🇪🇸 Spanish AI Exercise Maker

AI-powered fill-in-the-blank Spanish exercises that run entirely on your machine.

Type the Spanish words you want to practise, get an AI-generated exercise, submit your answer, and receive instant feedback — no API keys, no internet connection required after setup.

Features

• 📝 Fill-in-the-blank exercises generated from your own vocabulary list
• 🤖 Self-reviewing AI — the model checks its own output before showing it to you
• ✅ Instant answer grading with grammar tips
• 🔒 Fully local — powered by Ollama, nothing leaves your machine
• ⚡ Single command setup

Demo

Words:  hablar, comer, dormir, vivir

Exercise:  "Ella suele ____ fruta por la mañana."
Your answer: comer
Result: ✅ Correct! Great job.

Getting started

Prerequisites

• Node.js 18+
• Ollama installed and running

1. Pull a model

ollama pull llama3

Any model works. Other good options: mistral, phi3, gemma2.

2. Clone and install

git clone https://github.com/afiren/spanish-ai-exercises
cd spanish-ai-exercises
npm install

3. Configure environment

cp .env.example .env.local

Edit .env.local if you want a different model or Ollama URL:

OLLAMA_URL=http://localhost:11434
OLLAMA_MODEL=llama3

4. Run

npm run dev

Open http://localhost:3000.

How it works

You enter words
      ↓
POST /api/generate
      ↓
Ollama generates a fill-in-the-blank sentence
      ↓
Second Ollama pass reviews grammar and blank placement
      ↓
You answer in the UI
      ↓
POST /api/check
      ↓
Ollama grades your answer with an explanation

Project structure

app/
  page.tsx                 # Full UI — input → exercise → result
  layout.tsx               # Root layout
  globals.css              # Design tokens
  api/
    generate/route.ts      # Generates exercises with self-review
    check/route.ts         # Grades answers
lib/
  ollama.ts                # Ollama client + all prompt templates

Configuration

Variable	Default	Description
OLLAMA_URL	http://localhost:11434	Ollama server URL
OLLAMA_MODEL	llama3	Any model you have pulled locally

Roadmap

• v0.1 — Fill-in-the-blank, Ollama, self-reviewing prompt
• v0.2 — Multiple exercise types (translation, multiple choice, sentence builder)
• v0.3 — Session progress and score tracking
• v1.0 — Shareable exercise sets via URL

Contributing

Pull requests are welcome! See CONTRIBUTING.md for how to add a new exercise type or LLM provider.

Author

Made by @afiren

License

MIT