This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle
What I Built
I built Open Shelf, a local-first listing studio for independent sellers.
Add a few notes about an item—and optionally a product photo—and Open Shelf drafts a marketplace title, description, search tags, category, pricing thought, and a checklist of details to confirm. It’s designed for makers and small sellers who spend too much time turning product details into listings.
The goal is a practical tool sellers could use without sending their product photos or notes to a hosted AI service. [Add a sentence about who you think would pay for it and why.]
Code
https://github.com/MRIDUL-MB/hacktoberfest-2026-week-2.git
How I Built It
Open Shelf uses the open-weight Qwen2.5-VL 3B model through Ollama, an open-source local inference runtime. The model can work with product notes and an optional image to generate structured listing content.
The app is built with a small Node.js server and a plain HTML, CSS, and JavaScript frontend. The server sends generation requests to the Ollama instance running on the same computer. It doesn’t save submitted product information.
To try it locally, install Node.js and Ollama, download the model with ollama pull qwen2.5vl:3b, then run npm start.
Why Does Open Innovation Matter?
For a small seller, an AI API bill or a third-party upload can make a simple tool less appealing. With a local open-weight model, product photos and notes can stay on the seller’s computer, and there’s no per-listing API charge or dependency on one provider’s pricing.
After the model is downloaded, the app can generate listings without an internet connection. The model can also be changed, which leaves room to try alternatives or adapt the tool for a particular shop or marketplace.
Local inference still uses the computer’s resources, and the generated text needs a human review—but for this use case, local control and the option to run without a recurring AI fee are meaningful advantages.
Top comments (0)