💀 Inspiration
This project began back in 2021 , when I worked on automating the migration of ASP.NET Web Forms to ASP.NET Core and Angular for my final-year thesis using parse trees and template engines. That rigid approach required months of engineering per technology and was limited to a single tech stack.
So the question became: what if migration didn’t have to be rebuilt from scratch every time a technology disappears?
What if AI could understand any legacy stack and bring it back to life?
StackOverGrave is my attempt to answer that.
⚡ What it does
StackOverGrave takes old, abandoned codebases and resurrects them into modern, maintainable applications.
🎯 Core Features:
📦 Multi-Source Import
- Paste public GitHub/GitLab/Bitbucket URLs
- Upload ZIP files of legacy projects
- Automatic branch detection (main/master/develop)
🔍 Intelligent Analysis
- Detects technology: VB6, Flash ActionScript, Silverlight, old .NET Framework
- Scores files by criticality (entry points, models, business logic)
- Validates project size and complexity
🧙 AI-Powered Resurrection
- Small projects (≤20 files): Fully automatic conversion
- Medium projects (≤50 files): Top 15 critical files + detailed migration guide
- Context-aware conversion maintains consistency across files
- Generates modern .NET 8 C#, TypeScript/Angular, or Blazor
📁 Production-Ready Output
- Downloadable ZIP with proper project structure
- Comprehensive README with setup instructions
- Migration guide with find/replace patterns
🎃 Spooky UX
- Halloween-themed graveyard dashboard
- Tombstones crack open as files resurrect
- Real-time progress with ghost particles
🛠️ How we built it
Tech Stack:
- Frontend: Angular 17 (TypeScript) with spooky animations
- Backend: ASP.NET Core 8 Web API (C#)
- AI: OpenAI GPT-3.5-turbo for cost-effective conversions
- Storage: SQLite (in-memory for demo)
- Development: Kiro AI-powered IDE (95% code generation)
Architecture:
🌐 Frontend (Angular)
↓
🔧 GitRepositoryService → Downloads & extracts
↓
📊 RepositoryAnalysisService → Scans, scores, selects files
↓
🤖 AI ConversionService → Context-aware batch processing
↓ (for medium projects)
📖 MigrationGuideService → Generates detailed guide
↓
📦 PackagingService → Creates downloadable ZIP
↓
⬇️ User downloads resurrected project
Kiro's Role:
95% of code generated by Kiro in 3 days:
- ⚡ Vibe coding: Rapid UI prototyping - complete file upload system with animations in 15 minutes
- 📋 Spec-driven: Complex Repository Resurrection backend - 6 interconnected services in one pass
- 🎨 Steering docs: Halloween theme consistency - reduced manual styling from 40% to 5%
🔥 Challenges we ran into
💰 Token Budget Constraint: OpenAI costs could spiral out of control. We solved this with intelligent file selection (top 15 critical files only) and switching from GPT-4 to GPT-3.5-turbo, reducing costs from $8 to $3 per conversion.
🔗 Context Loss Between Files: Converting files independently produced inconsistent code. We built a stateful service that accumulates context—each file's prompt includes patterns from previous conversions, maintaining architectural consistency.
📏 Project Size Limits: Can't realistically convert 200+ files in 3 days. We turned this limitation into a feature: small projects get full conversion, medium projects get core files + expert migration guide.
🏆 Accomplishments that we're proud of
⚡ 5 Days → 15 Minutes: Kiro generated production-ready code (file upload, editor, animations, theme) in 15 minutes. What normally takes a week of manual coding.
🧠 Context-Aware Intelligence: Not just another ChatGPT wrapper. StackOvergrave learns from each file conversion to maintain architectural consistency across the entire project—a technical achievement requiring sophisticated prompt engineering.
🎨 Actually Enjoyable: Legacy modernization is usually tedious. We made it fun with cracking tombstones, ghost particles, and darkly humorous messaging. The Halloween theme transforms a boring task into an entertaining experience.
💡 What we learned
Kiro changed how we code. It's not about writing less—it's about thinking at a higher level. We focused on architecture and UX while Kiro handled implementation. The 95% code generation isn't laziness—it's force multiplication.
Real-world insight: My thesis taught theory; this hackathon taught reality. People don't need perfect conversions of 200 files—they need the 15 most critical files done perfectly, plus guidance for the rest. User needs > technical completeness.
Breakthrough discovery: AI isn't just for single-file generation. By maintaining context across conversions, we achieved consistency rivaling human refactoring. This pattern—stateful, context-aware AI workflows—extends far beyond code conversion.
🚀 What’s next for StackOvergrave: Where dead stacks rise again
Become the Stripe of code modernization - simple, powerful API that handles complexity behind the scenes. Every company with legacy code becomes a potential customer. The total addressable market is massive (Fortune 500 alone spend billions annually on maintenance).
StackOverGrave isn't just a hackathon project - it's the foundation of a real business solving a real problem.

Log in or sign up for Devpost to join the conversation.