This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle
ListingForge ⚡ — The Open-Weight AI E-Commerce Listing Studio
What I Built
ListingForge is a local-first, open-weight AI listing studio that turns raw product photos into high-converting, platform-optimized listings for Etsy, eBay, Shopify, and Poshmark in under 30 seconds.
Who Would Pay for It?
Full-time vintage resellers, boutique owners, thrift flippers, and handmade Etsy artisans.
The Pain Point & Problem It Solves
Taking photos of inventory takes 2 minutes. But writing the listings is an agonizing bottleneck that takes 15 to 30 minutes per item:
- Etsy Search Demands 13 Strict Tags: If you don't use all 13 tags, your visibility is slashed. If any tag exceeds 20 characters, Etsy rejects it.
- eBay Cassini Penalizes Wasted Space: eBay gives you exactly 80 characters for titles. Missing structured "Item Specifics" drops your ranking, while missing condition notes triggers costly "Item Not As Described" return disputes.
- Cloud AI is Too Expensive: Multi-photo vision requests to proprietary APIs (GPT-4o or Claude 3.5 Sonnet) cost \$0.15–\$0.25 per item. For a reseller listing 100 items a week, cloud API bills eat into narrow margins.
The Business Model
- The ROI: Cuts listing creation from 20 minutes to 30 seconds. For an active seller listing 25 items a week, it saves 7+ hours of admin work every week.
- Pricing: A \$19–\$29/month subscription or a \$49 lifetime desktop license (similar to iconic Mac productivity apps).
- Zero Marginal Cost: Because inference runs on open-weight models locally, the operating cost per listing is \$0.00.
Demo
- GitHub Repository: https://github.com/Srijanhub/ListingForge
- Instant 1-Click Presets: The studio includes 5 built-in, pre-configured items with vector SVG mockups (Vintage 90s Leather Biker Jacket, Hand-thrown Ceramic Pour-Over Dripper, Retro 1980s Casio Gold Watch, Mid-Century Teak Lamp, Y2K Levi's 501 Jeans). Evaluators don't even need photos on hand—just click a preset to see instant multi-platform generation!
# Clone and run locally in seconds
git clone https://github.com/Srijanhub/ListingForge.git
cd ListingForge
npm install
npm start
# Open http://localhost:3000 in your browser
Code
ListingForge ⚡
Open-Weight AI E-Commerce Listing Studio for Etsy, eBay, Shopify & Poshmark
Turn product photos into polished, platform-optimized, top-ranking listings in seconds using 100% open-weight vision & language models.
đź’ˇ Why Someone Would Actually Pay For This
Every day, millions of independent e-commerce sellers—vintage pickers, Etsy artisans, boutique owners, and full-time eBay resellers—face the same grueling bottleneck:
- The Problem: Photographing products is fast, but writing listings is excruciating. Crafting high-converting SEO titles, researching 13 Etsy search tags under 20 characters, filling out eBay's algorithmic Item Specifics, and writing dispute-proof condition notes takes 15 to 30 minutes per item.
- The Cost of Bad Listings: Missing tags mean zero Etsy search impressions. Vague condition disclosures cause painful "Item Not As Described" returns. Suboptimal titles get buried by the eBay Cassini algorithm.
- Why Cloud AI Fails Sellers: Cloud APIs (GPT-4o, Claude Sonnet) cost $0.15–$0.25 per multi-photo item. For a seller listing…
How I Built It
ListingForge is built around a flexible, multi-tiered open-source AI architecture:
-
Local Open-Weight Vision Models (Ollama):
- Direct connection to local Ollama running on
http://127.0.0.1:11434. - Automatically probes
GET /api/tagsto discover installed open-weight vision models: Llama 3.2 Vision (llama3.2-vision), Qwen2-VL (qwen2-vl), Moondream2 (moondream), and LLaVA (llava). - Takes multi-photo base64 arrays to inspect texture, fabric, brand labels, and natural wear/patina locally on Apple Silicon GPU/MPS.
- Direct connection to local Ollama running on
-
Hugging Face Open-Weight Router:
- Fallback to open-weight models (
Qwen/Qwen2-VL-7B-Instruct) hosted via the Hugging Face router for users without local GPU hardware.
- Fallback to open-weight models (
-
Built-in Domain Heuristic Engine:
- A comprehensive e-commerce rule and taxonomy engine that extracts vintage dating rubrics and category benchmarks with zero setup required.
-
Marketplace Compliance & Quality Scorer:
- Real-time SEO Auditor (0–100 score, Grade A–F): Live recalculation testing title keyword density, Etsy 13-tag counts, 20-character tag limits, and dispute-proof condition notes.
- Profit Margin & Fee Calculator: Deducts platform transaction fees (Etsy 6.5%, eBay 13.25%, Shopify 2.9%), Cost of Goods (COG), and shipping to display true net profit.
-
Bulk Exporters:
- RFC-4180 compliant CSV export engine ready for Etsy Bulk Upload, eBay File Exchange / MIP, and Shopify Products CSV.
-
Automated Testing:
- Comprehensive 63-point test suite (
node test.js) validating platform constraints, tag lengths, and CSV syntax.
- Comprehensive 63-point test suite (
Why Does Open Innovation Matter?
Open innovation and open-weight models are the only reason a product like ListingForge can exist as a sustainable side hustle:
- Zero Marginal Inference Cost: Closed API tokens turn a side business into an API billing liability. Open-weight models running on the user's laptop make inference completely free forever.
- Intellectual Property & Privacy: Artisans selling bespoke handmade pieces or vintage pickers discovering rare uncataloged items often refuse to send photos to cloud servers before public release. On-device models ensure client photos never leave their laptop.
- Offline Flea Market & Estate Sale Utility: Estate sale pickers and flea market resellers often operate in basements, warehouses, or rural locations with zero Wi-Fi. With local open-weight models, sellers can inventory and batch-generate listings on the go.
- Deterministic Rule Constraints: Open-source prompt engineering paired with local structured JSON outputs lets us strictly enforce platform rules (e.g., Etsy's unforgiving 20-character tag cutoff and eBay's 80-character title limit) without vendor lock-in.
My Agent Session
Built with the assistance of Google Antigravity, leveraging agentic pair programming to architect the multi-engine adapter pattern, design the responsive Studio UI, construct the 63-point test suite, and implement RFC-4180 CSV exporters.
My First User
I shared an early build with a friend who runs a vintage clothing resale side-hustle on Etsy and eBay. Every Sunday, she spent 4 to 5 hours measuring, researching tags, and drafting descriptions for 20 garments.
Using ListingForge with 4 photos per jacket/denim, her Sunday listing time dropped to under 35 minutes. Her immediate feedback: "If you can give me 13 tags that don't get rejected by Etsy's character limit and format the eBay specifics table automatically, I would happily pay \$20 a month just to get my Sunday afternoons back."
Prize Categories
- Side Hustle (Primary)
- Local Inference & Open-Weight AI
- Developer Productivity & Creator Tools
Top comments (0)