CODEMIND — AI Coding Mentor

Train Smarter. Memory Enhanced. An AI-powered coding platform that remembers every mistake and crafts personalised hints, feedback, and problems tailored to your weak areas.

---

What is CodeMind?

Most coding platforms treat every session as isolated — they forget you the moment you close the tab. CodeMind is different. It builds a persistent memory profile using the Hindsight vector memory engine and uses that history to power every hint, feedback, and AI-generated problem you receive.

The result: an AI mentor that genuinely gets smarter the more you use it.

---

Features

AI Problem Generator

• Generates unique coding problems tailored to your weak areas
• Powered by AWS Bedrock — Claude Haiku
• Select topic (Arrays, Strings, Loops, Linked Lists, Dynamic Programming, Stacks)
• Select difficulty (Easy / Medium / Hard)
• Includes title, description, examples, test cases, and starter code in Python, JavaScript & Java

Memory-Powered Hints

• Before/After Memory toggle — compare personalised vs generic AI responses side-by-side
• Memory ON: Hindsight recalls past mistakes relevant to the current problem and warns proactively
• Memory OFF: standard generic hint with no context

Code Editor & Execution

• Monaco Editor — the same editor used in VS Code
• Python, JavaScript, and Java support
• Real-time execution against test cases on the server
• Per-test-case results with expected vs actual output
• Anti-cheat: tab switch detection + countdown timer

AI Submission Feedback

• Mistake classification: null_check, index_error, time_complexity, syntax_error, edge_case, logic_error
• Three-metric AI score: Correctness / Efficiency / Style
• Feedback stored back into Hindsight memory for future personalisation

Memory Profile

• INSIGHTS tab: weak areas, common mistakes, streak, AI recommendation
• HINDSIGHT tab: cloud memories stored via Hindsight API
• HISTORY tab: local session attempt history with timestamps

Leaderboard

• Global ranking by score (10 pts per problem + accuracy bonus)
• Live stats: problems solved, accuracy %, sessions

---

Tech Stack

Layer	Technology
Frontend	React 18 + Vite
Code Editor	Monaco Editor
Styling	Custom CSS3 + Tailwind
Backend	Python + FastAPI
AI	AWS Bedrock — Claude Haiku
Memory	Hindsight API
Code Runner	Python asyncio subprocess (multi-language)
Frontend Hosting	Vercel
Backend Hosting	Render

---

Getting Started

Prerequisites

• Python 3.11+
• Node.js 18+
• AWS account with Bedrock access enabled (us-east-1)
• Hindsight API key from hindsight.vectorize.io

Backend Setup

cd backend
pip install -r requirements.txt

Create backend/.env:

AWS_BEDROCK_API_KEY=your_bedrock_key
AWS_ACCESS_KEY_ID=your_access_key        # alternative auth
AWS_SECRET_ACCESS_KEY=your_secret_key    # alternative auth
AWS_REGION=us-east-1
HINDSIGHT_API_KEY=your_hindsight_key
HINDSIGHT_BASE_URL=https://api.hindsight.vectorize.io
HINDSIGHT_BANK_ID=CodeMind

Run the backend:

uvicorn main:app --reload
# Starts on http://127.0.0.1:8000

Frontend Setup

cd frontend
npm install
npm run dev
# Starts on http://localhost:5173

---

Deployment

Backend → Render

Setting	Value
Root Directory	backend
Runtime	Python 3
Build Command	pip install -r requirements.txt
Start Command	uvicorn main:app --host 0.0.0.0 --port $PORT

Add all environment variables in Render → Settings → Environment.

Frontend → Vercel

Setting	Value
Root Directory	frontend
Framework	Vite
Build Command	npm run build
Output Directory	dist

Add environment variable in Vercel → Settings → Environment Variables:

VITE_API_URL=https://your-render-url.onrender.com

---

API Reference

Auth

Method	Endpoint	Description
POST	/signup	Create account
POST	/login	Login
POST	/guest	Guest session

Problems

Method	Endpoint	Description
GET	/problems	Get problem library
POST	/generate-problem	AI-generate problem for user
GET	/daily-challenge	Today's daily challenge

Execution & Feedback

Method	Endpoint	Description
POST	/run-code	Execute code against test cases
POST	/submit	Submit + get AI feedback
POST	/hint	Get memory-powered or generic hint
POST	/optimize	Code optimization analysis

Memory & Analytics

Method	Endpoint	Description
GET	/learning-insights/:user_id	AI learning insights
GET	/memory/:user_id	Hindsight memories + local history
GET	/leaderboard	Top 20 leaderboard
GET	/status	API health check

---

How Memory Works

Session 1: User struggles with null checks in arrays
→ Backend stores: "Failed Two Sum | mistake: null_check | language: python"
   via Hindsight retain()

Session 2: User attempts a new arrays problem
→ Backend calls Hindsight recall()
→ Returns: past null_check pattern
→ AI hint: "⚠️ Based on your history, you often miss null/empty checks —
            verify nums is not None before accessing indices."

The Before/After toggle in the navbar lets you instantly switch between memory-enhanced and generic AI responses to see the difference.

---

Project Structure

CodeMind/
├── backend/
│   ├── main.py              # FastAPI app — all routes and business logic
│   ├── hindsight_client.py  # Hindsight API wrapper (retain, recall, list)
│   ├── problems.json        # Curated problem library (6 problems)
│   └── requirements.txt
└── frontend/
    ├── src/
    │   ├── pages/
    │   │   ├── Dashboard.jsx        # Problem list + AI generator
    │   │   ├── ProblemSolver.jsx    # Editor + execution + hints
    │   │   ├── MemoryProfile.jsx    # Insights + history + memories
    │   │   ├── Leaderboard.jsx      # Global rankings
    │   │   └── AuthScreen.jsx       # Login / signup / guest
    │   ├── api.js           # Centralised API helper (uses VITE_API_URL)
    │   ├── App.jsx          # Root — nav, memory toggle, auth state
    │   ├── icons.jsx        # Custom SVG icon set
    │   └── index.css        # Dark theme, animations, component styles
    ├── vite.config.js
    └── package.json

---

Live Demo

codemind-tau.vercel.app

• Click Continue as Guest to explore without creating an account
• Try the AI Generator — select a topic and click Generate
• Solve a problem and click Smart Hint — toggle memory ON/OFF to see the difference
• Submit a solution to get AI feedback and a score

---

Future Scope

• Persistent database (MongoDB Atlas / Supabase) to replace in-memory stores
• JWT authentication with refresh tokens
• C++, Rust, Go language support
• 100+ problem library with adaptive difficulty
• Streaks, badges, and daily goals
• Mobile app (React Native)
• Multi-language UI (Telugu, Hindi, Tamil)

---

Built with AWS Bedrock + Hindsight Memory | Deployed on Vercel + Render