Skip to main content

Algorepo (α)

PyPI Version Python Version License: MIT Contributions Welcome

Algorepo is a CLI utility that scaffolds, organizes, and locally tests your algorithmic solutions — giving you a structured personal repository you control entirely.

Algorepo Demo

Table of Contents

The Idea

Solving DSA problems consistently is one of the most reliable ways to prepare for technical interviews. The harder part is building a habit and having something to show for it.

Algorepo's premise is simple: your solutions should live in your Git repository, structured the way you want, runnable from your terminal. LeetCode and CodeWars have their own tooling — sync integrations, notes, browser extensions. Algorepo is for those who prefer a local-first workflow: write in your editor, run tests in your terminal, commit to your own repo, and let the history speak for itself.

A well-maintained solutions repository is also something a potential employer can actually read — real code, organized and annotated the way you think.

How it Works

Pass a problem URL and Algorepo fetches the statement, extracts the function signature, and opens a ready-to-edit file in your editor:

❯ algorepo [https://leetcode.com/problems/zigzag-conversion/description/](https://leetcode.com/problems/zigzag-conversion/description/)
6. Zigzag Conversion [Medium]
Python3
/Users/you/Solutions/LeetCode/6. Zigzag Conversion.py

The generated file looks like this:

from lc import *
# ================================================
# 6. Zigzag Conversion
# https://leetcode.com/problems/zigzag-conversion/
# ================================================

class Solution:
    def convert(self, s: str, numRows: int) -> str:
        pass

test("""
Example 1:
Input: s = "PAYPALISHIRING", numRows = 3
Output: "PAHNAPLSIIGYIR"
Example 2:
Input: s = "PAYPALISHIRING", numRows = 4
Output: "PINALSIGYAHRPI"
# [truncated for readability]
""")

Write your solution, then test it with one command:

❯ algorepo test "6. Zigzag Conversion.py"
PASSED Test 1: args "PAYPALISHIRING", 3 result "PAHNAPLSIIGYIR" expected "PAHNAPLSIIGYIR"
PASSED Test 2: args "PAYPALISHIRING", 4 result "PINALSIGYAHRPI" expected "PINALSIGYAHRPI"

No compiler setup, no boilerplate main(), no copy-pasting.

Features

  • Multi-platform — LeetCode and CodeWars.
  • 10 language runners — C, C++, C#, Go, Java, JavaScript, Kotlin, Python, Rust, TypeScript. (more to come..)
  • Smart scaffolding — function signature, type hints, and full problem statement extracted automatically.
  • Flexible language selection — use your configured default priority list or override per problem with --lang.
  • Editor integration — opens the new file in any editor (vim, nvim, nano, code, notepad…).
  • Local test runner — isolated execution, data passed via JSON marshaling.
  • Share-ready — consistent structure, ready to be shared after each session..

Prerequisites

  • Python: 3.10+
  • Supported environments: Linux, macOS, Windows (WSL recommended).

Installation

Recommended (Global / CLI-wide):

# Via uv (fastest)
uv tool install algorepo

# Via pipx (industry standard for CLIs)
pipx install algorepo

Development / Local:

# Clone and install dependencies
git clone https://github.com/elsmirroad/algorepo.git
cd algorepo
uv sync

Quick Start

1. Configure

On the first run, Algorepo creates a config.yaml. Edit it to set your solutions directory, preferred language, and editor:

general:
  # The directory where your solutions will be saved
  solutions_dir: ~/Solutions

  # Custom templates directory (Optional)
  # If not set, uses ~/.config/algorepo/templates
  # templates_dir: ~/.config/algorepo/templates

  # Priority list: if Python3 isn't available for a problem, it falls back to C++
  language_priority:
    - Python3
    - C++
  
  editor: vim         # nano | vim | nvim | notepad | code | ...
  open_editor: true

  # LeetCode Authentication (RECOMMENDED)
  # Without these tokens, Cloudflare protection may block requests (403 Forbidden).
  # Required to bypass 403 on Premium problems and some public ones.
  # How to get them: Browser DevTools -> Application -> Cookies
  leetcode_session: ""
  leetcode_csrf_token: ""

2. Download a problem

# Shortest form — uses your default language priority
algorepo https://leetcode.com/problems/two-sum/

# Explicit subcommand
algorepo download https://leetcode.com/problems/two-sum/

# Override language for this specific problem
algorepo download https://leetcode.com/problems/two-sum/ --lang Java

3. Solve and test

Algorepo automatically finds the file in your solutions_dir by its name and extension.

algorepo test "1. Two Sum.py"

# Test a problem from a specific platform
algorepo test "Two Sum" --platform leetcode

4. Review your solutions

❯ algorepo list
SOLUTIONS:

LeetCode
    1. Two Sum.py
    6. Zigzag Conversion.py

# Filter by platform
algorepo list --platform codewars

A Note on Test Runners

⚠️ The runner system was designed with AI assistance and covers the common cases well. Complex or unusual data structures — certain tree variants, custom graph formats — may need manual test case adjustments.

If you find a bug or want to make a runner more idiomatic for your language, contributions are very welcome. See Contributing.

Roadmap

Roughly in priority order:

  • HackerRank, CodeRun, Codeforces support — third major platform.
  • Interactive CLI — search and select problems in the terminal via questionary.
  • Stats engine — generate a STATS.md with activity graphs and progress over time.

Contributing

If you're experienced in one of the supported languages and see a way to improve its runner — that's the highest-value contribution right now. Issues and PRs are open.

Please check CONTRIBUTING.md for local setup instructions and architectural guidelines.

License

MIT © 2026.

Metadata

Release files for algorepo 0.1.3

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for algorepo 0.1.3
File Size Uploaded
algorepo-0.1.3.tar.gz 102.2 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for algorepo 0.1.3
File Interpreter ABI Platform
algorepo-0.1.3-py3-none-any.whl Python 3 none any Details

Total release size: 160.9 kB

Release files / algorepo-0.1.3.tar.gz

Download URL algorepo-0.1.3.tar.gz
Size 102.2 kB
Tags Source
SHA-256 checksum
How to use checksums
49eed00cd73a49457a3bef59f938bc11d29902b515f4ed08d07c7d8c2677cdee
BLAKE2b-256 checksum
How to use checksums
d8718041acd04a08aa1920716807860d86763c5af7407835e8fd2b89d34f7ef3
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.13.0 {"installer":{"name":"uv","version":"0.13.0","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Arch Linux","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

Release files / algorepo-0.1.3-py3-none-any.whl

Download URL algorepo-0.1.3-py3-none-any.whl
Size 58.7 kB
Tags Python 3
SHA-256 checksum
How to use checksums
8d0cdf743a6d04ce137be6bfe6b15df615b8972583ed927dcda34fe0dc7d47c2
BLAKE2b-256 checksum
How to use checksums
fdaaddfd31251d15dc50f1b6f16d5d7f94065fc5b8012e451f53690dfa4230cc
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.13.0 {"installer":{"name":"uv","version":"0.13.0","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Arch Linux","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

Release history Release notifications | RSS feed

This release

0.1.3 This release

2 release files

0.1.2

2 release files

0.1.1

2 release files

0.1.0

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page