max-processor-output-viewer

Interactive visualizer for the JSON output of the max-processor-output service (omniplatform/OmniServices/max-processor-output). It draws every bounding box — group / block / shelf / product — plus product dots in the original pixel coordinate space, and lets you inspect the full data structure of any item.

Zero dependencies — just Node (>= 14). Runs via npx, opens a local page in your browser.

Usage

# from a file
npx max-processor-output-viewer output.json

# from a pipe (e.g. straight from curl / a log)
cat output.json | npx max-processor-output-viewer

# no input — paste JSON in the browser
npx max-processor-output-viewer

During local development (from this directory):

node cli.js path/to/output.json

Options

| Option | Description | | --- | --- | | -p, --port <n> | Port to listen on (default: a random free port) | | --no-open | Don't auto-open the browser | | -h, --help | Show help |

What it shows

• Bounding boxes for groups (dashed grey), blocks (colored, labeled block N · shelves×columns), shelves (S<n>), and products (colored, labeled with SKU).
• Product dots (productDot) placed in the inferred image space.
• Reference product (referenceProduct) marked with a ★.
• Empty facings (isEmpty) drawn dashed; unclassified products drawn with a dotted border.

Interaction

• Scroll to zoom, drag to pan, Fit to reset the view.
• Hover a product for a quick tooltip; click it for a full detail panel (all fields + raw JSON).
• Structure tree (block → shelf → column → product) on the left; click a leaf to select and center it.
• Color by block, group, confidence (cossum, red→green gradient), or SKU.
• Search SKU to highlight all matching facings and dim the rest.
• Layer toggles for groups / blocks / shelves / products / dots / labels / empty facings / reference.
• Summary stats: block/shelf/product/unique-SKU counts, average cossum, low-confidence count, empty facings, groups.

Input format

Accepts either the full response wrapper ({ "data": { … }, "taskId": … }) or a bare data object. Multiple concatenated JSON documents (batch output) are supported — a document selector appears in the header, and stray/garbage fragments between documents are skipped.

How coordinates are interpreted

All boxes use the service's pixel coordinates (x, y, w, h) with y increasing downward, matching image space. productDot is normalized 0..1; the viewer infers the image dimensions from the products' box centers vs. their dots to place the dots and draw a dashed image frame.