Repolode
nichochar

nichochar/ari

A local-first AI tutor that builds curricula and grades your handwritten pen-and-paper work, using a camera over your desk.

Stars
104
Forks
5
Open issues
0
Last commit
yesterday

What is it?

Ari is a learning app built around pen-and-paper problem solving. You describe a subject to an AI tutor, which proposes an outline and turns it into a curriculum with exercises. You work on paper at a study desk, optionally with a document camera above the page, and the tutor can see what you are writing so you can ask questions mid-task or ask it to grade your work when finished.

The app runs on your own machine. Chats, progress, uploaded materials and submitted photos are stored locally under a `.local/` directory, and personalization files such as a tutor style guide and a learner profile live in a workspace folder in your home directory. Tutoring requests use your local Codex login with a ChatGPT subscription rather than an API key, and submitted work and relevant context are sent to OpenAI for tutoring.

Setup requires Node.js 24+, git, ffmpeg and a ChatGPT account. The built-in camera bridge is Mac-only, while browser cameras and photo uploads work without it. A backup command exports learning data and personalization files, excluding credentials.

Key features

  • 01Generates a curriculum with exercises from a plain-language description of what you want to learn.
  • 02Reads handwritten work from a document camera so the tutor can answer questions while you work.
  • 03Grades submitted work on request and lets you save reflections that guide later lessons.
  • 04Stores chats, progress and materials locally, with a backup command for learning data.
  • 05Uses a local Codex login tied to a ChatGPT subscription instead of an API key.
  • 06Lets you edit the tutor's teaching style and learner profile directly in Settings.

Use cases

  • Working through math or physics problems on paper with an AI checking your steps.
  • Self-studying a new subject with a structured, exercise-based curriculum.
  • Keeping learning history and personal notes on your own machine instead of a cloud service.

Quick start

  1. 1

    Clone the repository

    git clone https://github.com/nichochar/ari.git
  2. 2

    Enter the folder and install dependencies

    cd ari && npm ci
  3. 3

    Build the app

    npm run build
  4. 4

    Start Ari and open http://127.0.0.1:3001

    npm start

Things to know

  • The built-in camera bridge works only on Mac; browser cameras and photo uploads work elsewhere.
  • You need a ChatGPT account and Node.js 24+; models must be available on your account.
  • Submitted work and learning context are sent to OpenAI for tutoring.
  • It is designed as a single-user app meant to run on localhost.

Topics

FAQ

What is ari?
A local-first AI tutor that builds curricula and grades your handwritten pen-and-paper work, using a camera over your desk. Ari is a learning app built around pen-and-paper problem solving. You describe a subject to an AI tutor, which proposes an outline and turns it into a curriculum with exercises. You work on paper at a study desk, optionally with a document camera above the page, and the tutor can see what you are writing so you can ask questions mid-task or ask it to grade your work when finished.
Is ari free to use? What license does it have?
ari is open source under the MIT license. Check the license terms before using it in commercial products.
Is ari still maintained?
Actively maintained. The last commit to ari was on Oct 19, 2026.
What can I use ari for?
Working through math or physics problems on paper with an AI checking your steps. Self-studying a new subject with a structured, exercise-based curriculum. Keeping learning history and personal notes on your own machine instead of a cloud service.
How do I install ari?
Clone the repository: git clone https://github.com/nichochar/ari.git Enter the folder and install dependencies: cd ari && npm ci Build the app: npm run build Start Ari and open http://127.0.0.1:3001: npm start
Where can I get ari?
The source code and installation instructions are on GitHub: https://github.com/nichochar/ari

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects