I Built Shree AI OS — A Deterministic AI Runtime for Java 21
After several months of development, I’m releasing the Developer Preview of Shree AI OS, a Java-first AI runtime designed for building intelligent backend applications.
Most AI frameworks focus on chatbots. My goal was different: create a reusable runtime that developers can embed directly into Java applications with memory, planning, reasoning, and hybrid retrieval.
Why I built it
While building Spring Boot applications, I found that adding AI often meant stitching together multiple services. I wanted a deterministic runtime that provides:
- Hybrid RAG with PostgreSQL pgvector
- Episodic and semantic memory
- Planning and reasoning modules
- Multi-agent orchestration
- Privacy-first architecture
- Java 21 native SDK
Architecture
The platform follows a layered architecture:
- Application Layer
- SDK Layer
- Runtime Orchestration
- Kernel Services
- LLM Provider Layer
Every request flows through an explainable cognitive pipeline instead of directly calling an LLM.
Developer Portal
I also built an official documentation website using Next.js 16, React 19, Tailwind CSS 4, and Framer Motion.
It includes:
- Architecture documentation
- SDK reference
- Quickstart guide
- Security model
- Interactive playground UI
Tech Stack
- Java 21
- Spring Boot
- PostgreSQL + pgvector
- Next.js 16
- React 19
- Tailwind CSS 4
Feedback Welcome
This is still a Developer Preview, and I’d love feedback from Java and backend developers.
GitHub: github.com/darshanrathod04/shree-ai-os
Website: shree-ai-os.netlify.app
Thanks for reading!