# Background Remover

## Objective

Turn an uploaded photo into a clean cutout on a transparent background.

A processing step that separates the subject of an uploaded photo from its background and stores the result as a cutout with transparency.

## Before You Begin

This feature is being added to an application that already exists and already
works. Do not scaffold a new project, and do not assume a blank slate.

Inspect the codebase first and establish:

- The existing application structure and where code of this kind already lives.
- The framework and version in use.
- The existing design system — colours, spacing, typography, and component conventions.
- Existing UI components you can reuse instead of writing new ones.
- The existing database structure, if this feature needs to persist anything.
- The existing authentication and authorization system, if this feature is user-scoped.
- Dependencies already installed, so you don't add a library that duplicates one.
- The existing test setup and conventions.

Only start writing code once you understand the above. If the application
already implements part of this feature, extend it rather than replacing it.

## Implementation Instructions

1. Run removal as a queued background job on the app's existing job infrastructure, and reuse the existing job progress reporting rather than blocking the upload request while the image is processed.
2. Write the result in a format that carries an alpha channel. Saving a cutout as JPEG silently fills the transparency with white and destroys the entire point of the feature.
3. Cap the input before processing: enforce a maximum resolution and file size server-side, and downscale oversized images to a working resolution rather than letting a single enormous photo exhaust memory.
4. Keep the original untouched alongside the cutout so the user can revert, retry with different settings, or use the full photo elsewhere.
5. Do not present a failed or poor separation as a finished result. When confidence is low or no subject is found, say so plainly, return the original, and offer manual touch-up instead of shipping a mangled cutout.

## UI and UX Requirements

Match the application's existing design system exactly. Reuse its components,
spacing, and typography. This feature should look like it was always there.

## Responsive Requirements

Works on mobile, tablet, and desktop. Touch targets are large enough to hit on a
phone, and nothing overflows horizontally at 320px.

## Accessibility Requirements

- Fully keyboard navigable.
- Correct semantic elements and ARIA roles.
- Visible focus states.
- Meets WCAG AA contrast.
- Dynamic changes are announced to screen readers.
- Respects prefers-reduced-motion.

## Edge Cases

- Removal takes seconds to minutes depending on image size. Queue the work, return immediately, and show progress on the asset, so the upload form is never held open waiting for a result.
- The output must be written in a format that supports alpha. Any pipeline step that re-encodes or flattens the image afterwards, including a thumbnail generator or an optimisation pass, must be checked for the same trap.
- Some photos have no clear subject, or several. When the model cannot find one confidently, tell the user, keep the original, and do not present the best guess as a clean cutout.
- An unbounded input will exhaust memory or run for minutes. Reject files above a server-side size limit and downscale beyond a maximum working resolution before processing, then scale the mask back up.
- Hair, fur, and thin edges are where automatic removal fails visibly. Provide a brush that adds and removes from the mask so the user can repair the edges, with undo, rather than forcing a retry.
- Processing costs money or compute. Meter it against the app's existing usage quotas, and make the limit visible before the user starts a batch rather than failing partway through.
- A transparent cutout is invisible against a white page. Preview it over a checkerboard so the user can see what they actually got.
- A job that fails or times out must leave the original asset intact and retryable, not a half-written file in the library.

## Testing

Exercise the feature end to end in the running application. Cover every edge case
above, then run the existing test suite and confirm nothing regressed.

## Acceptance Criteria

- [ ] Removal runs as a queued job with visible progress and never blocks the upload.
- [ ] Output is written in a format preserving transparency, and no later pipeline step flattens it.
- [ ] Oversized inputs are rejected or downscaled before processing, with limits enforced server-side.
- [ ] The original image is retained and the user can revert.
- [ ] Low-confidence results are reported honestly rather than presented as finished.
- [ ] A manual brush with undo is available for repairing edges.
- [ ] Failed jobs leave the asset intact and retryable.
- [ ] The feature matches the existing design system.
- [ ] No existing functionality is broken.

## Adaptation Rules

- Match the existing design system. Do not introduce a new colour palette,
  spacing scale, or component library.
- Reuse existing components and utilities wherever they fit.
- Follow the naming, file layout, and code style already present.
- Do not upgrade, replace, or remove existing dependencies to make this
  feature fit. Adapt the feature to the app, not the app to the feature.
- Do not break existing functionality. If a change is genuinely required in
  existing code, make the smallest one that works and say so.
- If something in these instructions conflicts with how the application is
  built, follow the application and explain the deviation.

## Final Verification

Before you report the work as done:

1. Re-read the acceptance criteria above and check each one against what you
   actually built.
2. Run the application and exercise the feature end to end.
3. Run the existing test suite and confirm you have broken nothing.
4. Check the feature on mobile, tablet, and desktop widths.
5. Check keyboard navigation and focus handling.
6. Summarize what changed: files added, files modified, and anything you
   deliberately did differently because of how this application is built.

If any acceptance criterion is unmet, fix it before reporting completion.