I built The Repair Line: turning small repairs into shared relief

A developer known as Reet24-del built The Repair Line, a community-repair platform that lets residents upload photos of broken shared infrastructure such as hand pumps, roofs, and fans. The app uses Google Gemini to assess repair photos and provide structured estimates, with a React and Vite frontend deployed on Vercel and Snowflake as the data layer. The project aims to make small-scale generosity practical and transparent.

This is a submission for Weekend Challenge: Generosity Edition https://dev.to/challenges/weekend-2026-09-03 I built The Repair Line , a community-repair platform for the small shared things that keep a neighbourhood moving: a broken public hand pump, a leaking roof above a ration counter, a damaged study desk, or a clinic fan that has stopped working. The goal is to make generosity visible and practical. Someone uploads a clear photo of a repair need, receives an AI-assisted assessment, and can add it to a local repair line where people can see the impact, the directional cost, and the amount already pledged. Each repair includes: The experience is designed to feel warm and community-led rather than like a generic fundraising dashboard. It starts with local repairs, but the city selector supports Lucknow, Kanpur, Prayagraj, Varanasi, and an all-cities view. Live app: The Repair Line https://therepairline.vercel.app Try the flow: The project is built with React and Vite, with API routes prepared for Vercel Functions. Repository: github.com/Reet24-del/the-repair-line https://github.com/Reet24-del/the-repair-line Key pieces of the implementation: I wanted to focus on generosity that is easy to understand. Not every community need requires a huge campaign. A ₹2,000–₹10,000 repair can make a real difference when it restores a shared tap, desk, roof, or fan. The product flow became: photo in, fixed out . The repair assessor accepts a photo and sends it to a server-side Gemini route. Gemini is asked to return structured JSON with: { "severity": "Low | Medium | High", "cost": "a directional INR range", "plausibility": "Likely genuine | Needs review | Unclear", "reasoning": "a concise explanation" } I require a photo before an assessment can run. This prevents the interface from creating a confident-looking repair result without visual evidence. The cost is intentionally presented as directional guidance rather than a quote. Repair records include the repair title, location, severity, estimated cost, people helped, reason, image, and pledged amount. The Snowflake schema and query layer support feed filters for city, severity, and sorting. The feed initially keeps the page focused by showing a smaller set of repairs. A Show all repairs control expands the complete list, while the city selector changes the listings and heading to keep the experience locally relevant. The frontend is deployed on Vercel, while Vercel Functions host the assessment, repair, and pledge endpoints. Sensitive credentials stay in server-side environment variables rather than the browser bundle. The project uses Google Gemini to analyse an uploaded repair photo and produce a structured, user-facing assessment. Gemini helps translate a visual problem into understandable severity, plausibility, repair guidance, and an INR cost range. Snowflake provides the repair data layer for listings and pledges. The app is designed to query repairs by city, severity, and priority, so a growing local repair line remains filterable and useful. The Repair Line is a small idea on purpose: make a modest, shared repair easy to see, trust, and complete.