MythMD

Hackathon project developed by Bezhan, Ashley, Briand, and Abdullah (BABA)

MythMD is an AI-powered health claim fact-checker that verifies claims against real peer-reviewed research. Enter any health claim and get a verdict backed by PubMed studies, a plain-language summary, a research support score, and text-to-speech readout via ElevenLabs.

---

Stack

• Frontend: React + Vite
• Backend: FastAPI
• AI: Anthropic Claude (claude-sonnet-4-6)
• Research: PubMed API (NCBI)
• Text-to-Speech: ElevenLabs

---

Setup

Prerequisites

• Python 3.10+
• Node.js 18+

1. Clone the repo

git clone https://github.com/jabrilo/hack-brooklyn-baba.git
cd hack-brooklyn-baba

2. Backend

cd backend
python -m venv venv
venv\Scripts\activate        # Windows
# source venv/bin/activate   # Mac/Linux
pip install -r requirements.txt

Create a .env file in the backend directory:

ANTHROPIC_API_KEY=your_anthropic_api_key
NCBI_API_KEY=your_ncbi_api_key
ELEVENLABS_API_KEY=your_elevenlabs_api_key
[email protected]

Start the backend:

uvicorn app.main:app --reload

3. Frontend

cd frontend
npm install

Create a .env file in the frontend directory:

VITE_BASE_API_URL=http://localhost:8000

Start the frontend:

npm run dev

4. Open the app

Go to http://localhost:5173

---

How it works

1. Enter a health claim (e.g. "Sugar causes hyperactivity in children")
2. Claude extracts keywords and searches PubMed for relevant studies
3. Abstracts are fetched and analyzed by Claude
4. You get a verdict (Supported / Unsupported / Uncertain), a research support score, a plain-language summary, and cited sources
5. Hit the speaker button to hear the summary read aloud