Embeddings convert text to vectors. Vector stores enable semantic search by finding the nearest neighbors. This post covers embedding models, vector databases, and similarity search queries.
Embeddings
from langchain.embeddings import OpenAIEmbeddings
embeddings = OpenAIEmbeddings(model="text-embedding-3-small") vector = embeddings.embed_query("What is semantic search?") print(f"Embedding dimension: {len(vector)}")
Output:
Embedding dimension: 1536
Vector Stores
from langchain_community.vectorstores import FAISS, Chroma
from langchain.embeddings import OpenAIEmbeddings
embeddings = OpenAIEmbeddings()
# FAISS for in-memory search vector_store = FAISS.from_documents(documents, embeddings)
# Query results = vector_store.similarity_search("What is the main topic?", k=3) for doc in results: print(f"- {doc.page_content[:100]}...")
Output:
- The main topic is semantic search and retrieval-augmented generation...
FAISS is fast and lightweight; Chroma persists to disk.
Conclusion
Embeddings and vector stores are the retrieval backbone of RAG. OpenAI embeddings are high-quality; FAISS and Chroma offer flexible storage and search. Understanding vector databases enables building effective semantic search. Next: we'll build a complete RAG pipeline from data ingestion to generation.