Johannes Rabauer
@rabauer.bsky.social
66 followers 160 following 140 posts
☕ Java coder @xdev.software | 🤖 AI wrangler | 🎤 Conference speaker | 🎮 Full-time dad, part-time gamer
Posts Media Videos Starter Packs
Reposted by Johannes Rabauer
quarkus.io
Securing LLM Responses in Java: Guardrails with Quarkus and LangChain4j
Learn how to protect your AI applications from prompt injection, unsafe outputs, and risky prompts with practical guardrail patterns in Quarkus.

www.the-main-thread.com/p/java-quark...
Securing LLM Responses in Java: Guardrails with Quarkus and LangChain4j
rabauer.bsky.social
It looks like you changed passwords since the leak: I tried to login to localhost:8080 with password "pw", but access was denied 😂

Good for you! :)
rabauer.bsky.social
Quanta tracks file changes, re-vectorizes updates, and returns instant, intelligent answers; all locally.
We debugged Quarkus quirks, Tika headaches, and got Ollama humming 💪
youtube.com/live/OenMqG5...
youtube.com
rabauer.bsky.social
Under the hood:
⚡ Next.js (frontend)
☕ Quarkus + LangChain4j (backend brains)
🦙 Ollama (LLaMA 3.2 embeddings)
🐘 PostgreSQL + pgvector
📄 Apache Tika for text extraction from any file type
rabauer.bsky.social
Just went live building Quanta: an AI-powered file searcher that understands your files 🤖📂
It watches your filesystem, vectorizes docs, and lets you query them in natural language.
Let’s make local AI search actually useful 👇
rabauer.bsky.social
Favorite input field 😅
rabauer.bsky.social
Unfortunately because of an emergency in the family i have to postpone the session again...i am really sorry.
rabauer.bsky.social
Thank you very much! :)
rabauer.bsky.social
Happy birthday Catherine! :)
rabauer.bsky.social
Thanks! I am unsure about PostgreSQL with pgvector, but since i have the most experience with it, i'm willing to try ^^
Why did you choose PostgreSQL, if you did?
rabauer.bsky.social
To make this work, I’m combining:
@nextjs.org frontend
@quarkus.io + @langchain4j.dev backend
⚡ Ollama (LLaMA 3.2) for AI
⚡ PostgreSQL + pgvector for embeddings

A lot of moving parts → one AI-powered file explorer.
rabauer.bsky.social
Quanta will:
📂 Scan your filesystem
🗂️ Catalogue files automatically
📝 Summarize their contents
🔍 Let you find them later with natural language search

No more “where did I save that doc?” moments.
rabauer.bsky.social
After building an AI storytelling app for fun…
…it’s time to create something actually useful. 🚀

Meet Quanta – an AI file searcher.
Reposted by Johannes Rabauer
oracleace.bsky.social
Excited to welcome Johannes Rabauer as our newest Oracle ACE Associate! 🚀
Johannes crafts secure, scalable Java solutions. Cloud, AI, frameworks & clean code—he’s passionate about making developers’ lives easier.
Learn more: ace.oracle.com/ords/ace/pro...
#OracleACE
rabauer.bsky.social
Just unlocked a new achievement: 🏆 Oracle ACE Associate!
Big thanks to the Oracle community 🙌
Catch me dropping Java wisdom at JCON or on YouTube 😂
#OracleACE #Java #DevCommunity @oracleace.bsky.social
A smiling man wearing glasses, a black t-shirt, and a lanyard points to an Oracle ACE Associate certificate. The certificate is awarded to Johannes Rabauer in recognition of technical and community leadership contributions to Oracle communities from July 2024 to August 2025. It is signed by Gerald Venzl, VP of Database Product Management, ACE Program, and dated September 2025. In the background, there is an illustration of a golden trophy with a spade symbol.
rabauer.bsky.social
Part 10: Agents came alive in LangChain4J + Ollama 💡
Part 11 (Sat 20:00 CET): Can we give the entire system memory that persists in PostgreSQL?
One landscape, many agents, real continuity.
Join the experiment 👉 youtube.com/live/Jpid86_...
#Java #AIAgents
Can We Give Our AI Agents Real Memory? (LangChain4J + Ollama + Java Live Coding)
Part 11 – Giving Agents Real Memory in LangChain4J (Live Java Coding with Ollama)In this live session, we’ll tackle one of the hardest (and most exciting) ch...
youtube.com
rabauer.bsky.social
Highlights with timestamps:
00:52 Fixing persistence
12:41 Fighting LLaMA 3.2:1B
29:00 Google ADK adventures
49:50 Agent prompts
01:13:00 An actual response (finally 🎉)

Full replay 👇
youtube.com/live/xppcyTU...
youtube.com
rabauer.bsky.social
Next session = me going all-in on LangChain4J native agents (no extra SDKs this time).

If you’re a Java dev curious about AI storytelling, Ollama, LangChain4J, or just enjoy watching someone debug themselves into a corner:
👉 Watch the chaos, link below
rabauer.bsky.social
So yeah, Google ADK looked like a shiny enterprise hammer 🛠️…
…but LangChain4J already had the screwdriver I actually needed. 🔧

Lesson learned.
rabauer.bsky.social
It did *kind of* work… but super strangely.
Like, “AI gremlin whispering from the shadows” strangely.

After more head scratching than I’d like to admit, I realized:
👉 I didn’t need ADK at all.
rabauer.bsky.social
Then came my *brilliant* idea: “Let’s integrate Google’s ADK with LangChain4J. It looks so professional. Surely this is the way.”

Spoiler: it was not the way. 😅
🔗 google.github.io/adk-docs/
Agent Development Kit
Build powerful multi-agent systems with Agent Development Kit
google.github.io