VisionRepo_Engine

InnovateX Hackathon Project

Day 1 – Project Setup & Initial AI Integration

Overview

Day 1 focused on establishing a solid foundation for both the frontend and backend. The goal was to set up the core tech stack, ensure UI consistency, and integrate the AI layer with a working API endpoint.

---

🎨 Frontend Setup (React + Vite)

Project Initialization

• Initialized a React + TypeScript project using Vite
• Verified development server and build configuration

UI Library Setup (shadcn/ui)

• Installed and configured shadcn/ui
• Configured path alias (@/) across:
  • tsconfig.json
  • tsconfig.app.json
  • vite.config.ts
• Added and verified core UI components:
  • Button
  • Card
  • Input
  • Badge

Routing

• Set up React Router
• Created a base routing structure for future pages

Branding & UI Polish

• Applied initial branding decisions
• Implemented background design for a clean, modern UI using AI

---

🧠 Backend Progress (AI Integration)

Server Setup

• Installed required backend dependencies
• Initialized and started the backend server successfully

AI Integration

• Designed a structured AI prompt
• Created a dedicated AI API endpoint
• Integrated AI logic into the backend
• Tested the AI endpoint successfully using Postman

---

✅ End of Day Status

• Frontend foundation established
• UI system configured and rendering correctly
• Backend server running
• AI integration functional and tested

Total Time Spent on day 1 is 5.5 hours

Day 2 – Backend Architecture & Prisma Setup

Overview

On Day 2, the focus was on setting up the backend foundation for VisionRepo using Prisma with MongoDB and understanding proper backend architecture.

Key Learnings

1. Database & ORM Decision

• Chose MongoDB + Prisma
• Avoided using Mongoose alongside Prisma
• Learned why only one data access layer should exist

2. Prisma Setup

• Initialized Prisma
• Configured MongoDB as the datasource
• Used prisma generate and prisma db push
• Understood why Prisma migrations are not used with MongoDB

4. Service Layer Responsibility

• Services define business logic
• Services should never execute logic on import
• Controllers decide when to call services

5. Dashboard Service Logic(took 40 min to learn how to create and to know how dashboard works xd)

Implemented core service functions:

• getDashboardMetrics()

  • Aggregates repositories and analyses
  • Computes totals and averages for dashboard cards

• getRepositoriesHealth()

  • Returns per-repository health data
  • Acts as a controlled data exposure layer

Status

• Prisma setup completed
• Service layer logic completed
• Ready to implement controllers and routes next

Total Time Spent on day 2 is 4 hours

Day 3 – Backend Stabilization & API Debugging

1.Backend & Environment Recovery

• Faced unexpected backend breakage due to Node.js reinstallation
• Switched to Node.js v20 (LTS) to restore a stable runtime environment
• Reinstalled dependencies to align with the new Node version

2.Prisma & Database Fixes

• Encountered Prisma client initialization failures
• Identified issues caused by:
• Missing url in schema.prisma
• Incorrect usage of prisma.config.ts
• Removed prisma.config.ts and used schema.prisma as the single source of truth
• Downgraded Prisma to v5.22.0 (stable)
• Successfully regenerated Prisma Client
• Verified MongoDB connection and database access

3.Express Routing & Middleware Debugging

• Fixed route conflicts where generic /api routes intercepted /api/dashboard routes
• Reordered route registration to ensure specific routes are mounted before generic ones
• Resolved hanging requests caused by middleware not calling next() or returning a response

4.Dashboard APIs (Working & Verified)

• Implemented and tested:
• GET /api/dashboard/metrics
• GET /api/dashboard/repositories

5.Confirmed end-to-end flow:

• Express → Controller → Service → Prisma → MongoDB
• Successfully tested APIs using browser and Postman
• Analyze Endpoint (/api/analyze)
• Endpoint wiring and request validation implemented
• Request reaches controller and service correctly

6.main problem

-Encountered 500 errors originating from Gemini AI integration after environment reset -Issue identified as AI model / environment sensitivity, not routing or backend logic

7.What Was Fixed Today

• Node.js version stabilized (v20 LTS)
• Prisma configuration and client generation fixed
• Express routing and middleware issues resolved
• Dashboard APIs fully operational and tested
• What Needs to Be Fixed Next
• Finalize stable Gemini AI model selection
• Improve defensive JSON parsing for AI responses
• Enhance error transparency for /api/analyze

Total Time Spent on day 3 is 6 hours

✅ What We Accomplished DAY 4,5,6 (Day Summary) 🧠 Core Architecture

✅ Completed end-to-end analysis flow

GitHub repo URL → backend → GitHub parsing → Gemini AI → structured output

✅ Removed Prisma + MongoDB completely

No unused DB, no dead code

Cleaner, safer for deadline

📊 Dashboard & UX

✅ Dashboard now renders FULL AI analysis

Repo metadata

Tech stack

Components

Architecture flow

Onboarding steps

Missing context

Confidence score

✅ Fixed invisible text issue (CSS / text color)

✅ Dashboard survives refresh using localStorage

🕘 History & Persistence (Temporary but Solid)

✅ Added temporary save using localStorage

last_analysis

analysis_history (last 10 reports)

✅ “View Dashboard” now works even after refresh

✅ Added Clear History functionality

With confirmation warning to prevent accidental deletion

⏳ Loading & Feedback

✅ Added proper loading UI

Disabled button

Spinner

Full-screen loading overlay

Clear explanation of what’s happening

🧹 Cleanup & Polish

✅ Removed unused dashboard services/APIs

✅ Removed Prisma/Mongo configs & dependencies

✅ Fixed TypeScript errors (repo scope issue, guards, etc.)

✅ Kept landing page preview UI and repurposed it correctly

✅ Analyze page input validation fixed

Total time spent from Day 4 - 6 is 11 hours