Skip to content

Home en

margrop edited this page Oct 31, 2026 · 4 revisions

Smart Wrong Notebook Wiki

Smart Wrong Notebook turns mistakes captured by photo or typed as text into a searchable, practicable and printable question bank, and tracks mastery from the student's actual answer history.

This Wiki covers margrop/wrong-notebook. It is the English edition of the Chinese Wiki, checked against release v2026.10.31 (7ebab6b) and last updated on 2026-10-31. The project is forked from wttwins/wrong-notebook and released under the MIT License.

Question cards, knowledge points and filters in a notebook

Start here

What you want to do Page
Run it locally and sign in for the first time Quick Start
Deploy with Docker or installers (desktop, fnOS, Android, npm), enable HTTPS or upgrade Deployment and Upgrades
Configure MiniMax, GLM, DeepSeek, Kimi, Qwen, StepFun, Volcengine, Anthropic, Azure, OpenAI or Gemini, plus cloud voices AI Configuration
Add, organize, review and search questions User Guide
Similar-question practice, exam papers, printing and recording results Practice and Exams
Preferences, the app time zone and user permissions Settings and Users
Export personal data or back up the whole site Backup and Restore
Batch-import questions from an external Openclaw agent Openclaw Integration
Code structure, tests and contribution rules Development
Sign in after a fresh install; troubleshoot recognition, formulas or screenshots FAQ

Key features

  • Image upload, clipboard paste, text input, cropping and multi-question selection; optionally review before saving.
  • AI recognizes the question, answer, analysis, mistake cause, subject, knowledge points and a 1–5 star difficulty.
  • Duplicate detection by question similarity; a repeated import records another wrong attempt on the original question.
  • Markdown, LaTeX, SVG figures, SVG animated explanations with voice narration and optional GeoGebra interactive demos.
  • Similar-question practice with AI grading, daily review, exam papers built from mistakes, answer history, mastery probability, statistics and print export.
  • Per-user data isolation, daily AI limits, admin settings, personal preferences and Openclaw batch import.
  • Fresh installs set up the admin with a one-time setup link (the install wizard on fnOS) instead of a fixed default password, then the setup wizard helps configure AI and create a notebook.

Recommended workflow

  1. Deploy the app and create the admin account: the desktop and standalone Android apps open the setup page automatically, fnOS asks in the install wizard, and Docker and npm use the one-time setup link from the startup log.
  2. Configure and test the AI; image recognition needs a model that accepts image input.
  3. Create a notebook with its school stage and subject, then add and review questions.
  4. Review with similar questions and exam papers, and record the results.
  5. Regularly export personal data and back up the site.

Project links

Source code · Releases · Issues · Project page

Clone this wiki locally