AtCoder GUI

A browser automation tool for AtCoder with a CLI interface designed to streamline competitive programming workflows.

Features

• Browser Automation: Launch a browser with UI using Playwright
• Problem Interaction: Seamlessly navigate to problem directories and interact with them
• Code Generation: The gen command extracts information from the current page to generate problem templates, support test case fetching, and manage contest structures.
• Workflow Automation: Build, test, and submit solutions directly from your terminal.
• Configurable Commands: Define allowed system commands in config.json5 for direct execution.
• Session Management: Persistent browser sessions with cookie exporting capabilities.

Usage

Interactive CLI

Start the interactive command prompt:

pnpm run start

Key Commands

• gen <contest-id>: Generate a problem directory and template from the current page or a given contest ID.
• submit <filename>: Submit your solution to the current problem.
• test: Run test cases for the current problem.
• build: Build the source code.
• open <URL>: Open a specific URL in the browser.
• cd <directory>: Change the current directory and automatically detect problem metadata.
• config: Display the current configuration.

Customizing Commands

You can define custom shell commands that are allowed to be executed directly from the CLI by adding them to the allowedCommands array in config.json5:

allowedCommands: ["cp", "copy", "del", "dir", "ls", "make", "pwd", "rm", "code"],

Development

Testing

Run the test suite with:

pnpm test

Parser regression tests

Input-format prediction is covered by scenario tests: every JSON file under test-resources/expected-results/ is automatically picked up by src/tests/scenario.test.ts, which fetches the problem HTML and compares the parsed result against the JSON.

Principle: whenever you add handling for a problem whose input format previously failed to parse (or was parsed incorrectly), add a corresponding test-resources/expected-results/<taskId>.json capturing the expected parse result. This locks in the fix and guards against future regressions.

You can generate the expected JSON by running generateParseResult on the problem and dropping the formatTree, samples, judgeType, yesStr, noStr, mod, and errorTolerance fields (see src/tests/scenario.test.ts for the exact shape that is compared).

Technology Stack

• Node.js
• TypeScript
• Playwright
• Vitest
• ESLint