{% extends "base.html" %} {% block title %}Privacy - AnZar{% endblock %} {% block content %}
AnZar runs on this machine. There is no AnZar server, no account and nothing to log in to.
AnZar sends data to two places, both of them set by you: the Canvas instance of the active profile (courses, questions, grades) and the model endpoint of the active profile (answers and grading criteria). Submissions contain student names, so the endpoint is worth choosing with that in mind. A local model in LM Studio or Ollama keeps the answers on this machine, a university-hosted endpoint keeps them on the campus network, and a commercial API does neither.
The data folder chosen during setup (Documents\AnZar by default) holds the submissions, the criteria, the grading runs and the logs. The Canvas token and the API keys live in profiles.json in the config folder. Whoever holds that file holds those credentials, so the file and any archive of its folder are better not shared.
No analytics, no telemetry, no data collection of any kind. It makes the network calls needed to fetch from Canvas and to call the model, and otherwise it works offline.
Deleting the data folder removes the grading data with it. Grades already uploaded live in Canvas and are removed there.
AnZar works with student data. The institutional rules for processing it and for automated assessment apply regardless of the tool used, and the teacher stays responsible for the outcome. The About page says what to arrange before a first use.