Why Does the World Need Another Database?

Building a new generation of applications will inevetably require new ways of storing data, such as vectorstores, and we need a database that will acommodate these structures as well as those that are created next. At NexusDB, we believe that optimizing for a given structure might work well in the short term but cannot be durable and ultimately leads to inflexibility and decline.

When I first started building with AI models I was frustrated by the database options available to me. I needed to store user information in tables, so I spun up a MySQL database in Azure and spent a few hours troubleshooting the connection string, CORS, etc. I wanted my app to allow users to view documents they upload and ask questions about them to a chatbot. Because I wanted the documents to be stored long term, I needed to set up a new database using Azure Blob storage, which of course means a new connection string, CORS issues, and a more complicated schema for tracking which documents belong to each user, what their title is, and their location within the blob store.

Now I have an app that takes new user info and stores it in a table, can look up documents uploaded by the logged-in user and serve them on a web page, but don't have any way to feed the PDF to a chatbot. Enter data store number three: Pinecone.

Vector databases such as Pinecone are great, but require me to maintain at least one other databse for user information. The other major drawback is that they require users to upload everything they might want to search for, unless you add a search engine API as part of your agent flow.

What's Next?

Naturally, I got to thinking that there must be a better way. What if, instead of attempting to synchronize multiple different databases, we could store everything in one? What if the vector search could optionally include shared information, so you don't have to upload things considered to be common knowledge or waste time adding web search to your chatbot?

If such a database could be built, what could we build with it? I started this journey because I wanted to build a smarter chatbot - if common knowledge could be shared by all users, we could get away with smaller, faster, AI models and feed them all the information they need. This is still a high value use case, but the potential for an all-in-one knowledge engine is far greater than the backend for a single app. This is why we shut down our chatbot and put all our energy into releasing this database as a hosted product, so anyone can connect to it and build next-generation applications that were previously impossible or impractical.

Some ideas I've been thinking about:

• Email with an integrated CRM (see Yohei Nakajima's MindGraph, which I built a NexusDB integration for)
• Linked Notes app where AI makes connections for you (see Chia Jeng Yang's WhyHow.AI demo)
• Long term memory for an AI assistant

If you want to build these or have other ideas, let people know in our community on X! If you have any questions or want help with your integration, reach out to us at info@nexusdb.io.

I can't wait to see what you're building!

- Will Humble