Skip to content

Python Vector Database: Effortlessly Manage and Query Vectors


Python Vector Database: Embeddings and Vector Databases With ChromaDB

Table of Contents

The era of large language models (LLMs) is here, bringing with it rapidly evolving libraries like ChromaDB that help augment LLM applications. You’ve most likely heard of chatbots like OpenAI’s ChatGPT, and perhaps you’ve even experienced their remarkable ability to reason about natural language processing (NLP) problems.

Modern LLMs, while imperfect, can accurately solve a wide range of problems and provide correct answers to many questions. However, they do have their limitations. LLMs have a limited amount of training data and can only process a certain number of tokens. As a result, LLMs might not be able to provide relevant responses about topics that are not in their training data.

To address these limitations and scale your LLM applications, you can use a vector database like ChromaDB. A vector database allows you to store encoded unstructured objects, like text, as lists of numbers that can be compared to one another. This enables you to find relevant documents or information for a given question that you want an LLM to answer.

In this tutorial, you will learn about representing unstructured objects with vectors, using word and text embeddings in Python, harnessing the power of vector databases, encoding and querying documents with ChromaDB, and providing context to LLMs like ChatGPT with ChromaDB.

Represent Data as Vectors

Before diving into embeddings and vector databases, it’s important to understand what vectors are and how they represent data.

Vector Basics

A vector can be thought of as an array of numbers. In Python, the NumPy library is often used to work with vectors represented as arrays. For example, a vector can be represented using a NumPy array as follows:

import numpy as np
v = np.array([1, 2, 3])

Vector Similarity

The similarity between two vectors can be measured using various techniques, such as the cosine similarity. The cosine similarity calculates the cosine of the angle between the two vectors, resulting in a value between -1 (completely dissimilar) and 1 (completely similar).

import numpy as np
from scipy.spatial.distance import cosine
v1 = np.array([1, 2, 3])
v2 = np.array([4, 5, 6])
similarity = 1 - cosine(v1, v2)

Encode Objects in Embeddings

In the next section of this tutorial, you will learn about embeddings and how to encode objects using embeddings.

Word Embeddings

Word embeddings are dense vector representations that capture semantic information about words. These embeddings can be generated using pre-trained models, such as Word2Vec or GloVe, which have been trained on large corpora of text.

Let’s see an example of how to use word embeddings in Python:

from gensim.models import KeyedVectors
word_embeddings = KeyedVectors.load_word2vec_format('path/to/word2vec.bin', binary=True)

Text Embeddings

Text embeddings are similar to word embeddings but capture the meaning of larger chunks of text, such as sentences or paragraphs. These embeddings can be generated using models like Universal Sentence Encoder or InferSent, which take into account the context and structure of the text.

Here’s an example of using text embeddings in Python:

import tensorflow as tf
import tensorflow_hub as hub
text_embeddings = hub.load('')

Get Started With ChromaDB, an Open-Source Vector Database

ChromaDB is an open-source vector database designed to efficiently store and query large collections of vectors. It provides a simple API for encoding, indexing, and querying vectors, making it a powerful tool for working with vector data in Python.

What Is a Vector Database?

A vector database is a specialized database that is optimized for storing and querying vectors. It allows you to store vectors as well as perform operations like similarity search, which enables you to find vectors that are similar to a given query vector.

ChromaDB is one such vector database that offers features like fast nearest neighbor search and efficient indexing for high-dimensional vectors.

Meet ChromaDB for LLM Applications

ChromaDB can be a great addition to LLM applications, as it allows you to store and query vectors representing unstructured objects, such as sentences or documents. With ChromaDB, you can provide relevant context to LLMs and enhance their ability to generate accurate responses.

To get started with ChromaDB, you will need to install the chromadb Python package and create a database connection:

from chromadb import Database
db = Database(database_uri='mongodb://localhost:27017/mydatabase')

Practical Example: Add Context for a Large Language Model (LLM)

In this practical example, you will learn how to use ChromaDB to add context to a large language model (LLM). The goal is to improve the LLM’s ability to generate accurate responses by providing relevant context from a collection of documents.

Prepare and Inspect Your Dataset

Before adding context to the LLM, you need to prepare and inspect your dataset. This involves gathering the relevant documents and organizing them into a format that is compatible with ChromaDB. You should also take a look at the structure and content of the documents to ensure they meet your requirements.

Create a Collection and Add Reviews

In this step, you will create a collection in ChromaDB and add the prepared documents to the collection. This allows you to efficiently store and query the documents using their vector representations.

collection = db.create_collection('reviews')
for doc in documents:
vector = encode_document(doc)

Connect to an LLM Service

To connect the LLM to ChromaDB, you need to establish a connection between the LLM service and the ChromaDB database. This allows the LLM to retrieve relevant documents from ChromaDB based on the provided context.

llm_service = LLMService()

Provide Context to the LLM

Finally, you can provide context to the LLM by retrieving relevant documents from ChromaDB and passing them to the LLM for processing. This allows the LLM to take the context into account when generating responses.

context = get_context_from_user()
relevant_documents = collection.find_similar(context)
responses = llm_service.generate_responses(relevant_documents)


ChromaDB is a powerful tool for working with vector data in Python. In this tutorial, you learned about representing data as vectors, using embeddings to encode objects, and how to get started with ChromaDB, an open-source vector database. You also explored a practical example of adding context to a large language model (LLM) using ChromaDB.

With the knowledge gained from this tutorial, you can now leverage vector databases like ChromaDB to enhance and scale your NLP and LLM applications.