LearnShift AI

Responsible AI Disclosure

Models and APIs

The backend can use Google Gemini for embeddings and generation when `GOOGLE_API_KEY` is configured. Without a key, it uses deterministic grounded fallback responses from the local OOP chunks.

Frameworks and Data

FastAPI, SQLAlchemy, Alembic, MySQL, Next.js, TypeScript, and Tailwind CSS. Data comes from the provided OOP academic CSV chunks, misconception bank, hint ladders, and clearly labeled demo assessment records.

Limits and Privacy

Only fictional or anonymized learners should be used. This is not AGI, not a clinically validated cognitive profiler, and not a replacement for teacher judgment.

Human Review

AI-generated content and questions should be reviewed by a human educator before classroom use. The educator dashboard exposes evidence so teachers can verify why an intervention was recommended.