ProductKit

An agentic platform to automatically generate product listings for user-facing pages, Shopify integrations, and FAQ/documentation.

Description

ProductKit is an intuitive, all-in-one toolkit for product designers, small businesses, and others to easily set up and maintain their storefront.

The configuration of a typical storefront (Shopify, in this case) could take up to multiple weeks. We automate the following:

• Product documentation generation
• Product image showcase in multiple scenes, with a consistent background and 4K quality
• Product video showcase
• Product interactive 3D model showcase
• Auto sync/propagation to your Shopify storefront

We do this with the following pipeline:

PDF Manual Parsing: NVIDIA Nemotron NIM parses uploaded PDF manuals to extract technical specifications and product details, enriching the AI-generated marketing copy with accurate information from documentation.

Image Gen: Image name, description, and content are fed into Anthropic's Claude Sonnet 4.5 to get a textual representation of the image and generate N number of dynamically generated prompts for the new showcase images. Showcase images are sent (prompt + img) to FLUX.2.

Video: Veo2 on fal.ai is used to generate a showcase video based on the original images submitted along with the product.

3D Model: omnipart on fal.ai is used to stitch together and generate a 3D model (.glb) and visualize it on the website using ThreeJS.

Storage & Deployment: MongoDB is used as the backing store, DigitalOcean is used for deployment and blob storage, Vercel's NextJS is used for the frontend, and Kotlin is used for the backend.

Tech Stack

BFL (FLUX.2), fal.ai (media model APIs), Anthropic (Claude Sonnet 4.5 via API for text generation), NVIDIA (codegen and documentation via NIM), Vercel (Next.js frontend), DigitalOcean (deployment & S3/Spaces object storage)

Getting Started

Public Instance

Visit our public instance: ProductKit

Local Development

1. Install dependencies in frontend:

   bun install

2. Configure your frontend API server with a .env file in the frontend directory:

   NEXT_PUBLIC_API_URL=https://prod-api-pkit.liftgate.io
   NEXT_PUBLIC_WS_URL=wss://prod-api-pkit.liftgate.io

3. Ensure you have a MongoDB instance running locally (if running fully local)

4. Start the Ktor backend in backend with a .env file:

   # fal.ai
   FAL_KEY=my:key
   # DigitalOcean Spaces
   DO_SPACES_KEY=MYKEY
   DO_SPACES_SECRET=my/secret
   DO_SPACES_BUCKET=productkit
   DO_SPACES_ENDPOINT=https://sfo3.digitaloceanspaces.com
   # Anthropic API
   ANTHROPIC_API_KEY=sk-ant-api03-my-key