README.md

---
license: other
license_name: permissive-mixed
license_link: LICENSE
task_categories:
- text-generation
- fill-mask
- feature-extraction
language:
- en
tags:
- code
- github
- ai-training
- llm
- fine-tuning
- code-generation
- python
- javascript
- typescript
- rust
- go
- permissive-license
- apache-2.0
- mit
- commercial-safe
pretty_name: HSH Intelligence — GitHub Code AI Training Corpus (Sample)
size_categories:
- 1K<n<10K
---

# HSH Intelligence — GitHub Code AI Training Corpus

**A 1,000-record evaluation sample of the HSH Intelligence GitHub Code AI Training Corpus.**

This is a curated, license-audited sample of source code from top-rated public GitHub repositories — prepared for large language model training, fine-tuning, and code understanding research.

The full corpus contains **5.6 TB** of source code (**211 million+ files**, **7.05 billion lines**) across **14 production languages** with **0.978 average quality score**.

---

## What's In This Sample

| Metric | Value |
|---|---|
| Records | 1,000 (curated subset) |
| Languages | 5 (Python, JavaScript, TypeScript, Go, Rust) |
| Format | Apache Parquet (Snappy compression) + CSV |
| Schema | 24 fields per record |
| Quality score (avg) | 0.978 / 1.000 |
| License coverage | 100% commercial-safe |

---

## Full Corpus Specifications

| Metric | Value |
|---|---|
| Total dataset size | 5.6 TB (raw) / 391 GB (Parquet, compressed) |
| Total records | 211 million+ code files |
| Total lines of code | 7.05 billion |
| Unique repositories | 3,710+ permissive-license repos |
| Programming languages | 14 production languages |
| Average quality score | 0.978 / 1.000 |
| Updates | Daily incremental |

**Languages covered:** Python, JavaScript, TypeScript, Go, Rust, Java, C++, Ruby, Swift, Kotlin, PHP, C#, Scala, Solidity

---

## Schema (24 Fields)

| Field | Type | Description |
|---|---|---|
| `id` | string | Unique record identifier |
| `repo_name` | string | GitHub repository slug (owner/repo) |
| `repo_owner` | string | GitHub username or organization |
| `repo_url` | string | Full HTTPS URL to repository |
| `file_path` | string | Relative path within repo |
| `file_name` | string | Filename with extension |
| `file_sha` | string | SHA-256 hash for deduplication |
| `code` | string | Raw source code content |
| `language` | string | Detected programming language |
| `language_extension` | string | File extension |
| `line_count` | integer | Total lines of code |
| `char_count` | integer | Character count |
| `token_count` | integer | Estimated tokens (tiktoken cl100k) |
| `license` | string | SPDX license identifier |
| `license_source` | string | Where license was detected |
| `license_confidence` | float | Detection confidence (0.0–1.0) |
| `commercial_safe` | boolean | Commercial-use safe flag |
| `repo_stars` | integer | GitHub star count |
| `repo_forks` | integer | GitHub fork count |
| `repo_description` | string | Repository description |
| `repo_primary_language` | string | Repository's dominant language |
| `repo_created_at` | timestamp | Repository creation date |
| `repo_updated_at` | timestamp | Last activity timestamp |
| `data_quality_score` | float | Composite quality score (0.0–1.0) |

---

## License Coverage (Commercial-Safe Only)

| License | Status | Notes |
|---|---|---|
| MIT | INCLUDED | Most permissive |
| Apache-2.0 | INCLUDED | Permissive with patent grant |
| BSD-2-Clause | INCLUDED | Permissive |
| BSD-3-Clause | INCLUDED | Permissive |
| ISC | INCLUDED | Permissive |
| GPL-2.0 / GPL-3.0 | EXCLUDED | Copyleft |
| AGPL-3.0 | EXCLUDED | Strong copyleft |
| LGPL-2.1 / LGPL-3.0 | EXCLUDED | Copyleft |
| No license / Proprietary | EXCLUDED | Default copyright |

---

## Quick Start

```python
from datasets import load_dataset

# Load sample
ds = load_dataset("HSH-Intelligence/github-code-corpus-sample")

# Inspect
print(ds)
print(ds["train"][0])

# Filter to high-quality Python only
python_only = ds["train"].filter(
    lambda x: x["language"] == "Python" and x["data_quality_score"] >= 0.95
)
```

```python
import pandas as pd

# Or load directly with pandas
df = pd.read_parquet(
    "hf://datasets/HSH-Intelligence/github-code-corpus-sample/github_code_sample_1000.parquet"
)

print(df.head())
print(f"Total records: {len(df):,}")
print(f"Languages: {df['language'].value_counts()}")
```

---

## Live API Demo

Try the full corpus via the live API sandbox (no signup required):

```bash
curl -H "X-API-Key: demo-key-12345" \
  "https://api.hshintelligence.com/api/v1/github-code-corpus?language=Rust&license=MIT&page_size=5"
```

Or run the interactive Google Colab notebook:
👉 https://links.hshintelligence.com/github-demo

---

## Use Cases

- **LLM pre-training** — multi-language code corpus for foundation models
- **Code completion fine-tuning** — Copilot-style models
- **Code search and retrieval** — embedding training
- **Code understanding research** — academic benchmarks
- **Vertical AI** — domain-specific code assistants

---

## Why This Corpus

| Vs. Alternative | HSH Edge |
|---|---|
| The Stack v2 | License-audited per file with provenance trail |
| Common Crawl code | Pre-filtered, deduplicated, quality-scored |
| Custom GitHub scraping | Saves 4+ months of engineering work |
| Internal datasets | EU AI Act Article 10 compliance ready |

---

## Compliance & Provenance

- **EU AI Act Article 10** ready (training data governance)
- Per-record license audit trail
- Source attribution retained (repo_name, repo_owner)
- Quality scoring per record
- No PII, no API keys, no secrets (sensitive content filtered)

Full provenance and audit report: 👉 https://links.hshintelligence.com/github-docs

---

## Full Corpus Access

This is a **1,000-record evaluation sample**. The full corpus is available via commercial license:

- **Format options:** Apache Parquet (default) / JSONL on request
- **Delivery:** Cloud signed URL / AWS S3 cross-cloud / sFTP
- **Custom subsets:** Filter by language, license, repo, or quality threshold
- **Updates:** Daily incremental updates included with annual licenses
- **Licensing:** 1-year non-exclusive commercial license

**Contact:** sales@healingsunhaven.com  
**Website:** https://www.hshintelligence.com  
**Live API:** https://api.hshintelligence.com

---

## About HSH Intelligence

HSH Intelligence is a Data Division of Healing Sun Haven LLC, building world-class AI training datasets and B2B intelligence products.

We build production-grade datasets across AI training, B2B intelligence, and decision-support — purpose-built for frontier AI labs and enterprise teams.

---

*This dataset is provided for evaluation purposes. The full 5.6 TB corpus is available under commercial license. License audit and provenance documentation included with all enterprise contracts.*