FlipCheck, a new open-source tool submitted for the Hacktoberfest Open-Source AI Challenge, aims to solve the tedious sourcing problem for electronics repairers and resellers. Developed by Quang Vinh Lee, the application aggregates listings from eBay Australia and Facebook Marketplace into a single workspace, using Gemma 4 running locally via Ollama to extract and validate product specifications. This approach eliminates the need for hosted model API costs per listing, a significant factor for users scanning hundreds of items weekly.

Technical Architecture and Workflow

The stack is pragmatic and builder-friendly: Next.js and TypeScript handle the frontend, while a Python backend manages marketplace discovery, validation, and local SQLite storage. Live data ingestion relies on SerpApi for eBay and Apify for Facebook Marketplace, with the backend normalizing listings and retaining seller photos. Gemma 4, specifically the gemma4:e2b variant, processes the text to extract facts, but crucially, every extracted detail must be backed by a supporting excerpt from the original listing. This validation step ensures that users can trust the automated assessments before contacting sellers.

Precision in Valuation and Validation

FlipCheck distinguishes itself by enforcing strict criteria for its "Worth a closer look" badge. This status requires at least five recent matching verified sales, confirmed product details, known shipping or pickup costs, and a margin meeting the user’s target based on lower-quartile resale estimates. The developer noted an early lesson where nearly everything was labeled "Needs review," prompting a redesign to make specific missing data pointsβ€”such as unknown condition or missing completed-sale evidenceβ€”visible to the user. This transparency helps repairers quickly identify why a listing lacks sufficient data for a confident purchase.

Real-World Feedback and Limitations

The tool is currently a working local MVP, with manual scan initiation and a focus on resale valuation rather than parts-only or faulty items. The first user, a friend of the developer who runs an electronics repair business, indicated willingness to pay for a more complete version but has not yet committed financially. The developer admits that resale accuracy and realized profit still require real-world validation, and the current version excludes a dedicated repair and parts workflow, which is planned as a future extension. The MIT-licensed code is available on GitHub, though live discovery still requires API keys for SerpApi and Apify.

Key Takeaways

  • Local inference with Gemma 4 via Ollama removes per-listing API costs for high-volume scanning.
  • Strict validation rules and visible "Needs review" reasons improve user trust in automated extraction.
  • The tool is currently an MVP focused on resale valuation, with repair-specific workflows still in planning.
  • Real-world profit validation and user retention metrics are the next critical hurdles for the project.

The Bottom Line

FlipCheck proves that local LLMs can handle complex validation logic without recurring API bills, but its true value for repair shops won't be realized until it expands beyond basic resale metrics into the nuanced world of parts harvesting and repairability assessment.