OpenStar Memory

OpenStar Memory is the visible profile behind personalization.

Memory is a control surface

Memory should not feel like a hidden recommendation model. Users should be able to inspect what OpenStar learned, correct it, and see the feed change.

GitHub taste

Stars and public repositories create the first map of languages, topics, and habits.

OpenStar actions

Save, Watch, open, and hide actions turn casual browsing into stronger preference signals.

Chat intent

Questions reveal jobs to be done: adoption, alternatives, risk review, or monitoring.

Explicit edits

User edits should override weak inferred taste when recommendations drift.

Signal types

Signal	Example	Strength
Positive	Save a repo, watch a repo, open a repeated topic.	Strong when repeated.
Negative	Hide a repo, mark a topic stale, ignore repeated clusters.	Strong for reranking.
Explicit	Edit interests or tell Chat what matters.	Strongest user-owned signal.
Background	Public stars, owned repos, language distribution.	Useful for cold start.

What Memory should explain

What topics, languages, and repo families OpenStar thinks matter to the user.

Which recent actions changed the profile.

Which signals are weak, stale, or overrepresented.

How to tune the feed without deleting the whole account.

No fake personalization

If OpenStar has too few signals, it should say so and explain how to build Memory instead of pretending to know the user.