.

### Confidence threshold

The **Confidence Threshold** is a numeric value (0–100) that controls how certain the agent must be before returning an answer. If the retrieval score for the best matching content falls below this threshold, the agent delivers the fallback message instead of a low-confidence answer. Raise the threshold to make the agent more conservative; lower it to allow answers even when source relevance is weaker.

<Warning>
  Setting the confidence threshold too low may cause the agent to return loosely relevant or misleading answers. Start with the default value and adjust based on what you observe in the Chat Playground.
</Warning>

## Conversation settings

These settings govern how the agent manages multi-turn conversations within a session.

### Max turns per session

Set the maximum number of message exchanges allowed in a single session. Once the limit is reached, the agent informs the user and closes the conversation. This prevents runaway sessions and helps manage API usage costs.

### Context window behavior

Choose how much conversation history the agent carries forward when generating each response:

* **Full history** — The agent considers the entire conversation when answering, which is best for complex multi-step support flows.
* **Last N turns** — The agent only looks at the most recent N exchanges, reducing token usage for long conversations.

### Session timeout

Set the number of minutes of inactivity after which a session expires. When a session times out, the next message from the user starts a fresh conversation with no prior context.

## Testing configuration changes

After updating any behavior setting, verify the effect before it reaches your users.

1. Open the **Chat Playground** tab on the agent's detail page.
2. Type a sample question that reflects a real user query.
3. Review the response for tone, length, citations, and accuracy.
4. Adjust settings and repeat until the output meets your expectations.

Configuration changes take effect immediately — you do not need to redeploy the agent after updating behavior settings.

<Accordion title="What is a system prompt?">
  A system prompt is a set of instructions delivered to the AI model before any user message is processed. It is invisible to the end user but shapes every response the agent produces. You can use the system prompt to give the agent a name and persona, restrict it to certain topics, define a communication style, or specify how it should handle edge cases. Think of it as the briefing you give a new team member before their first customer interaction — it sets the rules and expectations the agent will follow throughout every conversation.
</Accordion>