Embeddings map text to vectors where nearby means similar, the substrate under semantic search and RAG — and the quality of every retrieval is bounded by the quality of the embedding model.
Not yet scoped.
Why
The PoC embeds a corpus, runs nearest-neighbor search, and shows where semantic similarity diverges from what a user actually meant.
How it works
Not yet built.
Where it lands in Jayverse
Rabbit/Personas: if semantic search over skills, personas, or mandate lookups is ever added, budget for the embedding model's quality as the retrieval ceiling, and test where similarity diverges from intended meaning before shipping.
Number: any RAG-style lookup over Number's readings should log retrieval misses — semantically similar but wrong-intent matches — as a metric, not assume nearest-neighbor equals correct.
Key expressions
Words and phrases from this page worth keeping, with the Korean meaning and the sentence they come from.
Expression
뜻 · 쓰이는 자리
map ... to
~을 ~로 매핑하다, 대응시키다 · 한 형태의 데이터를 다른 형태로 변환해 연결할 때. "Embeddings map text to vectors"
the substrate under
~의 밑바탕이 되는 기반 · 여러 기능이 그 위에서 돌아가는 근본 토대. "the substrate under semantic search and RAG"
bounded by
~에 의해 상한이 정해지다, 제한받다 · 전체 품질이 한 요소의 한계를 못 넘을 때. "bounded by the quality of the embedding model"
diverge from
~에서 벗어나다, 갈라지다 · 결과가 원래 의도했던 것과 어긋날 때. "diverges from what a user actually meant"
nearest-neighbor search
최근접 이웃 검색 · 벡터 공간에서 가장 가까운 항목들을 찾는 검색 방식. "runs nearest-neighbor search"
RAG
검색 증강 생성(Retrieval-Augmented Generation) · 임베딩으로 관련 문서를 검색해 LLM 답변에 활용하는 방식, 임베딩의 대표적 활용처. "the substrate under semantic search and RAG"