Struct Headless 

pub struct Headless { /* private fields */ }

A reusable offscreen renderer. Creating one compiles vello’s shaders, so tests should create it once and render many scenes through it.

Implementations

impl Headless

pub fn new() -> Result<Self, ShellError>

Acquires a compute-capable adapter (no surface required) and builds a vello renderer on it.

pub fn max_dimension(&self) -> u32

The largest texture dimension the device supports; render sizes are clamped to it.

pub fn clamp_size(&self, width: u32, height: u32) -> (u32, u32)

Clamps a requested render size on both axes to the supported 1..=max_dimension() range.

pub fn render( &mut self, scene: &Scene, width: u32, height: u32, base_color: Color, ) -> Result<RgbaImage, ShellError>

Renders scene at the given pixel size over base_color and reads the result back into an RGBA image. The size is clamped to 1..=max_dimension() on both axes, so hostile dimensions cannot trigger wgpu’s fatal validation handler.

pub fn render_plan( &mut self, frame: &Frame, fonts: &mut Fonts, state: &mut FrameState, width: u32, height: u32, base_color: Color, ) -> Result<RgbaImage, ShellError>

Renders frame with real frosted-glass backdrop blur and foreground ElementFilters via the two-pass pipeline, falling back to a single pass when the frame has neither.

The fast path is byte-for-byte identical to render: a frame with no filtered node produces an empty plan, so the backdrop scene is the final image and nothing is read back or reprocessed. Otherwise the backdrop scene (every glass subtree skipped) is rendered and read back, each region is blurred/filtered on the CPU (process_specs), and the final scene composites those images. width/height are physical pixels.