Babbelaar is an open source language learning app that helps you learn a foreign language through personalized conversations with an AI teacher. It offers Text-to-Speech and Voice Recognition to be able to listen to and practice your pronunciation. Babbelaar does not have any server components and uses a free LLM service. To use all functionality you can provider your own OpenAI API key.

Is the output supposed to make better use of markup formatting? I'm getting the following as raw output:

"{ "message": "message": Bonjour! Je suis heureux de parler avec quelqu’un qui aime la nourriture et le vin. Moi aussi, j’aime découvrir de nouveaux vins quand je voyage. Tu as déjà goûté un vin spécial lors de ton voyage ?", "explanation": "I'm greeting you in French and mentioning your interests in food and wine to make the conversation engaging. I ask if you've ever tasted a special wine during your travels, which encourages you to talk about your experiences.", "reply": "Bonjour ! Oui, j’ai goûté un vin rouge en France, dans la région de Bordeaux. C’était délicieux ! Et toi, quel est ton vin préféré ?", "suggested": ["bonjour", "oui", "vin", "région", "délicieux"], "translations": { "Bonjour": "Hello", "oui": "yes", "vin": "wine", "région": "region", "délicieux": "delicious" } }"

Which is quite difficult to read and learn from.