feat: Initial project structure for WebTriviasMulti

- Backend: FastAPI + Python-SocketIO + SQLAlchemy
  - Models for categories, questions, game sessions, events
  - AI services for answer validation and question generation (Claude)
  - Room management with Redis
  - Game logic with stealing mechanics
  - Admin API for question management

- Frontend: React + Vite + TypeScript + Tailwind
  - 5 visual themes (DRRR, Retro, Minimal, RGB, Anime 90s)
  - Real-time game with Socket.IO
  - Achievement system
  - Replay functionality
  - Sound effects per theme

- Docker Compose for deployment
- Design documentation

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

backend/app/services/ai_validator.py

@@ -0,0 +1,80 @@
+import json
+from anthropic import Anthropic
+from app.config import get_settings
+
+settings = get_settings()
+
+
+class AIValidator:
+    def __init__(self):
+        self.client = Anthropic(api_key=settings.anthropic_api_key)
+
+    async def validate_answer(
+        self,
+        question: str,
+        correct_answer: str,
+        alt_answers: list[str],
+        player_answer: str
+    ) -> dict:
+        """
+        Validate if the player's answer is correct using Claude AI.
+
+        Returns:
+            dict: {"valid": bool, "reason": str}
+        """
+        prompt = f"""Eres un validador de trivia. Determina si la respuesta del jugador
+es correcta comparándola con la respuesta oficial.
+
+Pregunta: {question}
+Respuesta correcta: {correct_answer}
+Respuestas alternativas válidas: {', '.join(alt_answers) if alt_answers else 'Ninguna'}
+Respuesta del jugador: {player_answer}
+
+Considera válido si:
+- Es sinónimo o variación de la respuesta correcta
+- Tiene errores menores de ortografía
+- Usa abreviaciones comunes (ej: "BOTW" = "Breath of the Wild")
+- Es conceptualmente equivalente
+
+Responde SOLO con JSON: {{"valid": true/false, "reason": "breve explicación"}}"""
+
+        try:
+            message = self.client.messages.create(
+                model="claude-3-haiku-20240307",
+                max_tokens=150,
+                messages=[
+                    {"role": "user", "content": prompt}
+                ]
+            )
+
+            response_text = message.content[0].text.strip()
+
+            # Parse JSON response
+            result = json.loads(response_text)
+            return result
+
+        except json.JSONDecodeError:
+            # If JSON parsing fails, try to extract the result
+            response_lower = response_text.lower()
+            if "true" in response_lower:
+                return {"valid": True, "reason": "Respuesta validada por IA"}
+            return {"valid": False, "reason": "No se pudo validar la respuesta"}
+
+        except Exception as e:
+            print(f"Error validating answer: {e}")
+            # Fallback to exact match
+            player_lower = player_answer.lower().strip()
+            correct_lower = correct_answer.lower().strip()
+
+            if player_lower == correct_lower:
+                return {"valid": True, "reason": "Coincidencia exacta"}
+
+            for alt in alt_answers:
+                if player_lower == alt.lower().strip():
+                    return {"valid": True, "reason": "Coincide con respuesta alternativa"}
+
+            return {"valid": False, "reason": "Respuesta incorrecta"}
+
+
+# Singleton instance
+ai_validator = AIValidator()