OnlyCode

Local web app for practicing medium-difficulty LeetCode-style Python problems.

Stack

• Backend: FastAPI
• Database: SQLite
• Execution: subprocess + resource limits (resource.setrlimit)
• Frontend: single-page HTML + JS + Monaco Editor (CDN)

Setup with uv

1. Create/update the environment and install dependencies:

   uv sync

2. Start backend API:

   cd backend uv run uvicorn main:app --reload --port 8000

3. Open frontend in browser:

   Open frontend/index.html directly, or run a static file server and open it.

GPU exercises

Problems whose titles begin with [GPU] require an NVIDIA CUDA GPU and fail explicitly when CUDA is unavailable. On the GPU host, install their optional PyTorch and Hugging Face dependencies before starting the backend:

uv sync --extra gpu

The GPU problems use a longer per-submission timeout and disable the generic virtual-memory limit because CUDA reserves large virtual address ranges.

API

• GET /problems

• GET /problems/{id}

• POST /submit with JSON body:

  { "problem_id": "two_sum_variant", "code": "def solve(input_data):\n return []" }