vector-databases

Vector Databases SOP

Embed documents, store vectors, run similarity search, and build RAG retrieval pipelines using ChromaDB, pgvector, and popular embedding models.

When to Use

• User wants to build semantic search over a document corpus
• User wants to implement a RAG (Retrieval-Augmented Generation) pipeline
• User wants to store and query embeddings efficiently
• User wants to find similar items by meaning, not keyword
• User wants to choose between embedding models and vector stores

---

Part 1 — Embedding Model Selection

Model	Provider	Dims	Best For	Cost
text-embedding-3-small	OpenAI	1536	General purpose, fast	~$0.02/1M tokens
text-embedding-3-large	OpenAI	3072	Higher accuracy	~$0.13/1M tokens
all-MiniLM-L6-v2	sentence-transformers	384	Local, fast, English	Free
all-mpnet-base-v2	sentence-transformers	768	Local, balanced quality	Free
BAAI/bge-m3	HuggingFace	1024	Multilingual, local	Free
nomic-embed-text	Ollama	768	Local, good quality	Free
mxbai-embed-large	Ollama	1024	Local, high quality	Free

OpenAI Embeddings

import os
from openai import OpenAI

client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])

def embed_openai(texts: list[str], model="text-embedding-3-small") -> list[list[float]]:
    """Embed a batch of texts. Max 2048 inputs per call."""
    # Replace newlines — they degrade embedding quality
    texts = [t.replace("\n", " ") for t in texts]
    response = client.embeddings.create(input=texts, model=model)
    return [item.embedding for item in response.data]

# Single text
vec = embed_openai(["Hello world"])[0]
print(f"Dimension: {len(vec)}")

Sentence-Transformers (local)

# pip install sentence-transformers
from sentence_transformers import SentenceTransformer

model = SentenceTransformer("all-MiniLM-L6-v2")

def embed_local(texts: list[str]) -> list[list[float]]:
    embeddings = model.encode(texts, batch_size=64, show_progress_bar=True, normalize_embeddings=True)
    return embeddings.tolist()

vecs = embed_local(["Semantic search is powerful", "Vector databases store embeddings"])
print(f"Shape: {len(vecs)} x {len(vecs[0])}")