DEV Community: Ishaan

I got tired of getting lost in 10,000-file codebases, so I built a 3D visualization engine.

Ishaan — Thu, 18 Jun 2026 09:31:10 +0000

Whenever I join a new project, reading the codebase top-to-bottom feels impossible. Code isn't linear. It's a massive, tangled web of hundreds of files, circular dependencies, and hidden logic.

I wanted a way to just "fly through" an architecture to see how everything connects.

So over the last few weeks, I built Software MRI — an open-source, 100% local Electron app that parses directories using Abstract Syntax Trees (ASTs) and renders a massive 3D force-directed graph.

Here is how I built it, and the technical hurdles I had to overcome.

🏗️ The Architecture (Under the Hood)

Software MRI uses a chunked, multi-layered static analysis pipeline.

Because I built this in Electron, I was able to split the workload between the Node.js backend and the Chromium frontend.

The Parser (Main Process):
I use tree-sitter (compiled to WebAssembly) to parse the code of over 300+ programming languages locally. It rips through the codebase, extracting every import, export, and function definition without ever sending your proprietary code to a cloud server.
The Disease Analyzer:
Once the AST is built, I run basic graph traversal algorithms (like Tarjan’s strongly connected components) to detect "diseases"—things like circular dependencies or massive cyclomatic complexity spikes.
The Physics Engine (Renderer Process):
This is where things got tricky.

🏎️ Overcoming the UI Freeze

The biggest challenge I faced was the UI freezing when analyzing massive monolithic repos (like VS Code or the Linux Kernel).

If you try to run a d3-force-3d simulation on 10,000 nodes on the main thread, the browser window will completely lock up.

To solve this, I had to push the entire physics simulation into a background WebWorker. The WebWorker calculates the coordinate ticks in the background and streams the updated [x, y, z] positions back to the main thread in chunks.

🌌 Aggressive Level-of-Detail (LOD)

Even with WebWorkers handling the physics, rendering 10,000 text labels and 50,000 connecting lines in Three.js will melt your GPU.

I implemented an aggressive custom LOD (Level of Detail) camera system:

Zoomed Out: Files render as single instanced points (handling 10K+ nodes at a buttery smooth 60fps).
Flying In: As the camera gets closer, the points resolve into 3D spheres.
Close Up: Text labels appear, and the actual code reveals itself.

On top of this, the "diseased" files (the ones with circular dependencies) are given a custom glowing red WebGL shader so you can spot architectural flaws visually from a mile away.

🚀 Try it yourself!

I open-sourced the entire project today. You can download the Windows .exe directly from the Releases page, drag-and-drop your biggest local repository into it, and watch it come alive!

💻 GitHub Repo & Download: Ishaan-Sharma-tech/software-mri
🌐 Live Website: Software MRI Landing Page

I would absolutely love to hear your feedback on the architecture, or if you have any ideas on how I can optimize the AST parsing even further! Drop a comment below. 👇

Visualizing Database Design: From Interactive Canvas to Drizzle, Prisma, and SQL in Real-time

Ishaan — Sat, 23 May 2026 08:46:47 +0000

Every full-stack developer has been there: you are planning a new
feature, sketching out database tables, and spending half your time
looking up ORM syntax for many-to-many relationships or writing join
tables by hand.
I wanted a tool where I could visually draw my relational tables,
connect the columns, and immediately get clean, copy-pasteable code for
the three targets I use most: PostgreSQL SQL, Prisma ORM, and
Drizzle ORM.
So, I built SchemaDraw(https://schemadraw.netlify.app).
---

### The Architecture: Designing for ₹0 Scale

As an indie hacker, I wanted to launch a tool that could support

thousands of developers without leaving me with a heavy server bill.

To achieve this, SchemaDraw is built as a **100% client-side Single

Page Application (SPA)** using React and Vite:

1. **Client-side Compilation:** The engines that transform your

visual nodes into SQL or TypeScript are entirely written in JavaScript,
running in the user’s browser.
2. Local Storage Persistence: Your work auto-saves locally so
refreshing the tab won't lose your schema.
3. LZ-Compressed URL Sharing: Instead of saving diagrams to a
server database, SchemaDraw compresses the entire canvas layout and
schema structure into a URL hash component. You can copy the share link
and send it to a teammate—the app reads the hash and redraws the exact
layout with zero server storage costs.
---
### Key Features Built for Developers

* **Protruding Target Handles:** Most ERD tools place connection dots

inside nodes, leading to click collisions. SchemaDraw places handles - 5px outside the borders, lighting up green on valid targets to make
drawing links effortless.
* Interactive Presentation Lock: Toggle read-only mode to safely
present or inspect diagrams without making accidental changes.
* Grid Auto-Layout: Got a messy canvas? One click aligns all your
tables into a clean, alphabetical column-row grid.
* Cohesive Theme Engine: A single click swaps between a clean
Light Slate and a cohesive Pitch-Black interface.

---

### Check it out!

SchemaDraw is open, free, and requires no accounts or signups.

👉 **Try it here:** 
  
  


I’d love to hear your thoughts! What features, databases, or ORM

compilers would make this a permanent part of your bootstrapping
toolkit? Let me know in the comments below!

Building a Privacy-First Autonomous Agent with Gemini and LiveKit: Introducing Ved AI 🤖⚒️

Ishaan — Tue, 21 Apr 2026 12:17:31 +0000

Intro Snippet: Hey Dev Community! I've spent the last few months building a personal assistant that doesn't just chat—it acts.

Today, I’m open-sourcing Ved AI. It’s a local-first autonomous agent that uses Gemini Pro for reasoning and LiveKit for lightning-fast voice interactions.

Why I built this: Most assistants today feel like silos. Ved is designed to reach into your workflow—Gmail, Drive, System Control—and get things done without your data leaving your control.

The Tech Stack:

Backend: FastAPI & Python
AI: Google Gemini & Groq
Voice: LiveKit Realtime API
Memory: Knowledge Graph (NetworkX)
I’d love for you guys to check out the source code, play with the "Setup-First" UI, and let me know your thoughts!

Ishaan-Sharma-tech / ved-ai

Next-generation Autonomous AI Assistant with Real-time Multi-modal Voice, Agentic Swarms, and full Google Workspace integration. Built for Performance & Privacy.

🚀 Ved AI: Your Advanced Autonomous Personal Agent

Ved is a next-generation AI assistant designed to handle your digital life through voice and text. Built with a "Privacy-First" approach, Ved runs locally and connects to cutting-edge AI models to provide a seamless, sharp, and highly capable agentic experience.

🌟 Key Features

🎙️ Real-time Multi-modal Voice: Powered by LiveKit and Gemini Realtime API for natural conversations.
🐝 Agentic Swarms: Spawns parallel agents for Research, Synthesis, and Data Retrieval.
🛠️ Integrated Toolset: Full control over your local System, Gmail, Google Drive, Google Tasks, YouTube, and the Web.
📱 Telegram Integration: Secure remote control of your agent from anywhere.
⚙️ Setup-First UI: Configure everything (Identity, Keys, Paths) directly from the Web UI.

⚡ One-Click Quick Start (Windows)

The easiest way to get started is using our automated launcher:

Clone the repo: git clone https://github.com/your-username/ved-ai.git
Double-click start_ved.bat: This will automatically
- Create your .env…

View on GitHub

My First AI Agent Post

Ishaan — Sat, 21 Mar 2026 10:56:22 +0000

This is a live test from my local Python AI.

Guy's see this

Ishaan — Thu, 12 Mar 2026 04:59:22 +0000

dev.to / ishaansharma/i-built-an-ai-that-roasts-your-github-profile-heres-how-i-made-it-in-one-day-2b94

I built an AI that roasts your GitHub profile — here's how I made it in one day

Ishaan — Mon, 09 Mar 2026 09:52:10 +0000

The Problem

Every developer wants honest feedback on their
GitHub profile but existing tools are either
too expensive, too complicated, or just boring.

So I built something different.

Introducing GitRoast AI 🔥

GitRoast AI uses Gemini AI to analyze your
GitHub profile and roast you for your coding
habits — brutally but lovingly 😂

Paste any GitHub username and instantly get:

🔥 A savage AI roast of your profile
📊 Developer score out of 100
💀 Ratings for commit consistency, repo quality, social game & profile setup
💡 Actionable tips to actually improve
🏆 Global leaderboard — Hall of Fame & Hall of Shame
🐦 One click share to X and Reddit

How I Built It

Tech stack:

Frontend: React + Tailwind CSS
Auth & Database: Supabase
AI: Google Gemini API (gemini-1.5-flash)
GitHub Data: GitHub Public API
Hosting: Lovable

The entire app fetches your GitHub profile
data using the free GitHub public API —
no authentication needed. Then sends all
your stats to Gemini AI which generates
a personalized roast and developer score.

The Fun Part

The roasts are genuinely hilarious 😂

Try roasting these famous developers:

torvalds — Linux creator
gaearon — React creator
sindresorhus — npm legend

Try It Free

👉 gitroast-ai.lovable.app

Also launching on Product Hunt today —
would love your support! 🙏

👉 producthunt.com/posts/gitroast-ai

Let me know what score you get in the
comments below! 🔥