ChromaDB Tutorial For Beginners | Vector Database Explained | ChromaDB Crash Course | Simplilearn

[music] Have you ever wondered how AI applications like Chat GPT remember and retrieve relevant information from thousands of documents in just seconds? The answer often lies in vector databases. And one of the most popular ones is Chroma DB. Chroma DB is an open-source vector database designed specifically for AI applications. It helps store, organize, and retrieve embeddings, enabling features like semantic search, document retrieval, recommendation systems, and RAG, also short for retrieval augmented generation. In this beginner-friendly tutorial, we will explore what Chroma DB is, why it is important in modern AI workflows, how vector databases differ from traditional databases, and how to get started with Chroma DB step by step. By the end of this tutorial, you'll understand how Chroma DB serves as the memory layer behind many intelligent AI applications. So, let's get started. That's it. If these are the type of videos you'd like to watch, then hit that like and subscribe buttons, and hit the bell icon to get notified whenever we host. Also, just that you know, if you want to upskill yourself, master generative AI and artificial intelligence skills, and land in your dream job, or grow in your career, then you must explore Simplilearn's cohort of various generative AI courses and certifications. Simplilearn offers various certification programs in collaboration with some of the world's leading universities. Through our courses, you will gain knowledge and work-ready expertise in skills like advanced Python, machine learning, generative AI, and over a dozen others. After completing these courses, many have transitioned into an AI and machine learning role as a fresher, or moved on to a higher-paying job and profile. If you are passionate about making your career in this field, then make sure to check out the link in the pin comment and description box below to find a genuine VR program that fits your experience and areas of interest. Now, without further delay, let's test your knowledge on Chroma DB with a small quiz. So, here's the question. What is Chroma DB primarily used for? Your options are: A, video editing; B, vector data storage and retrieval; C, web hosting; and D, operating systems. Don't forget to leave your answers in the comment section below. Now, over to the training. So, have you ever wondered how AI tools like ChatGPT, Netflix recommendations, Spotify playlists, or modern chatbots seem to understand exactly what you're looking for, even when you don't use the exact words? That's interesting, right? So, the answer lies in a powerful technology called vector database. In today's tutorial, we're going to understand what a vector database is, why it has become one of the most important technologies in artificial intelligence, and how it powers modern AI applications such as RAG systems, semantic research, AI assistants, recommendation engines, and a lot more. So, let's get started. Now, let's get started by understanding one single point: why traditional databases are not enough. Imagine you're building an AI chatbot for a company. The company has thousands of documents, including employee handbooks, HR policies, training guides, product manuals, and customer support documents, etc. Now, let's suppose an incident. An employee asks, "Can I take a week off during December?" Inside the HR document, the information might be written as, "Employees are entitled to 20 annual vacations." Notice something interesting? The user's question never used the words "week off." A traditional database typically searches for exact keywords or exact matches. Since the word week off is different from vacation days or leaves, the database might struggle to find the answer or it may not even find the answer. This is where traditional search stops failing. Humans understand the meaning. Computers traditionally understand words. Vector databases help computers understand the meaning and that's what makes them so. Now, let's understand vectors before standing understanding vector databases. We first need to understand vectors. Now, for a moment, come out of mathematics for a moment. Whenever you listen to this term vector, think about how your brain works. When someone says the word doctor, your brain automatically connects with hospital, medicine, patients, health care, treatment, and diagnosis. You don't simply store the letters as d o c t o r. You store the meaning behind the word. AI models work in a similar way. They convert text, images, audio, or even videos into numerical representation called as a vector. A vector is simply a list of numbers that represent the meaning of the data. For example, take this sentence, "I love watching movies." might become a matrix of numbers. Let's say 0.24 0.78 0.11 0.56 and a lot more. Another sentence, which is "I enjoy films." might generate a vector that is extremely similar to the previous one. Although the words are different, the meaning is almost identical. As a result, they are going to be positioned close to each other. This process of converting data into vector is called as embedded generation. And the resulting numerical representation is called as embed. Now, why similar meanings create similar vectors? Imagine a 3D map which is giant. Every word, every sentence, image, or document is placed somewhere on this map. Items with similar meanings are positioned together. Items with unrelated meanings are placed far apart. Let's consider for an example, dog and puppy would be close. Car and vehicle will also be close. Pizza and pasta might be relatively close, but pizza and an airplane will be far apart. The AI model learns these relationships automatically during training. This allows machines to understand concepts instead of simply matching the words. And this is the foundation of vector databases. Now, the actual definition. What is a vector database? So, a vector database is a specialized database designed to store and search vectors efficiently. Instead of searching for exact keywords, it searches for similar meanings. When a user asks a question, the question is converted into a vector. The vector database compares it against millions of stored vector. It finds the vectors that are most similar. The corresponding documents are returned. This process is called as similarity search. The database isn't looking for matching words, it's looking for matching meanings, and that's the key difference. Now, a real-world example. Let's imagine an online shopping platform. A customer searches, "I need comfortable footwear for jogging." However, the product description says, "Lightweight running shoes with extra cushioning." A traditional search might miss some relevant products because the exact keywords don't match in this scenario. But with a vector database, it understands footwear similar to jogging footwear, which is also similar to running with comfortable cushioning. As a result, it can return highly relevant products even when the wording is completely different. This creates a much better search experience. Now, let's learn how vector databases power RAG. One of the most popular AI architectures today is RAG, which stands for retrieval augmented generation. Let's understand it using a simple analogy. Imagine you're taking an exam. Scenario one, closed book exam. You can only answer based on what you remember. If you don't know the answer, you're stuck. This is similar to a large language model working alone. It can only answer using information learned during the training. Now, let's consider a scenario two, open book exam. Now, imagine you have access to a textbook. When a question appears, you find the relevant chapter, read the information, use reasoning skills, write the perfect answer. This is exactly how RAG works. The LLM acts like the student, the vector database acts like the textbook. Whenever a user asks a question, the question becomes a vector. The vector database retrieves relevant information, and the information is sent to the LLM. The LLM generates a more accurate answer as the output. This dramatically reduces hallucinations and improves reliability. Now, let's go through the complete RAG workflow. Let's summarize the process. So, the first step, it collects the documents. Step number two, it converts the documents into embeddings using an embedding model. Step number three, it stores those embeddings inside a vector database. Step number four, user submits the question. And step number five, convert the question into an embedding. And step number six, perform similarity search. Step number seven, retrieve relevant documents. Step number eight, send documents plus user question to the LLM. And lastly, the step number nine, which is generate the final answer. This entire retrieval process happens within mere seconds. Now that we have a quick understanding of what a vector database is, then let's go through one of the popular vector bases today, which is the Chroma DB. A beginner-friendly option, easy to install, perfect for learning and building local rack projects. Beyond rack, are there use cases? Many people think vector databases are only used in AI chatbots. That's not true. They power numerous applications, like for example, Spotify recommends songs by comparing listening patterns, Netflix recommends movies based on similarity between user and content. Image search, like upload a photo of a chair, the system can find visually similar chairs even without text descriptions. This can also be used in fraud detection. Unusual transactions appear far away from normal transaction patterns in vector space. This helps identify suspicious activities and a lot more. Now, let's dive into the demo part of the Chroma DB. Now, let's start by searching Chroma DB on any of the browsers. I'm using Google here. So, you all you can do is just I'm also already on the website, but still, let's get started from the beginning. I'll open up a new tab. So, you can just type down Chroma DB or the website, actually, trychroma.com. So, Chroma DB. That's the first, chroma.com. And here you have all the related documentation, research, resources, pricing, etc. So, we have a lot of companies using the Chroma DB. And if you scroll down, here you can see they have 15 million monthly downloads, low latency, up to 10% cheaper, no engineering ops, etc. And you also have sparse vector search, vector search, full text search, and you also have the code right next to it, and you have the command line interface as well, which looks like the one on the right-hand side. And if you scroll up again, here you have the documentation. For today's tutorial, we'll be using the documentation for beginner-friendly understanding of how Chroma DB actually works. Now, here you have a quick introduction to Chroma DB, what Chroma DB offers. It offers document storage, embeddings, vector search, full text and re-query search, meta filtering, multimodal retrieval, and all of those. So, it's basically metadata filtering. And here also you have the quick start, or you can also go ahead with the server option. You have examples here, agentic search, code search, etc. And it is sponsored by the Apache 2.0, the one which owns Hadoop and Spark. And you also have the Chroma Cloud for this. And you can just click on get started here. The button is right next to here. You can just click start on that. And here you have the codes. So, if you want to go ahead with Python, they have their own dedicated tutorial. If you want to go with TypeScript, they have a tutorial for that as well. You can also go ahead with Rust. And now let's go ahead with it. So, you can get started by installing this first, and you can also install this with AI. And if you scroll down here, you have all the documentation required for that. Now, you can also install this manually. So, before you get started with this, make sure that you have Anaconda installed into your PC, because I'm going to start this on Jupyter Notebook. You can also go ahead with any of your So, you need to start Jupyter Notebook. And in case if you don't have Jupyter Notebook, don't worry. We'll show you the installation of Anaconda as well. And we also have a dedicated tutorial for that. We will link this in the tutorial for that as well. Now, if you don't have that, it's really simple. The Jupyter Notebook is now currently activated. I'll also show you how to install Anaconda in a minute. So, just click on a new tab. I'll open up a new tab here somewhere. I'll first open up a new Python terminal where we will be executing all these code segments according to the vector documentation. So that is beginner friendly. So you need to install Chroma DB first. You can either copy this command into the Anaconda prompt or you can also paste this in your Jupyter Notebook as well. So I'm opening up an Anaconda prompt. If you don't have, it's really simple. Just type down Anaconda download somewhere on the browser. You will have the links. Just follow the first link and follow the simple steps which are provided here in the distribution documentation. You just click on get started. It will download and it's really simple. You can just click on next next like that. Simple process. Don't worry about it. So I'll copy this command and go back to Anaconda prompt right which is right here. I'll just paste this command install command Chroma DB. So I've already installed this so you'll get requirements already satisfied. So you can if you're new, you'll just get some command prompt executed or if you are you know feeling uncomfortable with the command prompt, you can always paste this command onto Jupyter Notebook window as well. That will also work. So let me show you that. Shift enter. There you go. You'll have some commands running in the background. So requirement already satisfied since we've already installed this. Now let's get started by importing the Chroma DB. So I'll just copy the first line and paste it over here. And if the Chroma DB successfully installed, you'll not face any kind of errors here. You can just skip to the next cell. It's really satisfied here so you don't have to worry about any errors. Now let's create a client. So the client is really important here. Client is the one which helps you create collections and collections are where you will store your embeddings, documents and any additional metadata. Collections index your embeddings and documents and enable efficient retrieval and filtering. You can create collection with a name. So here they've already created collection with some name which says my collection. So, you can also create your customized name for the collection. I'll go with the name given by the documentation which is my collection. So, this is the collection which we will be creating using the client. Basically, let me type down a comment here. This is the one which allows you to have an instance created. It instantiates a collection and collection is the one which helps you to manipulate or do any operations on the vector database. Let me type down that here as a comment. And this client will create a collection in your vector database. Let's use as VDB for short. And that will help you to interact with your database or interact with the underlying data for the vector database. So, let's execute this. Shift enter. There you go. The collection has been successfully created. Now, let's go over the next step where we will add Okay, let me scroll down. Yeah. So, here is the next step. Let's create or add some documents. So, Chroma will store your text and handle embeddings and indexing automatically. You can also customize the embedding model. You must provide unique string IDs for your documents. So, let's copy this code and paste in the Jupiter notebook. And here you have two different documents. So, the first one is a simple string document which says this document is about pineapple. And the other one is this document is about oranges. Now, you can also see ID one and ID two. So, the first ID one is the ID for your first document. And the next ID two is the ID for your next document. Let me type down this as a comment. ID one is equal to this is the command for pineapple. ID two is for the next document which is about Let's copy and paste this again. This is the document about oranges. So, you can remove the additional space. So, you can remove that additional space there and shift enter. That's executed. Now, the next step, query the collection. So, the entire thing about creating a collection in the vector database is about querying the data, right? You can query the collection with list of queries and Chroma will return the n most similar results. It's that easy. So, the number of similar results which you found in the query or the retrieval process, that will be given to you. So, let's copy and paste it somewhere in the Jupiter notebook. So, this document is about Hawaii. Chroma will embed this for you and it will provide n number of results. So, I can also customize the results. So, for here you have two, so it's giving you the two results and it's printing the results. And now you can inspect, which is the next step. From the above, you can see the output query about Hawaii is systematically or semantically more similar to the document about pineapple. So, you'll get a similar result over here as well. So, let me paste that code right here and execute this. So, there you go. That's your output. This is the document about pineapple. This document is similar for you as So, you can next step is about trying it out yourself. What we have tried querying with this is a document about Florida and here's the full example. So, you can just copy paste these commands onto the Jupiter notebook and try your versions of the queries and you can also check switching these commands. So, here are the first one and this is the next one and you can also print the results for yourself. And this acts as a beginner-friendly guide for vector database. And that's the result when I execute this. And the next steps, in this guide you will be using Chroma's in-memory client for simplicity. It starts a Chroma server in memory and so any data you will ingest be lost when your program terminates. You can use this and here also you have some other operations to the collections. You can manage collections, add, update, delete the data, configure collections, and querying as well. And you have the CLI model, which you can have the command line interface, run as a chroma server, and others in cloud, and operating as an open source, migration, and troubleshooting as well. So, here is the documentation, and they also have a GitHub page. So, I'll drop this link in the description box below. You can also go through their GitHub page and all the resources you can go ahead with. And this link will be dropped in the description box below for your understanding. And with that, we have come to an end of this session on Chroma DB tutorial for beginners. And with that, we have reached to the end of this session on Chroma DB tutorial for beginners. If you found this useful, drop a like, and subscribe for more AI breakdowns. And if you want deeper dives into topics like prompt engineering, AI tools, or building your own chatbot, let me know in the comment sections below. This is a Simply Learn's video. Until next time, thanks for watching, and stay tuned for more from Simply Learn.