gRPC

Am I only basically building a certain part of my application in Rust using Hydro, and I could build other parts of it using whatever language I choose?

Yeah, I mean, you know, keep in mind it's the distributed systems world, so you can always take a piece of your system and turn it into a service, and then the endpoints are gRPC or REST or something like that, right? They can be cross language.

So certainly that's always an option.

You can write a microservice or tear out a piece of your system and turn it into a service and write it in Hydro even though your service was written in something else.

But Whisper doesn't speak that protocol.

Image generation models use their own thing, typically something like the stable diffusion web, UI API, or a custom gRPC interface.

Embedding models usually speak OpenAI compatible too, but not always.

So even if you standardize on the OpenAI API format for LLMs, you still have protocol diversity across model types, and something needs to normalize that.

The obvious objection first: we already have APIs.

REST, gRPC, OpenAPI specs, decades of tooling.

Why does an AI agent need a new protocol to call a database or a ticketing system? The answer is that traditional API integration was designed for developers, and agents are not developers.

Three gaps matter in practice.

So really just about every use case, every place I can imagine.

Messaging, gRPC, Kafka, integration, um, you know, um, web apps, uh, you know, everything.

Everything you can...

MCP for AI, you know? There's always a reason to, to use security and to think about security, and there's even more reason to use the Spring Auth Server.

I mean, just I, I think the, to the extent that we can make OAuth a s- you know, for everybody, you know, like there's, there's so many use cases.

One of the, one of the things that's, the, one of the strengths and also maybe not weaknesses, but, um, hiccups or bumps in the road for people adopting OAuth is that OAuth 2 is a framework for security.

It's not just a prescriptive set it once and forget, you know? And so to the extent that I can take it to my desktop, my CLI, my, my web application, my gRPC application, you know, the pieces are all there.

And, uh, now y- with what you're describing, there's, it sounds like there's, in that same framework, there's these new options for even the core things of like, like doing the authentication itself can become more secure, right? And it's still consistent with the overall intent of OAuth, right? It just, it's expanding.

Google has actually been using this internal design guide since 2014, and it's practically a masterclass in establishing a predictable API ecosystem.

Interestingly, it applies to both their traditional REST APIs and their high-performance gRPC APIs.

Let's talk about gRPC because it's a completely different beast than the normal web APIs most people are familiar with.

It relies heavily on protocol buffers or proto buffs instead of the standard JSON payloads we usually see.

To understand why this matters, you really have to look at the mechanism.

Through an ingenious process called transcoding, think of it like a real-time UN translator at the edge of their network.

OK, I like that.

They use API service configurations to automatically map standard incoming HTTP and JSON requests directly into internal binary gRPC calls.

It means you fire it up, you can send a request to it, you can access it in the browser.

There is a lot of, there's a pre-request to add a gRPC, which is an RPC interface, to add a Postgres-compatible front end, They're all fabulous ideas.

And of course I've been tempted, and in the advent of LLMs, you'd be tempted to sit down, tell the agent, add this, I'll be back in four hours.

But that is going to add a whole new operational element to it that is part of the reason I have resisted.