Stuck Doctor

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend What I Built I built Stuck Doctor for my friend Piku, who hesitates a lot before going to the doctor whenever she faces a health issue. Watching a friend put it off made me realize that waiting isn't laziness. It's usually fear, worry about money, awkwardness, not knowing where to start, or not knowing what to say. So instead of trying to replace a doctor, Stuck Doctor figures out why someone is stuck and gives them one tiny first step. What it does: - Speak or type in English, Hindi or Bengali, with read-aloud answers - Works out why you're avoiding care and gives you a 2-minute first move - Lets you rehearse booking a call with Meera, an AI clinic receptionist - Family and friend mode: gentle messages to encourage someone you love to go - Report decoder: checks values against the ranges printed on your report - Medicine schedule and reminders - After-visit notes: a clean summary and to-do list from what the doctor said - A "show the doctor" card for people too shy to say it out loud - Free-care finder (eSanjeevani, government hospitals, Jan Aushadhi, Ayushman Bharat) - An always-available SOS button (108, 112, 104, live location on WhatsApp) It deliberately does not diagnose, suggest medicines or doses, or read scans. It's a nudge, not a doctor. Demo Live app: https://kk-dev0.github.io/Stuck_Doctor/ (The AI backend may take a short while to wake up on the first visit.) Code https://github.com/KK-Dev0/Stuck_Doctor MIT licensed. How I Built It The AI engine is Gemma, Google's open-weight model family, run with Ollama. The backend uses Gradio and is hosted on Render, with a static front end on GitHub Pages. I built it in one day at Hacktoberfest 2026. Gemma handles the language-heavy parts: understanding why someone is avoiding care, writing the first-step suggestion, playing Meera in the call rehearsal, writing messages for friends and family, explaining report values, and summarizing the doctor's instructions. It also lets the app respond in English, Hindi and Bengali. I added safety limits on purpose: the app never diagnoses or recommends medicines, summaries use only what the user wrote, report values are compared with the ranges printed on the report, and there's an optional "double-check" mode that reads a photo a second time. Emergency help is always one tap away. AI assistance was used in building this project, which is also disclosed on the site. Why Does Open Innovation Matter? A health tool for people who are already worried about money and access shouldn't depend on a paid API that could change its price or rules. An open-weight model made it possible to build something this wide in a single day, support Indian languages, and keep the code open under the MIT license so others can fork it, translate it or improve its safety. It fits the theme perfectly: AI belongs to everyone, including the friend who has been too nervous, too busy or too broke to see a doctor. Prize Categories - Gemma - Ollama Top comments (0)