Vector store selection (Pinecone, Chroma, FAISS) in Agentic AI

Introduction

Vector stores help us quickly find similar items by comparing their numbers. Choosing the right one makes searching faster and easier.

You want to find similar images or texts quickly.

You need to search through a large collection of data by meaning, not just words.

You want to build a chatbot that remembers past conversations.

You need to organize and search data from sensors or devices.

You want to add smart search to your app without building everything from scratch.

Syntax

Agentic AI

vector_store = VectorStoreType(parameters)
results = vector_store.search(query_vector, top_k)

Replace VectorStoreType with Pinecone, Chroma, or FAISS depending on your needs.

parameters include things like API keys, index names, or file paths.

Examples

Using Pinecone with an API key to search top 5 similar vectors.

Agentic AI

import pinecone
pinecone.init(api_key='your_key')
index = pinecone.Index('example-index')
results = index.query(query_vector, top_k=5)

Using Chroma to find 3 closest matches in a collection.

Agentic AI

from chromadb import Client
client = Client()
collection = client.get_collection('my_collection')
results = collection.query(query_vector=query_vector, n_results=3)

Using FAISS to search 4 nearest neighbors with Euclidean distance.

Agentic AI

import faiss
index = faiss.IndexFlatL2(dimension)
index.add(data_vectors)
D, I = index.search(query_vector, k=4)

Sample Model

This program creates a small set of 3D vectors, builds a FAISS index, and searches for the 2 closest vectors to the query. It prints distances and indices of the closest matches.

Agentic AI

import numpy as np
import faiss

# Create some example data vectors (5 vectors, 3 dimensions each)
data_vectors = np.array([
    [1.0, 0.0, 0.0],
    [0.0, 1.0, 0.0],
    [0.0, 0.0, 1.0],
    [1.0, 1.0, 0.0],
    [0.0, 1.0, 1.0]
], dtype='float32')

# Build FAISS index
index = faiss.IndexFlatL2(3)  # 3 is the dimension
index.add(data_vectors)

# Query vector
query_vector = np.array([[1.0, 0.0, 0.0]], dtype='float32')

# Search top 2 nearest neighbors
D, I = index.search(query_vector, 2)

print('Distances:', D)
print('Indices:', I)

Important Notes

Pinecone is a cloud service, so you need an internet connection and API key.

Chroma is easy to use locally and good for small to medium data sets.

FAISS is very fast and works well for large data but needs more setup.

Summary

Vector stores help find similar data quickly using numbers.

Pinecone, Chroma, and FAISS each have strengths for different needs.

Choose based on your data size, speed needs, and whether you want cloud or local storage.