---
license: cc-by-4.0
task_categories:
- text-retrieval
tags:
- github
- code
- metadata
- software-engineering
- open-source
- astronomy
- machine-learning
size_categories:
- 1M<n<10M
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*.parquet
---

# GitHub All Repositories Metadata Dataset (>= 3 Stars)

A comprehensive metadata dataset covering **5,289,726 public GitHub repositories** with **3 or more stars (>= 3)** spanning the history of GitHub from 2008 to 2026.

## Data Recency & Snapshot Notice

> [!NOTE]
> - **Snapshot Methodology:** This dataset combines a comprehensive historical base archive (up to mid-2022) with continuous periodic crawler snapshots (2023 through 2026).
> - **Star Counts & Metrics:** Star counts and repository metrics reflect the recorded values at the time of each repository's latest crawl/snapshot. Actively maintained projects may have higher live star counts on GitHub.com today.
> - **Coverage:** Includes active and historical repositories created between 2008 and 2026.

---

## Key Features
- **Total Repositories:** 5,289,726
- **Rich Topics & Descriptions:** Native tag arrays (`topics`), repo descriptions, licenses, homepage links, and star/fork metrics.
- **Fast Search:** Designed for instant topical search (e.g., Astronomy, BioInformatics, WebAssembly, Robotics, Machine Learning) using DuckDB, Polars, or Hugging Face `datasets`.
- **Deduplicated & Cleaned:** Deduplicated by `owner/repo`, prioritized by latest activity and rich metadata.

---

## Dataset Schema

| Column | Type | Example | Description |
| :--- | :--- | :--- | :--- |
| `full_name` | `string` | `astropy/astropy` | Full repository name (`owner/repo`) |
| `owner` | `string` | `astropy` | Organization or username |
| `name` | `string` | `astropy` | Repository name |
| `description`| `string` | `Astronomy and astrophysics core library for Python` | Repository description |
| `topics` | `list<string>` | `['astronomy', 'astrophysics', 'python', 'fits']` | List of repository topic tags |
| `language` | `string` | `Python` | Primary programming language |
| `stars` | `int64` | `4820` | Recorded stargazer count at snapshot |
| `forks` | `int64` | `1540` | Recorded fork count at snapshot |
| `license` | `string` | `bsd-3-clause` | Open source license key |
| `homepage` | `string` | `https://www.astropy.org` | Project homepage / website |
| `created_at` | `timestamp` | `2011-08-25 21:00:00` | Repository creation date |
| `last_active`| `timestamp` | `2026-06-15 12:30:00` | Most recent activity / push date |

---

## Quickstart: How to Find Projects (e.g., Astronomy)

### 1. Using Hugging Face `datasets` (Streaming / No full download required)
```python
from datasets import load_dataset

# Stream the dataset directly from Hugging Face
ds = load_dataset("Mieaz/github-repos-metadata-ge3", split="train", streaming=True)

# Find all repositories with 'astronomy' or 'astrophysics'
astronomy_repos = []
for repo in ds:
    topics = [t.lower() for t in (repo.get("topics") or [])]
    desc = (repo.get("description") or "").lower()
    
    if "astronomy" in topics or "astronomy" in desc or "astrophysics" in topics:
        astronomy_repos.append(repo)
        if len(astronomy_repos) >= 20:
            break

for r in astronomy_repos:
    print(f"⭐ {r['stars']:>5} | {r['full_name']:<35} | {r['description'][:60]}")
```

### 2. Using DuckDB (Instant SQL Queries on Remote Parquet)
```python
import duckdb

con = duckdb.connect()
query = '''
SELECT full_name, stars, language, description, topics
FROM 'hf://datasets/Mieaz/github-repos-metadata-ge3/data/train-*.parquet'
WHERE list_contains(topics, 'astronomy') 
   OR description ILIKE '%astronomy%'
ORDER BY stars DESC
LIMIT 15;
'''
print(con.execute(query).df())
```

---

## License & Attribution
- Data compiled from public open-source event logs and repository records under Creative Commons Attribution 4.0 (CC-BY 4.0).