> ## Documentation Index
> Fetch the complete documentation index at: https://visorai.dev/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Evidence capture

> See how visor captures screenshots and UI source, what metadata it returns, and where files are written.

Evidence capture is one of visor's core capabilities. It turns app state into files and structured metadata that an LLM or human can inspect.

## Screenshot capture

The `screenshot` action captures the current screen and writes a PNG file.

Returned metadata includes:

* `label`
* `file`
* `path`
* `width`
* `height`

Behavior details:

* if a path is provided directly, visor writes to that path
* if no path is provided, visor writes `<label>.png`
* parent directories are created automatically
* visor uses Appium screenshot persistence

## UI source capture

The `source` action captures the current UI hierarchy and writes an XML file.

Returned metadata includes:

* `label`
* `file`
* `path`
* `format`, which is `xml`
* `bytes`

Behavior details:

* if a path is provided directly, visor writes to that path
* if no path is provided, visor writes `<label>.xml`
* parent directories are created automatically
* visor writes the current Appium `page_source`

## Automatic paths during scenario runs

When `screenshot` and `source` are executed inside a scenario run, visor injects deterministic per-run paths before the adapter call.

Injected directories:

* screenshots go under `<output>/<run-id>/screenshots/`
* UI sources go under `<output>/<run-id>/sources/`

The file name is based on the step label.

Examples:

* `<output>/<run-id>/screenshots/app-opened.png`
* `<output>/<run-id>/sources/counter-after-increment-ui.xml`

## Artifact collection

During a scenario run, visor records screenshot and source paths as run artifacts.

Later, the report writer copies these artifacts into the final report structure and renames them with an ordered numeric prefix.

Example final names:

* `001-app-opened.png`
* `002-counter-after-increment-ui.xml`

## Why both screenshot and source matter

The two capture types answer different questions.

| Evidence type | Best for                                                    |
| ------------- | ----------------------------------------------------------- |
| Screenshot    | Visual layout, styling, and state visible on screen         |
| UI source     | Structure, text values, identifiers, and machine comparison |

For reliable verification, many scenarios should collect both.

## Recommended labeling practice

Labels are not required in every case, but they are strongly recommended.

Good labels make artifacts easier to compare over time.

Examples of useful labels:

* `before-login`
* `after-login`
* `cart-opened`
* `checkout-form`
* `confirmation-screen`

In scenario validation, missing screenshot labels trigger a determinism warning.