GitHub Star Organizer

A GitHub Actions workflow system that automatically discovers, categorizes, and summarizes interesting repositories to help organize and expand your GitHub stars.

Features

• Discover Interesting Repos: Dual-model AI analysis (DeepSeek + Ollama) identifies 13 unique repositories per week that are unusual, innovative, or solve problems in creative ways
• Keyword-Based Categorization: Automatically categorizes repositories by topic using configurable keywords
• Weekly Organization: Creates GitHub issues for interesting discoveries with AI-generated summaries; uncategorized popular repositories are tracked in a local state DB (no issue/comment noise) for later keyword-growth distillation
• Smart Deduplication: Removes already-starred repos and previously flagged discoveries to prevent redundant suggestions

Workflows

discover-repos.yml

Runs weekly (Saturdays 11pm UTC) to discover new interesting repositories:

1. Fetches popular repos from GitHub (created after 2023, 1000+ stars)
2. Splits into uncategorized (Stage 1) and categorized (Stage 2+3)
3. Records new uncategorized repos in the state DB for keyword expansion
4. Uses DeepSeek and Ollama to identify 13 interesting repos per run
5. Creates GitHub issues with AI-generated summaries

Output: ~13 new interesting repositories + uncategorized repos for keyword growth each week

organize.yml

Automatically stars/unstarred repos based on discovery issue state

distill.py

Reads pending uncategorized repos from the state DB and uses DeepSeek to suggest config.json keyword/category updates, opening a PR with the result

Configuration

Edit config.json to customize:

• Category keywords for auto-categorization
• Exclusion patterns
• Repository search criteria

API Requirements

The workflows require:

• DEEPSEEK_API_KEY: DeepSeek API access
• OLLAMA_CLOUD_KEY: Ollama Cloud access
• GitHub token with issues:write and contents:read permissions

Testing

Run unit tests:

python -m pytest tests/unit/ -v

Project Structure

.
├── .github/workflows/      # GitHub Actions workflows
├── github_star_organizer/  # Main package
│   ├── categorizer.py      # Category matching logic
│   ├── gh_client.py        # GitHub GraphQL client
│   ├── issue_manager.py    # Discovery issue creation
│   └── state_db.py         # SQLite state (dedup, uncategorized tracking)
├── discover_repos.py       # Main discovery workflow
├── distill.py              # Category consolidation
├── config.json             # Configuration
└── tests/                  # Unit tests

How It Works

1. Search: Fetches up to 100 popular repos from GitHub
2. Split: Separates into categorized and uncategorized repos
3. Dedup: Removes user's starred repos and already-flagged discoveries
4. Identify: DeepSeek analyzes first half (identifies 7), Ollama analyzes second half (identifies 6)
5. Summarize: Both models generate detailed summaries for the 13 selected repos
6. Report: Posts results to GitHub issues for weekly review

Development

Changes to the discover-repos workflow are developed on feature branches and require PR review before merging to main.