Knowledge base overview

Stars vs library, capability map, and a suggested first path

Purpose

The knowledge base is the set of repositories you ingest so the RAG workspace can retrieve and answer. Starring a repo does not ingest it.

Sidebar Library plus toolbar Knowledge Q&A

Screenshot TODO · images/screenshots/knowledge-base/overview.webp

Stars vs knowledge base

	Stars	Knowledge base
Source	GitHub star sync	You mark ingest in Starcat
Scale	Thousands is normal	Curate (often dozens to a couple hundred)
List search	Yes	Yes (can limit to knowledge)
Default RAG scope	No	Yes, ingested only
Typical index	FTS and similar	Chunks + keywords + vectors

• Star does not auto-ingest
• Unstar does not auto-remove from the library
• Ingested but not starred: still askable, and you can still subscribe to Releases

What you need

1. Pro
2. Chat + Embedding → AI configuration
3. Some ingested repos → Adding repos

Without Embedding the workspace may open, but vector retrieval cannot build. See Index management.

Capability map

I want to…	Go here
Put a repo in the library	Adding repos
Ask a question	RAG workspace (⇧⌘K)
Open the cited source	Citations
Manage history	Sessions
Fix low coverage	Index management

Suggested path: ingest 20-50 repos (prefer using) → confirm Embedding and coverage → @ one repo first → then ask across repos.

Entries: sidebar Manage → Library; toolbar Knowledge Q&A or ⇧⌘K; knowledge-related Smart Collections.

Notes

• Semantic search is list-level vector search, not workspace Q&A.
• Ingesting everything you starred stretches index time and the Embedding bill.

Related

• AI overview
• Insights