Wrong Notebook智能错题本 GitHub 中文

A fresh start for every mistake

Turn every mistake
into a step forward.

Saving a mistake is only the first step.
AI helps you see why it went wrong and practice more like it,
so "I've done this" becomes "I've got this".

Get started Try the demo

Demo account demo@localhost / demo123456 · Shared data, please don't enter personal information

Open sourceInstallers for every platformPhone and computer
AI connects the dots
Your growing knowledge
Every practice leaves a traceAnswer → assess mastery → keep improving

Real app screenshot (English UI) · Click to enlarge

Built for steady learning

Primary · Middle · High school Choose your AI provider Desktop · NAS · ServerPWA Opens like an app

A LITTLE MISTAKE. A BIG STEP FORWARD.

From "I got it wrong" to "I've got it".

Spend less time copying questions out by hand and more time actually understanding them.

  1. 01

    Capture the mistake

    Snap a photo, paste a screenshot or type the question. AI recognizes it and sorts out the answer, analysis and knowledge points.

    Less busywork
  2. 02

    See how it works

    Review the cause of the mistake and the solution, with SVG animations that make abstract steps easy to follow.

    More "aha" moments
  3. 03

    Try a new one

    Generate variant questions on the same knowledge point at the difficulty you choose, and check that you really understand.

    From memorizing to mastering
  4. 04

    Watch yourself improve

    Every attempt is kept and mastery is tracked, so you can turn the questions that need work into an exam.

    Focus your next review

THOUGHTFUL TOOLS, REAL UNDERSTANDING

More than collecting mistakes:
it cares whether you learned.

From the details of one question to your whole web of knowledge,
the tools you need for learning, in one place.

Solutions you can watch

Abstract steps,
made clear at a glance.

AI turns the solution into a step-by-step SVG animation,
in English or Chinese to match the interface.
Pause, seek and replay at your own pace.

Practice beyond one question

Solve one,
then solve them all.

Generate fill-in-the-blank or free-response questions on the same knowledge point, at the difficulty you choose, to strengthen old weak spots with new problems.

Original · Linear equations3x + 6 = 21This time, understand the properties of equality.
Same concept, new practice
Variant · A new angle5x − 8 = 27This time, solve it on your own.
Illustration · Sample questions

A place for every concept.

Knowledge points are organized by school stage, subject and chapter. AI tags questions automatically, and you can add your own tags.

Organize on screen, practice on paper.

Build exams by knowledge point and mastery. Include answers and analyses if you like, then print or save as PDF.

Let practice say whether you've got it.

Mastery is estimated from your actual answer history and combined with mistake and practice statistics to find what's next.

×First mistake✓Variant practice✓Review
Based on answer history to guide your review
Several questions per photo Configurable duplicate detection LaTeX formulas Separate data per user

A CLOSER LOOK

See it in action.

The real running app with real study records.
Pick a tab to see each step. The interface also comes in English.

My notebook

Filter by knowledge point, grade, mastery and A/B paper to quickly find the questions you need to review.

Real app screenshots · English UI

YOUR NOTEBOOK, YOUR SPACE

Keep your study data
in your own space.

Desktop apps for Windows, macOS and Linux work out of the box;
fnOS, npm and Docker run it on a NAS or server.
Bring the AI provider you know and build your own notebook.

  • You own your data

    The SQLite database stays on your machine, with backup import and export.

  • Choose your AI provider

    MiniMax by default, or Zhipu GLM, DeepSeek, Kimi,
    Anthropic, Azure OpenAI, OpenAI and Gemini,
    plus services with compatible APIs.

  • Open it anytime

    Install it as a PWA on your phone or tablet home screen;
    the standalone Android app needs no server and keeps data on the device.

Read the full setup guide
QUICK START
Prebuilt image for amd64 / arm64
# 1. Create data and config directories
mkdir -p wrong-notebook/data wrong-notebook/config
cd wrong-notebook

# 2. Pull and start; data stays in the mounted directories
docker run -d --name wrong-notebook \
  --restart unless-stopped -p 3000:3000 \
  -e NEXTAUTH_SECRET="$(openssl rand -hex 32)" \
  -v "$(pwd)/data:/app/data" \
  -v "$(pwd)/config:/app/config" \
  ghcr.io/margrop/wrong-notebook:latest
Then open http://localhost:3000

The default admin is admin@localhost / 123456. Change the password after your first sign-in and set up your AI provider in Settings. Images are published at ghcr.io/margrop/wrong-notebook; pin a version with tags such as 2026.10.22. See the setup guide for upgrades, Docker Compose and data directories of each installer.

GOOD QUESTIONS

Before you start,
you might wonder.

More questions? Join the discussion on GitHub.

Report an issue or suggestion
Does it cost anything?

The project is free and open source under the MIT License, with source code, a Docker image, an npm package and installers for every platform to host yourself. AI features such as recognition and explanations need your own AI provider, which charges for usage; official MiniMax, Zhipu GLM, DeepSeek and Kimi channels are built in, and services with OpenAI- or Anthropic-compatible APIs work too.

Can I try it first?

Yes. Open the live demo and sign in with demo@localhost and password demo123456. The demo account is shared and its data is visible to all visitors, so please don't enter personal information. For real use, host your own instance.

Which systems and devices are supported?

The server runs as a desktop app on Windows, macOS and Linux, on fnOS NAS, and on any server with Docker or Node.js. Once it's running, use it from a browser on your phone, tablet or computer. The desktop app accepts local connections only; fnOS, npm and Docker setups can be reached across your LAN.

Which grades and subjects does it cover?

Each notebook is created for primary, middle or high school and a subject. The built-in knowledge points follow the Chinese K-12 curriculum and are organized by school stage and subject; custom tags are supported, and AI helps identify knowledge points and difficulty.

Does it work in English?

Yes. The interface is available in Simplified Chinese and English, and AI can recognize and explain questions in either language. The built-in knowledge-point library and most documentation are in Chinese.

Can I use it on my phone?

Yes. It works in browsers on phones, tablets and computers, and supports PWA. Choose "Add to Home Screen" from the Share menu in Safari on iPhone / iPad, or from the Chrome menu on Android, to open it like an app. On Android you can also install an APK from Releases: the standalone app has the server built in and keeps data on the device, while the client connects to your own server. Both support taking photos for upload, saving exports and system printing.

Where is my data stored? Can it run fully offline?

Questions and answer records are stored in a SQLite database on your own deployment, and so are AI settings. AI recognition, solving and generation send the relevant content to the AI service you configured, which must be reachable; PWA support does not mean fully offline use.

Can GitHub Pages run the notebook itself?

This website is hosted on GitHub Pages. The notebook app includes sign-in, a database and AI APIs, so it needs a running server: install the desktop app or the fnOS package, or deploy with npm or Docker to a computer, NAS or server. To see it first, try the live demo.

How is it related to the upstream wrong-notebook?

This project extends wttwins/wrong-notebook with bilingual SVG animated explanations, mastery tracking, configurable duplicate detection, entering exam results, official Chinese model channels, an English interface, installers for every platform and more. Thanks to the upstream authors. See the project README for the full feature list and differences.

KEEP THE LESSON. LEAVE THE MISTAKE.

Next time you meet it, you'll know.

Start with one mistake and collect every "now I get it".

Get started on GitHub

App screenshot