Local runtime

LayerForge browser interface

This page is usable as a project overview on GitHub Pages, but the run controls require the local layerforge webui server so the browser can call the existing Python pipeline on uploaded images.

Interactive workflow

Run LayerForge without the CLI

The local interface is organized around three tasks: select the operating mode, provide targeting hints only when they matter, and choose execution overrides when the default configuration is not enough.

Choose the run mode

Start with frontier selection when you want the strongest measured candidate bank, then move to the narrower extraction, transparent recovery, or recursive peeling modes when the task is specific.

Add only the hints you need

Prompts, points, and boxes are optional. Leave them blank for a general decomposition and add them only for targeted edits.

Review artifacts in-place

The browser keeps the manifest, metrics, DALG export, and preview images together so editors do not need to chase CLI output folders.

Core inputs

Image and pipeline mode

These controls define the main run. For most cases, choose the mode, upload the image, and keep the configuration defaults.

Mode Image Config Segmenter override Depth override Device

Targeting hints

Prompt, point, or box guidance

Use these only when the task is targeted. They are optional for a general layered decomposition.

Prompt Point hint Box hint Prompt source

Execution controls

Ordering and runtime overrides

Use these controls when comparing ordering behavior or constraining a peeling run.

Ordering Max layers (peel)

Skip parallax export for faster local runs Use frontier candidate-bank selection as the strongest base run for run, extract, and transparent modes

Start the local server with layerforge webui --open-browser.

Run output