Ben Sehat

Journaling
Homepage
Init connection with G-Drive
Chat Interface
Medical Record from G-Drive
Login

Inspiration

Sometimes AI can be frustrating when we ask about our health — the answers feel too generic because the AI doesn't know our context. We wanted to simplify this: an AI that already knows your health records, understands symptom urgency, and can recommend nearby facilities. The name "Ben Healthy" comes from Javanese — "Ben" means "let it be" (like "biar" in Bahasa), so together it means "Let us be healthy" or "Biar kita sehat."

What it does

You type your symptoms — Ben triages them into emergency, moderate, or monitor — then shows the closest health facility and what kind of doctor you need. It also tracks your mood daily, syncs your medical documents, and gives weekly reflections. All in one place.

How we built it

We used Python, FastAPI, Vite, React, TypeScript, Tailwind CSS, and Supabase for the backend and database. Groq (Llama 3.3 70B) powers the AI triage, Leaflet + OpenStreetMap handles facility maps, and we deployed on Vercel (frontend) and Railway (backend). We collaborated through Git, with one person coding through Opencode (Claude) and others handling design and testing.

Challenges we ran into

Typos, merge conflicts, and Git disagreements slowed us down since not everyone was comfortable with version control. The Google Maps API required verification that took longer than the hackathon — so we switched to OpenStreetMap mid-project. Google Drive API also needed OAuth verification, so we built a mock sync that demonstrates the same pipeline without the approval bottleneck. When we asked AI to generate code, vague prompts sometimes broke things, and we had to debug without AI help.

Accomplishments that we're proud of

We got the app running without errors. We caught bugs ourselves without asking AI. We finished within the time limit despite the Git chaos and API blocks. And we shipped a real deployed app — not just a local demo.

What we learned

You can't fully rely on AI to build everything — it makes mistakes it doesn't even realize. You still need to understand the code, review the output, and know when to step in.