PokeLearn 🎮

Learn by Talking. Talk to Your Pokémon.

Cursor Colombo Buildathon 2026 · Best Use of VALSEA Track

---

🧠 The Problem

Voice AI fails South Asian kids.

A child in Sri Lanka who says "Pikachu, what is photosynthesis ah?" gets gibberish back from Google, Siri, or Alexa — because these engines are trained almost entirely on Western English. Over 500 million school-age children across South and Southeast Asia are effectively locked out of voice-based learning tools, not because of ability, but because of accent.

✨ The Solution

PokeLearn lets kids pick any of 1025 Pokémon as their teacher and learn any subject by simply talking — in their natural accent, with their natural slang, in the way they actually speak.

VALSEA handles the voice layer. It understands Sri Lankan and South Asian accented English the way no mainstream speech engine does. Every question is transcribed accurately, every emotion detected, and every response adapted to how the child is feeling in that moment.

---

🎯 How It Works

Child speaks naturally
        ↓
VALSEA transcribes with accent awareness
        ↓
VALSEA sentiment detects emotion (confused? confident? frustrated?)
        ↓
OpenAI responds in character as the chosen Pokémon
adapted to the child's emotional state
        ↓
Every 3 questions → auto MCQ quiz with star rewards

---

🔥 Features

Feature	What it does
🎤 VALSEA Voice Transcription	Understands Sri Lankan and South Asian accented English accurately
💬 Sentiment-Driven Responses	Pokémon adapts tone based on VALSEA emotion detection
🐾 All 1025 Pokémon	Every Pokémon with a unique teaching personality + shiny toggle
🧪 Auto MCQ Quiz System	Quizzes generated every 3 questions with star burst rewards
📚 8 Subject Categories	Science, Math, English, Geography, History, Nature, Art, Technology — auto-detected
♿ Text Chat Fallback	Full accessibility for deaf users or noisy environments
⭐ Progress Tracking	Daily questions, stars, and streaks via localStorage
🛡️ Content Safety	All inputs filtered before reaching OpenAI. Child-safe on every call
🔐 Secure Deployment	API keys injected at Netlify build time via shell script — never in the repo

---

🛠️ Tech Stack

Layer	Technology	Why
Voice Transcription	VALSEA /v1/audio/transcriptions	Only API that handles South Asian accented English accurately
Sentiment Analysis	VALSEA /v1/sentiment	Drives emotional adaptation in every Pokémon response
AI Responses	OpenAI gpt-4o-mini	Fast, cost-efficient, strong character instruction following
Pokémon Data	PokéAPI	Free, open, all 1025 Pokémon with sprites and types
Frontend	HTML, CSS, Vanilla JS	No framework — zero build complexity, maximum speed
Hosting	Netlify	Free tier, automatic GitHub deploys, environment variable support
Secret Management	generate-config.sh	Writes API keys from Netlify env vars at build time

---

🚀 Running Locally

git clone https://github.com/dkrnr/pokelearn
cd pokelearn

# Create your config file (never commit this)
cp config.example.js config.js
# Edit config.js and add your API keys:
# CONFIG.VALSEA_KEY = "your_valsea_key"
# CONFIG.OPENAI_KEY = "your_openai_key"

# Serve locally
python3 -m http.server 8080
# Open http://localhost:8080

API keys needed:

• VALSEA — for voice transcription and sentiment
• OpenAI — for Pokémon character responses

---

🔐 Security

• API keys stored in Netlify environment variables
• generate-config.sh writes keys to config.js at build time
• config.js is .gitignored — never committed
• No user data collected or stored beyond anonymous localStorage progress
• Content safety filter on all inputs before any OpenAI call

---

📁 Project Structure

pokelearn/
├── index.html                    # App shell
├── style.css                     # Pokémon-themed UI
├── app.js                        # All app logic
├── pokemonPersonalities.json     # 1025 Pokémon personality prompts
├── config.example.js             # Safe template (no real keys)
├── generate-config.sh            # Netlify build script for key injection
├── netlify.toml                  # Netlify build config
└── README.md

---

👾 Easter Eggs

Find Ursaluna and ask who made this app. Find Vespiquen and ask who is it's trainer and more.

---

👥 Team

Name	Role
Dunith	Lead Developer — VALSEA integration, OpenAI prompt system, quiz engine, deployment
Ibaad	Frontend & Product — Pokémon selector, UI/UX, personality system, progress tracking, API integration
Zahrah	Lead Tester — QA Engineer, Analyst, Advisor, Editor, Feedback

Built at the Cursor Colombo Buildathon 2026 in 24 hours at Royal MAS Arena, Colombo, Sri Lanka.

---

📄 License

MIT — feel free to build on this.

---

Built with Cursor · Powered by VALSEA · Pokémon data from PokéAPI