# Safety

## Masking personal data

We recommend that when communicating with a client, you request their **consent to the processing of personal data**.

To enable masking, go to additional settings, select the Security tab and enable the Mask personal data flag.

<figure><img src="/files/C36WZIAA1oMS4B7KzzDk" alt=""><figcaption></figcaption></figure>

### How does the process of masking personal data work?

1. **Masking data**

* Personal data (for example, passports and phone numbers will be replaced with masked values)

2. **Transfer of anonymized data**

* Once masked, the data is no longer considered personal and is passed to the model for processing.

3. **Unmasking and transmitting the response**

* The received response will be unmasked back to the server of your country.

Thus, your personal data remains on the **servers of your country**, and the model receives **anonymized data**.

### What does masking look like in chat?

{% hint style="info" %}
In the chat manager and dialog export, we've added a modal window with information about how the model generated the response — what the message was before and after masking, what the instructions were, and what the dialog history is.\
All of this contributes to better understanding when analyzing the bot's responses.
{% endhint %}

To see the masking of **Personal Data** in the **Chat Manager** mode, click on the icon in the message field

<figure><img src="/files/TjVzGR2i0fDOzp8UiHus" alt=""><figcaption></figcaption></figure>

A modal window will open, along with the **bot's last message with a transcript**—what the message looked like when personal data was masked, and what message was **sent to the Client**:

<figure><img src="/files/XEvIkPfKVwwz29BabcnB" alt=""><figcaption></figcaption></figure>

### Media transfer in LLM

<figure><img src="/files/S69JN0jkTjovvnEZDVzL" alt=""><figcaption></figcaption></figure>

This setting prevents media files from being passed to the LLM model for processing.

{% hint style="warning" %}
Please note that this setting overrides the Image Recognition setting.
{% endhint %}

### Data storage

Here you can set the time period after which messages and dialogs will be deleted from all Savvi servers.

<figure><img src="/files/rF6nnIs1voo33Ml4Nfmq" alt=""><figcaption></figcaption></figure>

**Message storage (days)** — The storage period for dialog messages. After a specified period of time, all messages that exceed the specified time period will be deleted.

**Dialog  storage (days**) — the period after which dialogues will be deleted if no events have occurred in the dialogue during the selected period of time.

**Delete dialogs and message history** —  deleting all current dialogs (at the time of deletion activation) regardless of events in the dialog.

{% hint style="warning" %}
Please note that **by defaul**t, we store customer conversations for up to **90 days**. After that, messages and inactive conversations are deleted.

When integrating with MedFlex, the maximum data storage period is 7 days.
{% endhint %}


---

# 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.suvvy.ai/en/main-settings/dopolnitelnye-nastroiki/safety.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.