MRiabov/github-awesome-repo-descriptions

Awesome Final Repos Dataset

A dataset of “final” GitHub repositories and their descriptions discovered by traversing Awesome lists.
This dataset focuses on canonical GitHub repositories referenced by Awesome lists at a specified depth, with optional metadata enrichment like stars and README previews.

---

Dataset Overview

• Data file: awesome-repos.parquet
• Hugging Face mirror: MRiabov/awesome-repo-descriptions
• Core columns: name, link, description, source_repo
• Optional columns: stars, readme_preview

Each row corresponds to a repository that does not contain "awesome" in its name and is collected at the deepest traversal level of an Awesome list.

---

How the Dataset Was Built

1. Start from a root Awesome list (e.g., sindresorhus/awesome).
2. Fetch README markdown via the GitHub API and parse bullet entries like:

[Name](https://github.com/owner/repo) - Short description

3. Recursively traverse linked Awesome lists up to max_depth.
4. At the deepest level (depth == max_depth), collect only non-"awesome" repositories.
5. Optionally enrich with GitHub metadata:

• Repository description (if missing)
• Star count (stargazers_count)
• Short README preview

---

Quick Start

import pandas as pd

df = pd.read_parquet("awesome-repos.parquet")
print(df.head())

---

Intended Use

• Train repository-type classifiers (e.g., backend framework, UI library, dataset).

---

Reproducing the Crawl (Optional)

If you want to rebuild or extend the dataset:

Scripts & Config:

• awesome_final_repos.py – main crawler script
• github_api_utils.py – GitHub API helpers
• awesome_scrap_config.yaml – configuration

Install dependencies:

pip install -r requirements.txt

Run the crawler:

python3 awesome_final_repos.py \
  --root https://github.com/sindresorhus/awesome \
  --depth 2 \
  --workers 20

Key config options (awesome_scrap_config.yaml):

Key	Description
root	Root Awesome repo URL
depth	Maximum recursion depth (only repos at this depth are collected)
workers	Concurrency for async requests
output_dir	Output path for awesome-repos.parquet
fetch_stars	Whether to enrich repos with star counts
fetch_readme_preview	Whether to fetch first N chars of README

Authentication & rate limits:

export GITHUB_TOKEN=ghp_your_token_here

The crawler reads the GITHUB_TOKEN from the environment to increase GitHub API rate limits.

---

Output Schema

Column	Description
name	Repo name extracted from the Awesome bullet link
link	Canonical GitHub URL (e.g., https://github.com/owner/repo)
description	Short description (from Awesome bullet or enriched via GitHub API)
source_repo	The Awesome list repository referencing this entry at the deepest depth
stars (optional)	GitHub stargazers_count
readme_preview (optional)	First N characters of README markdown

Notes:

• Only “final” repos (non-Awesome) at depth == max_depth are included.
• Internal traversal bookkeeping is not written to the dataset.

---

Project Structure

.
├── awesome_final_repos.py      # Main crawler
├── awesome_scrap_config.yaml   # Configuration
├── github_api_utils.py         # GitHub API helpers
├── awesome-repos.parquet       # Dataset output
└── requirements.txt            # Dependencies

---

License

MIT.