GitHub Star Intelligence

GitHub Star Intelligence is a personal knowledge system for organizing GitHub starred repositories with LLM assistance.

The system will inspect starred repositories, infer an evolving taxonomy from repository content, classify new stars weekly, preserve category history, and maintain a searchable knowledge base of repository usefulness.

Core idea

GitHub Stars are treated as a research library, not a bookmark list. The system should explain why each repository matters, how it relates to other starred repositories, and whether it is worth revisiting, cloning, studying, building with, contributing to, or monitoring.

Initial scope

• Pull starred repositories for the GitHub account eazene.
• Detect new starred repositories incrementally.
• Inspect repository metadata, README content, topics, language, and documentation.
• Use an LLM to classify repositories into an emergent taxonomy.
• Reuse existing categories when possible.
• Create new categories only when justified.
• Store repository records, category history, similarity relationships, embeddings, and weekly reports.
• Generate weekly summaries for Hermes agent review.

Design principle

Do not begin with a fixed taxonomy. Let categories emerge from the repository corpus, while controlling category sprawl through similarity checks, confidence scores, and periodic taxonomy review.

Repository structure

docs/       Product, architecture, taxonomy, and Hermes agent specifications
schemas/    JSON schemas for repository, category, and weekly report records
prompts/    LLM prompts for classification and taxonomy review
src/        Future implementation code
tests/      Future tests
data/       Local runtime data, ignored by git

Status

Seed repository. Product requirements and system design are being established before implementation.