Enter your address, see your federal representatives and their recent votes explained in plain English. Nonpartisan.

• Frontend: React + Vite (port 5173)
• Backend: FastAPI + httpx (port 8000)
• APIs: Google Civic Information, api.congress.gov, Anthropic (Claude)

You need three API keys:

cd backend
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Create backend/.env with:

GOOGLE_CIVIC_API_KEY=your_key_here
CONGRESS_API_KEY=your_key_here
ANTHROPIC_API_KEY=sk-ant-your_key_here

Run it:

uvicorn main:app --reload --port 8000

First run will download the public legislators dataset (~700 KB) from theunitedstates.io and cache it locally.

cd frontend
npm install
npm run dev

Open http://localhost:5173.

1. User enters an address. Frontend calls /api/reps.
2. Backend hits Google Civic API, gets federal reps, looks up their bioguide_id from the local legislator dataset.
3. User clicks a rep. Frontend calls /api/votes/{bioguide_id}.
4. Backend hits api.congress.gov: lists recent chamber votes, then in parallel fetches each vote's roster to find that member's position. Falls back to a hardcoded set of real recent votes if the live calls fail.
5. User clicks "Explain in plain language". Frontend calls /api/explain.
6. Backend asks Claude (Haiku 4.5) for a 2-3 sentence neutral explanation, with the system prompt cached for cost savings on subsequent calls.

Open http://localhost:5173, paste a real address (e.g. 1600 Pennsylvania Ave NW, Washington, DC or your hackathon venue's address), click a senator, expand a vote, and watch Claude generate a neutral explanation live.

• State and local reps (federal only)
• Caching of vote data (calls api.congress.gov fresh each time)
• Auth, accounts, saved reps
• Mobile polish beyond a basic responsive layout

• The api.congress.gov roll-call vote endpoints are relatively new and the response shapes vary slightly. The backend is defensive: any failure falls through to a curated set of real recent votes so the demo never hard-breaks.
• The system prompt for explanations is cached on Anthropic's side using cache_control: ephemeral. Repeated explanations during a demo are cheaper and faster.