Offline coding practice, from LeetCode-style problems to real engineering projects.
中文 · Español · Website · Discussions · Issues
Solve LeetCode-style algorithm problems in JavaScript, TypeScript or Python, then build multi-stage engineering projects that are graded on concurrency, latency, resilience and code quality. Everything runs on your own machine; AI assistance is optional and uses the provider you configure.
No runtime dependencies are required. Download the latest release:
| Platform | Artifact |
|---|---|
| macOS (Apple Silicon) | AlgoLocal-*-macOS-arm64.dmg |
| macOS (Intel) | AlgoLocal-*-macOS-x64.dmg |
| Windows (Installer) | AlgoLocal-*-Windows-Setup.exe |
| Windows (Portable) | AlgoLocal-*-Windows-Portable.exe |
| Linux (AppImage) | AlgoLocal-*-Linux.AppImage |
| Linux (Debian, Ubuntu) | AlgoLocal-*-Linux.deb |
| Linux (Fedora, RHEL) | AlgoLocal-*-Linux.rpm |
These builds are signed ad-hoc rather than with an Apple Developer ID, so macOS cannot verify the developer and will refuse the first launch. To allow it:
- Move AlgoLocal to Applications and double-click it. macOS says it cannot verify the app is free of malware — click Done.
- Open System Settings → Privacy & Security, scroll to Security, and click Open Anyway next to AlgoLocal. Confirm with Touch ID or your password.
This is only needed once. On macOS Sequoia (15) and later the old Control-click → Open shortcut
no longer works, and xattr -rd com.apple.quarantine is neither needed nor sufficient.
Removing this warning entirely requires a paid Apple Developer ID certificate and notarization. Building from source avoids it.
Requires Node.js 18 or later and npm 8 or later.
git clone https://github.com/zxypro1/algolocal.git
cd algolocal
npm install
npm run build
npm startThe application is then available at http://localhost:3000. start-local.bat (Windows) and start-local.sh (macOS, Linux) perform the same steps and offer to write an AI configuration file on first run.
- Quick start
- Features
- Practice modes
- Language support
- AI features
- Usage
- Development
- Project structure
- Documentation
| Capability | Detail |
|---|---|
| Two practice modes | Algorithm problems and multi-stage engineering projects |
| Offline execution | Code runs in the browser through WebAssembly, with no server-side execution |
| Local data | Problems, drafts and statistics stay in local storage |
| Optional AI | Problem generation, hints, solutions and engineering review, via your own provider |
| Editor | Monaco, with per-problem and per-language draft persistence |
| Dashboard | Accuracy by difficulty and tag, activity heatmap, recent attempts |
| Platforms | Windows, macOS, Linux |
Standard interview-style practice: read the statement, implement the function, run the test cases. 29 problems are included and the library is extensible.
A project is a small system built across several stages in a multi-file workspace. Correctness is necessary but not sufficient; each stage also measures behaviour.
| Component | Description |
|---|---|
| Workspace | File tree, tabbed editor, read-only contract files, files unlocked per stage |
| Acceptance specs | Hidden test cases, executed in a Web Worker |
| Engineering gates | Assertions on measured metrics, for example peak concurrency at most 4, or 12 requests within 300ms |
| Virtual clock | sleep(200) costs no real time while latency and concurrency remain exactly measurable |
| Scoring | Correctness, concurrency, latency, resilience, encapsulation, elegance |
| AI review | Reads the code, the latest run and the static metrics, then reviews it as a production pull request |
Three projects are included:
| Project | Topics |
|---|---|
| Resilient fetch pipeline | Bounded concurrency, backoff, single-flight caching, cancellation |
| Event-driven order pipeline | Event bus, onion middleware, idempotency, dead letter queue |
| Resilient API gateway | Token bucket, circuit breaker, timeout budgets |
Projects can also be generated by AI. A generated project is executed against its own reference solution before it is accepted. See the Engineering Practice Guide for the authoring format.
| Language | Execution |
|---|---|
| JavaScript | Native browser execution |
| TypeScript | Transpiled by the TypeScript compiler, then executed |
| Python | Pyodide (CPython compiled to WebAssembly) |
All four features share a single provider configuration.
| Feature | Description |
|---|---|
| Problem generator | Produces a complete problem from a natural-language description: statement, test cases, templates, reference solution |
| Solution generator | Produces several approaches with complexity analysis and trade-offs |
| Chat assistant | Answers questions about the code currently in the editor without revealing the full solution |
| Engineering review | Reviews a completed stage on concurrency safety, failure behaviour, module boundaries and style |
Supported providers: DeepSeek, OpenAI, Claude, Qwen, and any OpenAI-compatible endpoint, local or remote (Ollama, LM Studio, vLLM, LocalAI, llama.cpp, or a hosted gateway). See AI_PROVIDER_GUIDE.md.
Algorithm problems:
- Select a problem and a language.
- Implement the solution in the editor.
- Select "Submit & Run Tests" to execute the test cases in the browser.
- Review the results, execution time, and the dashboard statistics.
Engineering projects:
- Open a project and read the stage brief.
- Implement the stage in the workspace.
- Select "Run acceptance" to execute the hidden specs and evaluate the gates.
- Review the spec results, metrics, score and optional AI review. Clearing a stage unlocks the next one.
Configuration and problem management:
- AI providers are configured in Settings (application menu in the desktop app,
/settingsin the browser). Desktop configuration is stored in~/.offline-leet-practice/config.json. - Problems can be added through the "Add Problem" page, by importing JSON, or by editing
public/problems.json. See MODIFY-PROBLEMS-GUIDE.md.
| Command | Purpose |
|---|---|
npm run dev |
Development server |
npm run build |
Production build |
npm run test:engineering |
Engineering runtime and preset project tests |
npm run test:ai |
Provider, streaming and JSON extraction tests |
npm run test:editor |
Draft persistence tests |
npm run projects:build |
Compile projects/definitions into projects.json |
npm run projects:verify |
Execute every stage against its reference solution |
npm run dist:mac / dist:win / dist:linux / dist:all |
Desktop builds, see DESKTOP-APP-GUIDE.md |
Built with React 18, Next.js 13, TypeScript, Mantine v7, Monaco Editor and Electron.
algolocal/
├── pages/
│ ├── api/ # Problem, AI and project endpoints
│ ├── problems/[id].tsx # Problem detail, with AI chat and solutions
│ ├── projects/ # Engineering Practice: list, workspace, generator
│ ├── generator.tsx # AI problem generator
│ ├── stats.tsx # Practice dashboard
│ ├── manage.tsx # Problem management
│ └── index.tsx # Problem list
├── src/
│ ├── components/ # React components
│ ├── hooks/ # WASM executor, project runner, AI configuration
│ ├── lib/engineering/ # Virtual clock, lab, module runtime, spec runner, scoring
│ ├── lib/server/ # AI provider, project store, prompts
│ └── workers/ # Stage runner worker
├── projects/definitions/ # Engineering project sources
├── public/
│ ├── problems.json # Problem database
│ └── projects.json # Engineering project database
├── electron-main.js
└── electron-builder.config.js
| Document | Contents |
|---|---|
| ENGINEERING-PRACTICE-GUIDE.md | How the engineering runtime works and how to author projects |
| AI_PROVIDER_GUIDE.md | Provider configuration, models, troubleshooting |
| MODIFY-PROBLEMS-GUIDE.md | Problem format and offline editing |
| DESKTOP-APP-GUIDE.md | Desktop build and packaging |
Contributions are welcome. Additional algorithm problems and engineering projects are the most useful additions; improvements to the analytics, interface and documentation are equally welcome.
MIT