Kann Magnific AI weg?

Build a closest honest personal substitute for Magnific AI in an empty repository.
Use Python 3.12, FastAPI, SQLite, ComfyUI as a local worker, and a small React frontend; do not offer alternative stacks.
The core loop is: queue local upscaling and enhancement experiments, submit jobs to a user-owned model server, and keep settings and outputs reproducible.
Make the first run work locally with one documented command.
Store all user data locally by default and make export straightforward.
Put secrets in .env, ship .env.example, and never commit credentials.
Create prompt, negative-prompt, seed, dimensions, model, and workflow controls.
Submit jobs only to the local ComfyUI endpoint configured in .env.
Record exact generation parameters and workflow JSON beside every output.
Build a searchable contact sheet with compare, favorite, annotate, and rerun actions.
Support local image-to-image and mask inputs without uploading them elsewhere.
Show estimated VRAM needs and fail clearly when a workflow or model is missing.
Include clear empty, loading, success, and recoverable error states.
Add input validation, safe filenames, and graceful handling of unavailable APIs.
Write focused tests for the core transformation and one end-to-end happy path.
Create a README with setup, architecture, permissions, data location, and backup steps.
Do not add accounts, billing, telemetry, analytics, or a hosted control plane.
Do not claim to reproduce proprietary data, network liquidity, regulated access, or frontier infrastructure.
Deliberately leave out training a new frontier model.
Deliberately leave out copying a vendor's proprietary model or dataset.
Deliberately leave out public generation hosting and moderation.
Finish by running the tests and listing the exact commands used.