Dgraph Labs is becoming part of Hypermode - Dgraph Blog

Blog
1

Dgraph Labs is becoming part of Hypermode

Dgraph Labs is becoming part of Hypermode

@Gajanan Chinchwadkar, CTO of Dgraph Labs | @Kevin Van Gundy, CEO of Hypermode.

Today we announce Dgraph Labs is becoming part of Hypermode! This ensures the continued development & support of Dgraph, the open-source graph database, as well as Badger and Ristretto.

Gajanan Chinchwadkar, CTO of Dgraph Labs, sat down with Kevin Van Gundy, CEO of Hypermode, to answer anticipated questions.

  • Why is Dgraph Labs joining Hypermode? What role will the Dgraph project play in Hypermode?
  • What happens to Dgraph and other open-source projects?
  • What happens to the Dgraph Labs Team?
  • What happens to Dgraph Cloud?
  • Where can I learn more?

GC & KVG will also host a livestream on Wednesday, November 8 at 10am Pacific. To join, please register below or subscribe on YouTube.

Why is Dgraph Labs joining Hypermode? What role will the Dgraph project play in Hypermode?

Gajanan Chinchwadkar (GC): With the tailwinds for graphs in AI, trends like RAG, interpretability, and vector search, Dgraph is seeing a surge of interest. The outreach from community members has shown us there’s a unique opportunity to make the world of AI more accessible. Dgraph (the database) is a great foundation; graph structures are used widely in AI/ML. Further, adding graphs to LLMs makes things like model trust and debugging far easier than with traditional structures. I wrote about this in my recent Graph + Vector blog post.

On Hypermode and its CEO – Kevin isn’t new to the Dgraph community. We met when I was working at NASA and he was at Neo4j, nearly ten years ago. We gave a seminar to the NASA team about graph structures. When I joined Dgraph Labs, we brought him on as an advisor. He’s been helping our team since 2022. So I’m excited for this move and to work with Kevin! The vision for Hypermode and how Dgraph’s team and technology amplify that vision is compelling, so Kevin and I have worked together for the past several months to get to this announcement.

Kevin Van Gundy (KVG): I’ve been a fan of Gajanan, graphs, and specifically Dgraph for a long time.

Every developer’s theme this year is firmly “experiment with AI”, and Hypermode sees a massive opportunity for simplifying how developers build AI into their sites and applications. Major projects involving AI shouldn’t automatically add severe and insurmountable complexity, let alone early POCs and experiments.

The toolchain we’re building breaks down into three main parts: an SDK, an inference engine, and a data plane.

The SDK creates shorthand methods for requesting input from your AI models, the inference engine allows you to chain together services in the programming language(s) you’re most comfortable with. This is all enabled by a data plane comprised of a fully managed vector-graph database with model management, embedding automation, and analytics.

Our goal is to make it easy for developers to augment their applications with artificial intelligence in a few lines of code. We can solve the fundamental problems of performance, trust, and complexity.

GC: Dgraph creates a foundation for us to marry vectors, structured data, and models. For example, when thinking about adding guardrails to an LLM, you can use graphs to trace the shortest paths between similar objects found via vector search. If no connections exist between those nodes in graph space, you might want to throw that result out as a possible hallucination.

By combining our backend architecture with a frontend SDK, we can optimize how we compute and serve responses to the frontend to provide more reliable infrastructure with better performance. We’re also making the entire platform extensible through integrations and key partnerships across the AI toolchain to benefit companies with engineering investments in specific technologies. Hypermode.com is a live demo application showing off our beta.

What happens to Dgraph and other open-source projects?

KVG: Dgraph, Badger, and Ristretto will not only remain open source, but we will ensure these projects continue to grow. We’re going to ensure that our open source users have all the tools and features they need to be successful with Dgraph in production, regardless of where they choose to host it.

While known largely for Next.js, one thing I enjoyed while at Vercel was investing in a variety of OSS, oftentimes because it felt right regardless of the “ROI”. This is clearly a sustainable practice as shown by Meta with React and Google’s Aurora team.

Highly permissive open source is the best way for any developer-first product to be built and adopted. It’s critical to both our credibility as a developer-first company as well as the best way to share our view of what good looks like for the future of AI.

GC: I joined Dgraph Labs because I’m passionate about building great open-source data products. I’m excited to continue investing in Dgraph’s open-source projects as a part of Hypermode. Dgraph will be a core part of how Hypermode’s platform operates. Dgraph now has a well-funded commercial entity whose success is predicated on Dgraph being a great database.

What happens to the Dgraph Labs Team?

GC: I will be joining Kevin as the CTO of Hypermode along with the Dgraph Labs team. We’re believers in Hypermode’s vision of easy, safe, and performant AI for the web. We will continue our focus on making Dgraph the best database for applications.

KVG: With Dgraph being a critical part of Hypermode’s infrastructure— Gajanan and I both wanted to retain as much of the team as possible. There are some truly exceptional engineers who will be joining Hypermode!

What happens to Dgraph Cloud?

KVG - Dgraph and Dgraph Cloud are mission critical to our users. We take that responsibility very seriously. We’re going to ensure that all users have a paved pathway that ensures their workloads continue operating without downtime. As is the responsibility of all critical infrastructure providers, changes we eventually make will be made very slowly with an abundance of communication.

We hope to see even more users from our community come over to Dgraph Cloud to explore all the new functionality we’re adding.

GC - There are hundreds of users signing up for Dgraph Cloud every month, we will continue to serve those teams. We’ve been investing behind the scenes into more scalable and reliable infrastructure over the last several months.

Where can I learn more?

Learn about Hypermode and request early access. Kevin and Gajanan will also host a livestream on Wednesday, November 8 at 10am Pacific. To join, please register below or subscribe on YouTube.

This is a companion discussion topic for the original entry at https://dgraph.io/blog/post/20231106-dgraph-part-of-hypermode
4 Likes
2

Exciting. A big part of why I picked DGraph was that I knew that a knowledge base structure would unlock a lot of possibilities when combined with AI, however, I expected I’d eventually have to build a very messy pipeline that would move data in and out of DGraph in order to perform predictions. Really excited about the prospect of this all being automated and streamlined.

Questions:

  • Am I right to assume that DGraph will eventually become one of Hypermode’s features, and that Hypermode will be an AI-backend, or will they remain separate products that just happen to work really well together?
  • Does this change the vision for DGraph and how it works? i.e. will GraphQL and DQL remain the primary interfaces for data in DGraph databases?
  • Will Hypermode eventually work with other databases?
2 Likes
3

Thanks for the support Ben!

In the long run, it’s probably somewhere in between.

We’ll be integrating Dgraph heavily within Hypermode, we think one of the best ways to create a great developer experience is through tight integration between the layers of the stack. For many users, they might not even know they’re using Dgraph.

That said, we expect that many teams with want to access the underlying data to power other non Hypermode applications. As such, we’ll also expose a hosted version of Dgraph. Those who currently rely on us for hosting will have a home with Hypermode both today and in the future.

Dgraph (and Badger and Ristretto) will remain permissively licensed. We’ll keep improving and investing in it. The goal is to over time make the Dgraph community more like PostGres/Cassandra/React where there’s a team of maintainers at a few commercial entities but the community is also able to contribute/steer to the core project.

As a “customer” of Dgraph we’re going to be leaning into GraphQL. Though our secret sauce will be the inference engines that will abstract the backend into “chains” of functions that resolve whatever intelligence your app needs.

Regarding DQL, there are just some things that you can’t express via GraphQL, as such DQL will remain supported as means to execute those queries.

There’s a much much much longer answer I owe the community about where we want to take the database, but regardless we’ll partner with you all to make sure that folks can get what they need out of the database.

Hypermode has an “open platform” mindset towards infra, models, databases, etc. We plan to make public APIs between each layer of Hypermode to make it easy for folks to map their stacks into our workflow. It’ll be limited to start, but over longer periods of time developers should be able to zero-config integrate other databases/data sources along with models, ml ops tools, etc.

2 Likes
4

I have a few questions:

  • Could you clarify a little further on how the developer experience pieces would change? Is Hypermode going to serve as a completely new product or will these features you are building be on top of Dgraph Cloud?

  • Is the plan to use lightweight vectorDB to support AI use-cases OR is the plan to use Dgraph w/ those vector stuff that was written in a previous blog?

  • The current website link to hypermode.com is a little confusing, could you clarify the vision there?

  • I would love a post or a discussion on your proposed architecture diagram that is available on hypermode.com. I am not sure if it is intentionally vague or if it is a WIP. Either way it is exciting to see this direction for Dgraph.

  • Could you please clarify where is the Vector work happening?.. it appears this is closed-source. Is the company switching to a closed-source model?

Some logistic questions

  • Is Hypermode also a new business? How long has it been operating? (from KVG’s linkedin it appears it is recent). Is this more of an acquisition to power the next gen AI apps quickly for Hypermode?
3 Likes
5

The foundation of Dgraph was for a “Distributed” Graph database. Is hypermode going to keep distributed as a key focal point?

That being the focal point has kept Dgraph’s head under water most of its life. Instead of adding basic features to make your data more queryable, the focus was on overoptimization on the distributed factor, and depending on clients to add their own real security and optimization layers.

GraphQL was added after users complained enough that while Dgraph claimed native GraphQL support the GraphQL± (now DQL) was not really GraphQL compliant. Enter a (still) half-baked GraphQL API with glaring security wholes and missing basic features and a renaming of the old to DQL. This just added complexity to the codebase and gave a false sense of performance and security to its users.

What we really want to know is twofold:

  1. Will Dgraph ever see any of the promised missing features?

  2. Will Dgraph’s glaring security wholes ever be fixed?

4 Likes
6

I wanted to express my gratitude to both the Dgraph team and the vibrant community here. The steadfast support and innovative spirit have always been the wind in the sails for Dgraph’s growth. :pray:t5: :heartbeat:

With the recent announcement of Dgraph Labs joining Hypermode, I’m thrilled about the potential for AI and its applications in serving communities through technology.

As someone deeply invested in integrating AI and LLMs with Dgraph, I’m curious about the path forward. The integration has been challenging, but now, with Hypermode’s involvement, it seems we’re at a crossroads.

For developers like myself, who are midway through fusing AI with Dgraph, what’s our best course of action? Do we continue with our current development strategies, or is it wiser to pause and see how Hypermode streamlines AI integration?

Timing is crucial. We’re seeking guidance on when Hypermode’s solutions will be ready for Dgraph developers as we aim to optimize our efforts and avoid redundant work. It would be beneficial to understand the timeline for these new integrations, ensuring our team’s work aligns with the upcoming advancements.

Thank you for any insights you can provide, and here’s to a future where AI and Dgraph drive innovation together! :muscle:t5:

3 Likes
7

Glad to hear that this is the direction you want to go. Please have a look at the backup/restore feature. It is one of the mayor roadblocks to seriously adopt Dgraph in a locally hosted manner. Backup/restore was once enterprise-only, then open-sourced and now again only available with an enterprise key.

For more information have a look at this thread: Backup restore documentation

4 Likes
8

Hi Everyone-- we tried to cover all of your questions here in the Community Forums stream. Here’s a link to the stream for those interested: https://www.youtube.com/watch?v=Bnnews6rlpY

I’ll also try to go through and answer the above directly later today.

Thank you all for you engagement and excitement as the team makes this transition. I’m so excited to share all the stuff we’re building with you all. We’re working through the Beta Signups list as fast as we can!

2 Likes
9

Highly permissive open source is the best way for any developer-first product to be built and adopted.

Hello! I initially thought this was very clear, but then I remembered the past few years in open-source and had some doubts. Can you please define what you mean by open-source? Is your definition of open-source the same as the OSI definition? I understand you can’t know the future, but would whatever license you choose in future releases likely be within the following list?

Thank you.

1 Like
10

Looking forward to your thoughts @KVG on the above questions.

Another question - How do we get to talk to someone from Hypermode for early access?

1 Like