# SentenceTransformers **SentenceTransformers (SBERT) is the Python library most people reach for to turn text into [[Embeddings]] locally, without calling an API.** It wraps transformer models behind a two-line interface and runs on CPU or GPU. ```python from sentence_transformers import SentenceTransformer model = SentenceTransformer("all-MiniLM-L6-v2") vectors = model.encode(["some text", "more text"]) ``` Why it's everywhere: - **Local and free.** No API key, no per-token cost, no data leaving the machine. That's why tools like [[CocoIndexCode]] bundle it as the default, with cloud providers (via [[LiteLLM]]) as the opt-in alternative. - **Big model zoo.** Hundreds of pretrained embedding and re-ranking models on Hugging Face, from tiny MiniLM to multilingual and domain-specific ones. - **Built for retrieval.** Bi-encoders for fast similarity search, cross-encoders for precise re-ranking, plus support for asymmetric search where the query and document get encoded differently. It's the embedding layer underneath a lot of [[Retrieval-Augmented Generation (RAG)]] and [[Semantic Search]] stacks. License: Apache 2.0. ## References - https://www.sbert.net/ - https://github.com/UKPLab/sentence-transformers ## Related - [[Embeddings]] - [[Semantic Search]] - [[Retrieval-Augmented Generation (RAG)]] - [[CocoIndexCode]] - [[qmd]] - [[LiteLLM]]