Language AI App

AI-powered language practice app with multi-provider LLM backend (OpenRouter, Ollama, Runware, Fal.ai).

Run locally

1. Create .env in project root:

PORT=3000
PROVIDER=openrouter # or ollama

# OpenRouter
OPENROUTER_API_KEY=your_key
OPENROUTER_MODEL=anthropic/claude-3.5-sonnet
APP_URL=http://localhost:5173

# Ollama local
OLLAMA_HOST=http://127.0.0.1:11434
OLLAMA_MODEL=qwen2.5:14b

2. Install and run:

npm i
npm run dev

• Frontend: http://localhost:5173
• Backend: http://localhost:3000

Build and run production

npm run build
npm start

Serves built frontend from the backend on port 3000.

Docker

Build and run with Docker:

docker build -t language-ai-app .
docker run --rm -p 3000:3000 \
  -e NODE_ENV=production \
  -e PROVIDER=openrouter \
  -e OPENROUTER_API_KEY=your_key \
  language-ai-app

Or with docker-compose (reads env from your shell):

docker compose up --build

For Ollama, ensure Ollama is running on your host (default 127.0.0.1:11434). On macOS, host.docker.internal is used inside the container via compose.

Persistent cache volume

• The app persists cache and analytics when run in Docker.
• A volume is mounted at /data inside the container (configurable with CACHE_DIR, default /data).
• Configure the host path via CACHE_HOST_DIR in your .env. The deploy script ensures the directory exists on the remote host.
• See docs/persistent-caching-plan.md for full design (schema versions in shared/schemaVersions.js, LRU, image persistence for Cloze, etc.).

Switching providers

Set PROVIDER to one of:

• openrouter
• ollama

Optionally override the default model via corresponding *_MODEL env vars.