# Visual Context

Source: https://everywhere.sylinko.com/en/docs/chat-tools/visual-context

Even if this tool is not enabled, Everywhere will attempt to retrieve the visual context if a visual element is included when sending a message in the chat window. This tool only provides additional functions.

## Functions

| Function           | Description                                                                       | Permission(s)                |
| ------------------ | --------------------------------------------------------------------------------- | ---------------------------- |
| List Windows       | Lists all windows on the screen.                                                  | Read Screen (risk: low)      |
| Capture UI Element | Capture a screenshot of a visual element.                                         | Read Screen (risk: low)      |
| Read Visual Tree   | Read the visual tree of a window or element on screen.                            | Read Screen (risk: low)      |
| Automate Actions   | Execute a set of automated actions, such as clicks, inputs, or sending shortcuts. | Access Screen (risk: medium) |

## Notes

### Automate Actions

The "Automate Actions" feature is **experimental** and may not work as expected. Use it with caution.

The decision to execute actions on UI elements is determined by the large model itself. In most cases, **the model may be reluctant to perform such actions**, and the execution results may sometimes be unsatisfactory.

### Software Compatibility

Since the visual context is obtained via UI automation, &#x2A;*content cannot be retrieved from software that does not support accessibility features (such as WeChat)**. Additionally, applications like games are not supported.

### Real-time Capability

The visual context acquisition works like a snapshot rather than a real-time stream, so tasks like real-time YouTube subtitle translation are not possible.
