Small-scale e-commerce sellers often spend 15-20 minutes crafting each product listing, a bottleneck that scales poorly for resellers and makers. A new open-source tool called Listing Writer addresses this by generating ready-to-paste titles, descriptions, and search tags for Etsy and eBay in seconds. Built for the Hacktoberfest 2026 Open-Source AI Challenge, the application leverages Google's Gemma 3 model running locally to eliminate per-request API costs and keep seller data private.

Local Execution Over Cloud APIs

The core architecture relies on Gemma 3 (4B), served via Ollama and exposed through a Gradio interface. This stack allows the entire workflow to run on a developer's laptop or a free Google Colab T4 GPU instance. By avoiding paid cloud APIs, the tool ensures that generating hundreds of drafts incurs no additional financial overhead, a critical factor for side hustles where margins are thin. The app accepts a product name, specific details, platform choice, and tone preference, optionally processing an image via Gemmaโ€™s vision capabilities.

Prompt Engineering Fixes Hallucinations

Initial testing revealed significant hallucination issues common with smaller models. The first version invented accessories, such as claiming a charging cable was included when it wasn't, and added fake user reviews. The developer rewrote the system prompt to enforce strict factual adherence, instructing the model to use only provided details or visible image data. This adjustment successfully eliminated fabricated specs, though the developer notes that human verification remains necessary for nuanced terms like waterproof versus splash and sweat resistant.

Key Takeaways

  • The tool generates Etsy-compliant 13 tags and titles under 140 characters.
  • Running locally ensures seller data, including prices and photos, never leaves the machine.
  • Future updates plan to include bulk CSV processing and niche fine-tuning for refurbished electronics.
  • The project demonstrates that 4B parameter models can handle structured business tasks with proper prompting.

The Bottom Line

Listing Writer proves that local LLMs are viable for specific, high-volume e-commerce tasks, offering a cost-effective alternative to cloud APIs. While prompt engineering is critical to prevent hallucinations, the privacy and zero-cost benefits make it a compelling tool for side hustlers.