Paste a ToS, get its DODI score. Live at dodi-web.onrender.com. The Digital Ownership Deception Index (0–100) measures how hard a document works to hide that "Buy now" means "revocable licence". Higher is more deceptive.
The free tier sleeps when idle, so the first request after a quiet spell takes about 30 seconds to wake the service.
This is the deployed companion to dodi-analysis, the study that scored ten platforms across a decade of ToS snapshots. Read the write-up for the findings and the validation against ToS;DR. The scorer here is the same math as DODI v1.1: deterministic, no LLM, no API calls, nothing stored. It counts whole words and their genuine forms (v1.1, September 2026).
DODI = 0.25 x readability penalty + 0.50 x licence ratio + 0.25 x red-flag score
pip install -r requirements.txt
uvicorn app.main:app --reloadOpen http://localhost:8000. Interactive API docs at http://localhost:8000/docs.
docker build -t dodi-web .
docker run -p 8000:8000 dodi-webThe container respects PORT, so it deploys unchanged to Render, Fly.io or Hugging Face Spaces.
One endpoint does the work:
curl -X POST http://localhost:8000/api/score \
-H "Content-Type: application/json" \
-d '{"text": "You are granted a limited, revocable license to access the service..."}'Response: the total score, the three weighted components (licence ratio, readability, red flags) and the raw counts behind them. GET /api/health for monitoring.
pytestCovers the scorer (determinism, weighting, bounds, the zero-ownership edge case and the counting cases that broke v1.0, such as "download" and "different") and the API (validation, error paths). CI runs the suite on every push via GitHub Actions.
Same as the study: keyword counting has no sense of negation, the score is gameable by padding a document with ownership words and it measures language, not legal substance. Treat the number as a signal, not a verdict.
Akshay A · axwolf13.github.io