🏛️ The 4 Pillars of Marketplace Policy Compliance
Immutable Slot #1 Protection (Unedited Camera Photo Rule)
Marketplace Standard: Pursuant to Handmade Marketplace Image Guidelines, physical item listings must contain authentic, unedited photography showing the real physical craft in its true condition.
ArtisanStudio Pro Technical Safeguard: Our platform immutably protects and reserves Slot #1 (the Primary Listing Image) for the artisan's unedited camera photograph. All staged compositions generated within our studio are exclusively routed to Slots #2 through #10 as supplementary contextual staging. Our API sync engine is hard-coded to reject any command that attempts to overwrite or delete Slot #1.
Dual-Engine Architecture & Bit-for-Bit Pixel-Lock Protection
Marketplace Standard: Digital staging or artificial intelligence must never alter, hallucinate, reshape, or misrepresent the authentic physical shape, color, texture, hallmarks, stitching, engravings, or flaws of a handcrafted item.
Dual-Engine Implementation:
- Engine 1 — 2D Optical Flatlay Compositor: Tailored for flat crafts (leatherwork, prints, textiles, jewelry trays). Isolates the craft with proprietary optical alpha-segmentation and places it on real curated surfaces (2000 × 1500 px) with color-calibrated contact shadows. Zero generative diffusion touches the craft pixels.
- Engine 2 — 3D Adaptive Studio Set with Pixel-Lock Sandwich: For volumetric crafts (mugs, candles, pottery). Generates perspective and lighting around the craft, followed by a mathematical Pixel-Lock Overlay that re-stamps the artisan's exact, unaltered camera pixels back onto the interior geometry, guaranteeing 100% optical fidelity.
Mandatory Human-in-the-Loop Review Gate (Anti-Bot Policy)
Marketplace Standard: Marketplace policies strictly prohibit unattended bot publishing, automated mass edits, or headless media mutations without active seller supervision.
Verification Gate: Neither single-item studio sessions nor batch queues can write to marketplace listings autonomously. Before the API write call (uploadListingImage) is unlocked, the seller is presented with an interactive Side-by-Side Split View and must manually check the verification gate: "I confirm that I have personally inspected this staged image, verified that my physical craft remains 100% unaltered, and approve its upload as a supplementary listing photo (Slots #2–#10)."
True-Scale Reference & Native 4:3 Aspect Ratio Safe-Zone
All images render at native 2000 × 1500 pixels (standard 4:3 marketplace search format) at 4:4:4 chroma subsampling. The studio includes an interactive Marketplace Safe-Zone overlay and a physical coin/credit card true-scale caliper tool to ensure buyers are never misled on product dimensions.
⚙️ Technical Specifications & API Governance
| Architecture Component | Production Standard |
|---|---|
| OAuth 2.0 Flow | Authorization Code Flow with PKCE (RFC 7636), SHA-256 S256 Challenge. |
| Least-Privilege Scopes | listings_r, listings_w, shops_r, profile_r. |
| API Rate Limit Throttle | Hard-throttled at maximum 2 requests/second. Automatically parses and honors X-Limit-Per-Second and Retry-After headers. |
| 429 Backoff Strategy | Exponential backoff with full randomized jitter (initial 1.5s, step x2 up to 30s) prevents server-side thundering herds. |
| Cache TTL & Purge | Volatile listings cache: 6 hours max; shop headers: 24 hours max. Automated daily cron purge at 03:00 UTC permanently erases all temporary mask cutouts. |
| Deauthorization Webhook | Upon seller disconnect or Etsy OAuth revocation, all OAuth tokens, refresh tokens, and cached listings are deleted immediately (O(1)). |
🔒 Zero Data Retention & No-Training Guarantee
100% Creator Ownership: The artisan retains full, unencumbered intellectual property rights over all uploaded craft photography and staged composite outputs.
No Model Training: Merchant product photos are strictly processed in ephemeral memory on isolated GPU microservices under enterprise Zero-Data Retention agreements. Under no circumstances are merchant images stored, shared, or utilized to train or fine-tune public or private AI models.