Listing Lab: Private, Local AI Copy for Independent Sellers
This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle . What I Built Listing Lab is a private marketplace copywriter for independent Etsy, eBay, and Facebook Marketplace sellers. A seller enters the real facts about a product, chooses a marketplace and tone, and gets a title, description, highlights, and search tags. The person I built it for is the small seller…
Listing Lab is a private, local AI copywriter tailored for independent sellers on platforms like Etsy, eBay, and Facebook Marketplace. Users input product details and select a marketplace and tone to receive a compelling title, description, highlights, and search tags. The tool is designed for small sellers seeking enhanced copy without the need for an additional subscription or uploading sensitive data to a third-party service.
The business model is straightforward: a one-time payment for a polished local tool or bundling it with a seller workflow, avoiding recurring API fees. The app is built as a responsive React experience, featuring a visible first-run model download, model caching, WebGPU acceleration, and fallback options for CPU and WASM. A plain-language privacy model ensures transparency.
The project is MIT-licensed and developed using React 19, TypeScript, Vite, Transformers.js, and the open-weight SmolLM2-360M-Instruct model. A public repository will be published before the final release. The entire inference process occurs within the browser, with the UI gathering product facts, buyer details, price, marketplace information, and tone.
The Transformers.js library lazily loads the quantized ONNX build of SmolLM2 only when the Generate button is pressed. Chrome utilizes WebGPU when available, falling back to WASM on the CPU, and model files are cached by the browser after the initial download.
The interface provides real-time feedback on bytes and percentage during the first model download, ensuring a smooth user experience. Listing Lab emphasizes honest costs, such as the substantial initial download and varying browser support, while avoiding hidden fees like per-generation API charges. The project highlights the importance of open weights, privacy, and the ability to swap, quantize, tune, or self-host the model without vendor constraints.
After the first model download, the core workflow can operate offline, ensuring the seller's prompt remains private.
The author reflects on the learning process, emphasizing the significance of model download and user experience design. The first user test will determine whether sellers prioritize privacy enough to accept the initial download, which marketplace is in demand, and whether they would opt for a one-time purchase over a subscription. References to similar projects include a React component for running Qwen2.5 in the browser via WebGPU and a collection of 21 AI tools built without server integration.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
