If you know me, you know I spend my weekends doing exactly what I do during the week - building.
This past weekend, I decided to have a little fun and wire up my Obsidian vault to a custom AI interface I built in Python. I affectionately named it J.A.R.V.I.S - MARK XLVII.
No, I’m not really Tony Stark, but when you’re deep in the AI rabbit hole, it’s hard not to lean into the aesthetic.
The Setup
The goal wasn’t just to make something that looked cool (though the audio waveform visualization definitely helps). The goal was to create a recursive learning loop directly on top of my second brain.
Here’s what you’re seeing in the video:
- The Interface (Left): A custom Python GUI running my local J.A.R.V.I.S instance. It’s built to handle audio input/output and display system logs in real-time.
- The Vault (Right): My live Obsidian vault, specifically the Graph view. You can see the nodes and connections - the raw data of my thoughts, projects, and AI architectures.
The Workflow
I gave Jarvis a simple verbal prompt: “Can you analyze my Obsidian vault and give me a summary of the main topics and how they connect?”
While I’m watching the Graph view, Jarvis is running a background process, crunching through 23 specific items across multiple directories in the vault. It’s extracting themes, mapping connections, and synthesizing insights.
The real magic here isn’t the voice interface; it’s the human-in-the-loop (HIL) execution. I’m not letting the AI run wild. I’m directing it to analyze my own structured data, waiting for the synthesis, and then I can act on those insights.
When it’s done, I tell it to carry on and put something chill on YouTube. Because even AI architects need a soundtrack while they work.
Why This Matters
This is a micro-example of what we’re building at scale with the Logoclothz OS and the RPG Content OS. It’s about taking raw, unstructured data (or in this case, a semi-structured markdown vault) and using AI to create immediate, actionable intelligence.
It’s not about replacing the human; it’s about giving the human a better lens to view their own data.
Credit where it’s due: The J.A.R.V.I.S interface you’re seeing is built by FatihMakes - specifically his Mark-XLVII project on GitHub. It’s the newest and most advanced model in his Jarvis series, and it’s genuinely impressive work. 333 stars and counting for a reason. Go give him a follow.
I’m running it locally and layering my own integrations on top - Obsidian vault analysis, custom prompts, and some additional tooling. But the core interface and architecture? That’s Fatih’s build. Full credit.
Related reading:
- From Freelancer to AI Architect
- Human-in-the-Loop AI: Why I Never Remove the Human
- AI Infrastructure Lessons Learned
Back to the lab.
Find me across the web
Stay curious, my AI friend. It's the secret sauce - think like you are seven. - Ryan