Hi Product Hunt,

I'm dong4j. A year ago, I realized my GitHub Stars had become useless.

Not because GitHub is bad — but because starring is too easy. Click, and it's saved. After a few years I had 1,800+ starred repos and no way to find the one I actually needed. I once spent 20 minutes scrolling, trying to remember the name of a Swift clipboard manager I'd starred months earlier.

So I built Starcat: a native macOS app (SwiftUI, not Electron) that treats your GitHub Stars as a personal map of tools, libraries, and ideas worth remembering.

What it does today:

• Sync Stars into a native three-column workspace; cache & read READMEs offline

• Tags, notes, reading status, full-text (FTS5) + semantic search

• AI summaries, tag suggestions, translation, and in-repo chat (suggestions need your confirm)

• Release tracking, Trending/Weekly feeds, similar-repo recommendations

The part I'm most excited about — Knowledge Base RAG:

A dedicated workspace for natural-language questions across your knowledge base (repos you actively add to your library — not all stars).

Example: "which of my saved SwiftUI projects use Core Data?" That's not a keyword search; it needs semantic retrieval across READMEs, notes, and summaries, then a citation-backed answer.

Retrieval is local-first hybrid: FTS5 + local vector search (vDSP-accelerated, ~23x vs naive Swift), optional remote reranker. RAG is read-only — it never auto-writes tags or notes.

AI toolchain:

Local MCP Service + starcat-cli so Claude/Codex/other agents can query your stars and pull structured repo context. You bring your own AI provider (OpenAI-compatible, DeepSeek, OpenRouter, Ollama, etc.). Data stays on your Mac by default.

Open source: https://github.com/starcat-app/Starcat

Would love feedback from anyone with a large Stars library, people doing open-source research, or folks building local RAG on macOS.

Website:

https://starcat.ink