Replit + Chroma: AI for the next billion software creators


Anton Troynikov

Replit x Chroma
Replit x Chroma

Guest post by Chroma

Today we’re announcing the Chroma template for Replit, the next step towards bringing the power of AI application development to the next billion software creators.

With the Chroma template, developers can easily create AI applications with state and memory. Want to make ChatGPT for your email? Or chat to your textbooks while you study? Want LLMs to know about the latest news stories? Together with Replit, Chroma makes all that and more easy.

Chroma Template Demo

The AI software revolution

We are entering a transformative era for software development. People around the world are building AI-powered applications, and exploring the potential of the latest large models. The frontier expands every day.

The tools and techniques for this new era are being worked out almost in real-time. New ideas which use more kinds of data, more ways to interact with AI, and more insights about how the models work are constantly being developed.

One of the most powerful ways to work with AI is giving it state, memory and pluggable knowledge. Adding these capabilities to LLMs allows developers to create question answering bots, personal assistant agents, and applications which can dynamically interact with other APIs.

The pace of change can be dizzying, but with Replit and Chroma, experimenting with, building, and sharing powerful AI-driven applications is easier than ever.

The secret sauce: Embeddings

Just like traditional applications use databases to store and retrieve data, AI-powered applications need an AI native storage and memory layer. To work with models in the loop, data needs to be represented in an AI native way - using embeddings.

Embeddings are a numerical vector representation of all kinds of data, generated using an embedding model. When two pieces of data are similar in meaning, for example, two sentences that are about the same thing, their embeddings are close together in vector space.

By representing data in this way, we can interact with the data the same way we do with the models themselves. For instance, we can find the most relevant text documents for a specific question or topic, and then use them as the context for the LLM to answer the query accurately.

The Chroma Replit template

Chroma is the easiest to use embeddings store for your AI application. And now with the Chroma template, you can get it right from inside a Repl. Just add your own data!

Chroma handles embedding documents and queries for you, and stores your documents alongside their embeddings. It supports filtering and dynamic updating out of the box, making it perfect for you next AI-powered application. Learn more from the Chroma getting started guide.

The template includes a demo for question-answering using OpenAI’s ChatGPT API. We’ve embedded the popular textbook “Artificial Intelligence: A Modern Approach” by Stuart Russell and Peter Norvig, letting you ask it questions in a conversational style.

Give it a try! We can’t wait to find out what you’ll build with Replit and Chroma!

More blog posts