All Posts

How GC advances changed the reality of running an MMORPG on the JVM.

A look at Omnilude Asset Studio's real implementation and generation flow.

Using the enhanced chatbot agent workflow as the basis, this post breaks down the routing, search, summarization, and response flow.

A summary of the production strategy and current implementation status from Story Quiz to Mystery to MMORPG.

One repository, but deployment is split by service. This is how I structured it.

In Omnilude's ai-service, `AiAgent` is not a separate AI but a workflow that connects assistants as nodes. This article explains the real execution structure through DTE, WorkflowEngine, and NodeExecutor.

Inspired by OpenAI's Assistants API, this post explains how Omnilude's ai-service implements an Assistant / Thread / Run structure.

Instead of repeatedly relying on external S3 just for tests, I built an on-prem S3-compatible storage stack with SeaweedFS and an Nginx CDN. The best part was keeping local and dev on-prem while still being able to swap production over to S3 or R2.

A look at how we split one backend project into an MSA and carried the change all the way to deployment.