OncoScan

AI-powered OSCC (Oral Cancer) detection from biopsy slides — Next.js dashboard + FastAPI ML backend.

---

How it works

Browser → Next.js (frontend) → Your FastAPI server → Swin Transformer model
                             → MongoDB Atlas         (auth + records)

---

1 — Set up the ML API (your own server)

The inference backend is open — duplicate it from HuggingFace Spaces:

https://huggingface.co/spaces/immohitsen/oncofastapi

1. Click "Duplicate this Space" (top-right button)
2. Give it a name and click Duplicate Space — HuggingFace builds it automatically
3. Once running, your API URL will be:

   https://YOUR-USERNAME-YOUR-SPACE-NAME.hf.space
   

4. Verify it works:

   curl https://YOUR-USERNAME-YOUR-SPACE-NAME.hf.space/
   # {"status":"online","model_loaded":true,...}

You can also run it locally — clone the Space repo and run:

pip install -r requirements.txt
uvicorn main:app --port 8000

---

2 — Set up MongoDB

1. Create a free cluster at cloud.mongodb.com
2. Create a database user and copy the connection string
3. Whitelist your IP (or 0.0.0.0/0 for dev)

---

3 — Run the frontend

git clone https://github.com/your-username/oncoscan.git
cd oncoscan
npm install

Create .env in the oncoscan/ folder:

FASTAPI_URL="https://YOUR-USERNAME-YOUR-SPACE-NAME.hf.space"
MONGODB_URI="mongodb+srv://<user>:<pass>@<cluster>.mongodb.net/oncoscan"

Start:

npm run dev
# http://localhost:3000

---

Tech Stack

Frontend	Next.js 16, React 19, Tailwind CSS
Auth	JWT + bcryptjs, MongoDB Atlas
ML API	FastAPI, Swin Transformer (HuggingFace)

---

Troubleshooting

Problem	Fix
Model not loaded	Your Space is still building — wait ~2 min and refresh
bad auth on MongoDB	Check username/password and IP whitelist in Atlas
Port 3000 in use	Next.js auto-switches to 3001