Security Practices
How we protect the product and its users.
API Key Authentication
All AI endpoints require a backend API key. The system raises an error on startup if the key is absent in non-development environments.
CORS Restriction
Cross-origin requests are restricted to explicitly listed origins. Wildcard origins are never permitted in staging or production environments.
Rate Limiting
AI endpoints enforce per-IP sliding-window rate limiting via Redis to prevent abuse. The limit is configurable per deployment environment.
No Identity Data at Core
The app does not require a student account, which minimises personal identity data stored or processed by the core practice flow.
AI Provider Sandboxing
Code submitted for AI review is treated as untrusted input. Users should not submit secrets or regulated data in code samples.
Infrastructure Security
The backend runs on Kubernetes with Gunicorn and Uvicorn workers. Redis caching reduces unnecessary AI provider calls and improves resilience.