AtCoder Problem Helper

AtCoder Problem Helper is a VS Code extension for fetching AtCoder problems, reading statements, translating problems with LLMs, generating AI explanations, running sample tests locally, and submitting code from the editor.

Features

• Fetch contest problems by contest ID or AtCoder contest URL.
• Read statements, constraints, input/output formats, and samples in the sidebar.
• Translate problem statements with LLMs and switch between original and translated content.
• Stream translation output directly in the Problems panel.
• Generate AI explanations with streaming output, chat bubbles, folded thinking blocks, and copy buttons for code blocks.
• Continue asking follow-up questions, regenerate answers, roll back the latest follow-up answer, and undo rollback.
• Use Thinking switches for translation and AI Tutor. Translation keeps thinking disabled by default for faster Ollama/Qwen-style reasoning models.
• Render problem statements and AI answers with KaTeX, including common bare LaTeX, block formulas, mixed delimiters, and Unicode math symbols.
• Save translations and AI explanations into problem history so they can be restored later.
• Configure multiple LLM providers, provider switches, model lists, Base URLs, API keys, and provider logos.
• Run sample tests locally and compare outputs.
• Submit solutions to AtCoder after login.

Supported LLM Providers

Built-in presets include:

• OpenAI
• Anthropic
• Google Gemini
• DeepSeek
• SiliconFlow
• Zhipu AI
• Kimi / Moonshot
• Alibaba Cloud Bailian / Qwen
• Ollama, using http://localhost:11434 by default and no API key
• OpenCode Go
• Custom OpenAI-compatible API

Only enabled providers that are configured and have at least one model appear in model dropdowns. Local providers such as Ollama do not require an API key.

Quick Start

1. Install the extension and open the AtCoder icon in the VS Code Activity Bar.
2. Open the Settings panel and configure at least one LLM provider and model.
3. In the Problems panel, enter a contest ID such as abc350, or paste a full AtCoder contest URL.
4. Select a problem, then translate it, generate an AI explanation, run samples, or submit your solution.

Common Configuration

{
  "atcoder.llm.provider": "deepseek",
  "atcoder.llm.deepseek.apiKey": "your-api-key",
  "atcoder.llm.deepseek.model": "deepseek-chat",
  "atcoder.llm.targetLanguage": "zh-CN",
  "atcoder.language": "cpp",
  "atcoder.cppCompiler": "g++",
  "atcoder.cppFlags": "-std=c++17 -O2 -Wall"
}

You can also configure providers, API keys, Base URLs, and model lists in the extension Settings panel.

Ollama

Start Ollama locally and pull a model first:

ollama serve
ollama pull qwen2.5-coder:7b

Then select Ollama in the Settings panel, keep the Base URL as http://localhost:11434, and fetch the local model list.

Usage

Fetch Problems

Enter a contest ID such as abc123 or arc180, or paste a full AtCoder contest URL, then click Fetch.

Translate a Problem

Open a problem and click Translate. Translation streams into the statement area and is saved to problem history. You can switch between original and translated content at any time.

AI Tutor

Click Generate to create a guided explanation. The AI Tutor supports streaming output, folded thinking blocks, code block copy buttons, follow-up questions, message copy, regeneration, rollback, and undo.

Run Samples

Open the solution file for the current problem and click Run Samples. The extension compiles and runs the configured language against all samples and shows the result for each case.

Submit Code

Log in to AtCoder, open your solution file, and click Submit. If AtCoder requires Cloudflare or Turnstile verification, the extension will prompt you to finish submission in the browser.

Commands

Development

npm install
npm run compile
npm run watch

Main directories:

src/
  extension.ts
  providers/
  services/
media/
out/

Packaging

npm install
npm run compile
npx vsce package

Install the generated .vsix with Extensions: Install from VSIX... in VS Code.

License

MIT