Déclic
From a photo to your store, in one click — the AI studio that turns a product photo into a publishable listing, rolls the catalog out across channels and translates it
Discuss a similar projectThe live site
Click the preview to open and test Déclic directly in your browser
declic-app-production.up.railway.app
Réponse courte
Déclic is an AI-powered e-commerce studio built by Digit-AI during the Vibe Coding Arena 2026 hackathon (the 'Retail & E-commerce' brief). It turns a simple product photo into a ready-to-publish listing — title, SEO description, bullet points, attributes, keywords and a staged visual — in about ten seconds, versus around twenty minutes manually. It then rolls an entire catalog out in batch, exports it across multiple channels (Shopify, Instagram) and translates it via DeepL. The chain relies on Anthropic vision and copywriting, fal.ai for the image, remove.bg for background removal, with all keys kept server-side and a deterministic fallback on every integration.
The product listing, the small merchant's bottleneck
For a maker or a small brand, creating a listing takes ~20 minutes, the catalog is poorly ranked and multilingual content stays out of reach
Independent merchants, makers and small brands spend about 20 minutes per listing writing a title, a description, attributes — with no time left for SEO or translation. The result: incomplete, poorly ranked, single-language catalogs. Déclic turns this manual work into an automated chain that starts from a simple photo.
~20 minutes per listing
Manual copywriting is slow and tedious, multiplied by every product in the catalog.
Neglected SEO
For lack of time and expertise, titles and keywords aren't optimized, and products stay invisible.
Raw visuals
Photos taken on the corner of a table, with no background removal: an unprofessional presentation.
No multilingual content
Going international stays inaccessible for lack of time to translate and adapt each listing.
Functional scope
Independent merchants, makers, small brands. Processes covered: photo capture, SEO listing generation, staged visual, inline editing and tone switching, batch catalog, multichannel export and translation.
Key constraints
Built in 24h (hackathon) · 100% Railway monorepo (web + api + Postgres) · Integration keys exclusively server-side · Free-tier, no scraping · Deterministic fallback per integration.
What Déclic solves
Three axes to go from a photo to a publishable, optimized and international catalog
Speed to market
Cut listing creation from ~20 minutes to about ten seconds, photo included, to publish without waiting.
Quality & ranking
Produce SEO-optimized listings and staged visuals with removed backgrounds, for a consistent professional presentation.
Multichannel & multilingual distribution
Roll the catalog out across multiple channels and translate it via DeepL to open up international markets without manual copy-paste.
What Déclic enables
From the photo to multichannel publishing, a complete chain for creating product listings
Photo → structured attributes
The Anthropic vision API analyzes the photo and extracts structured attributes (type, color, material, style) that form the backbone of the listing.
Complete SEO product listing
Generation of the title, SEO description, bullet points, attributes and keywords — a listing ready to publish, not just a paragraph.
Staged visual
Generation of a product visual via fal.ai, launched in the background and non-blocking: the text listing never waits for the image.
Inline editing & tone switching
Correct an attribute or change the tone (warm, premium, technical…) with immediate regeneration of the affected content.
Batch catalog & multichannel
Process several photos in parallel, then roll the catalog out across multiple channels (Shopify CSV export, Instagram caption).
Translation & SEO coach
Multilingual translation via DeepL, an SEO score with suggestions (local heuristics) and background removal via remove.bg.
Tech stack
A TypeScript monorepo deployed 100% on Railway, with all AI integrations orchestrated server-side
Frontend (apps/web)
Backend (apps/api)
Artificial intelligence
Integrations
Data & ORM
Auth & infra
Differentiating pattern — withFallback(): every external integration (vision, copywriting, fal.ai, remove.bg, DeepL) is wrapped by a single wrapper that handles timeouts, errors, quotas and a switch to a deterministic fixture. No integration is ever called without it — and the text listing, synchronous, stays the reliable core while the visual is generated in the background.
Architecture diagrams
A Railway monorepo: a React SPA that only talks to a Node/Express orchestration API, which drives all integrations and Postgres persistence
User
Merchant / maker
Captures or uploads a product photo
Web interface (apps/web)
React 18 + Vite SPA
Studio, catalog · TanStack Query · Tailwind
Orchestration API (apps/api)
Node 20 + Express + TypeScript
JWT auth, /listings, orchestration
withFallback()
Timeout + quota + fixture fallback
Integrations (server-side)
Anthropic Vision
Anthropic Copywriting
fal.ai (visual)
remove.bg (background removal)
DeepL (translation)
Persistence
Railway Postgres + Prisma
User · Listing (catalog) models
What the prototype demonstrates
Indicators from the prototype presented at the Vibe Coding Arena 2026 hackathon
| Indicator | Manual creation | With Déclic | Gain |
|---|---|---|---|
| Time to create a listing | ~20 minutes (manual) | ~10 seconds | −99 % |
| SEO quality | Variable, often neglected | Optimized title, description and keywords | Standardized |
| Product visual | Raw photo, careless background | Staged visual + background removal | Professional |
| Multichannel variants | Manual copy-paste per channel | Shopify CSV / Instagram export | Automated |
| Internationalization | No time for multilingual content | One-click DeepL translation | Unlocked |
Qualitative gains
A complete chain, not a wrapper
Déclic covers the whole chain from photo → publishable listing → variants → translation, where most tools stop at a text generator.
Key security
All integration keys (Anthropic, fal.ai, remove.bg, DeepL) stay server-side; the client only talks to the API.
Graceful degradation
Each integration is wrapped by withFallback(): a slow or rate-limited service switches to a deterministic fixture without breaking the flow.
Consistent brand image
Listings in a consistent format, staged visuals and removed backgrounds: a catalog that inspires trust, even for an independent merchant.
Mobile-first
The Studio is designed capture-first on mobile: take the photo and get the listing right after, with an anchored action bar.
Scalable catalog
An extensible data model (User, Listing) ready for batch processing, exports, translations and SEO scoring without a rebuild.
How about we start by talking?
A similar project in mind? Let's take 30 minutes to understand your situation and see how AI can speed up your e-commerce. First conversation free and with no commitment.
Discuss a similar project