feat: i18n full support, responsive UI, multi-model AI config, and bug fixes

Major Features:
- Internationalization (i18n) with auto-detection for ES/EN/PT
- Mobile-first responsive design for Studio and Experience
- Multi-model AI configuration (llama3.2:3b, qwen3.5:9b, gpt-oss:latest)
- Course language configuration (auto-detect or fixed per course)

Backend Changes:
- shared/common: ModelType enum for intelligent model selection
- LMS: log_ai_usage function migration (fix chat tutor 500 error)
- LMS/CMS: course language config fields (language_setting, fixed_language)
- LMS: /courses/{id}/language-config endpoint for language detection

Frontend Changes:
- Experience: Enhanced i18n with browser language detection
- Experience: Audio recording with HTTPS check and error handling
- Studio: Memory game with unique pair IDs and debug logging
- Studio: Expanded translations (250+ keys for ES, EN, PT)
- Both: Language selector in headers (mobile responsive)

Documentation:
- AI_MODELS_CONFIG.md: Multi-model configuration guide
- RESPONSIVIDAD_GUIA.md: Mobile-first design patterns
- I18N_RESPONSIVIDAD_IMPLEMENTACION.md: Implementation details
- DEBUG_AUDIO_RECORDING.md: Audio troubleshooting guide
- DEBUG_MEMORY_GAME.md: Memory game debugging steps

Bug Fixes:
- Fix chat tutor 500 error (missing log_ai_usage function)
- Fix audio recording (HTTPS check, browser compatibility)
- Fix memory game pair IDs (unique ID generation)
- Fix HotspotBlock TypeScript errors

Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>

shared/common/src/ai.rs

@@ -13,6 +13,59 @@ pub const DEFAULT_OLLAMA_URL: &str = "http://localhost:11434";
 /// Embedding dimensions for nomic-embed-text
 pub const EMBEDDING_DIMENSIONS: usize = 768;
 
+/// Model selection for different use cases
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum ModelType {
+    /// Fast conversational AI (chat, tutor, Q&A)
+    Chat,
+    /// Complex reasoning (analysis, recommendations, feedback)
+    Complex,
+    /// Advanced tasks (course generation, detailed analysis)
+    Advanced,
+    /// Embedding generation
+    Embedding,
+}
+
+impl ModelType {
+    /// Get the model name for this type from environment
+    pub fn get_model(&self) -> String {
+        match self {
+            ModelType::Chat => {
+                std::env::var("LOCAL_LLM_MODEL").unwrap_or_else(|_| "llama3.2:3b".to_string())
+            }
+            ModelType::Complex => {
+                std::env::var("LOCAL_LLM_MODEL_COMPLEX").unwrap_or_else(|_| "qwen3.5:9b".to_string())
+            }
+            ModelType::Advanced => {
+                std::env::var("LOCAL_LLM_MODEL_ADVANCED").unwrap_or_else(|_| "gpt-oss:latest".to_string())
+            }
+            ModelType::Embedding => {
+                std::env::var("EMBEDDING_MODEL").unwrap_or_else(|_| "nomic-embed-text".to_string())
+            }
+        }
+    }
+
+    /// Get recommended temperature for this model type
+    pub fn get_temperature(&self) -> f32 {
+        match self {
+            ModelType::Chat => 0.7,       // Balanced creativity/accuracy
+            ModelType::Complex => 0.5,    // More focused reasoning
+            ModelType::Advanced => 0.6,   // Balanced for analysis
+            ModelType::Embedding => 0.0,  // Deterministic for embeddings
+        }
+    }
+
+    /// Get max tokens for this model type
+    pub fn get_max_tokens(&self) -> u32 {
+        match self {
+            ModelType::Chat => 1024,
+            ModelType::Complex => 2048,
+            ModelType::Advanced => 4096,
+            ModelType::Embedding => 0, // Not applicable
+        }
+    }
+}
+
 #[derive(Error, Debug)]
 pub enum AiError {
     #[error("Ollama request failed: {0}")]
@@ -40,6 +93,29 @@ pub fn get_embedding_model() -> String {
     std::env::var("EMBEDDING_MODEL").unwrap_or_else(|_| DEFAULT_EMBEDDING_MODEL.to_string())
 }
 
+/// Get the best model for a specific task
+pub fn get_model_for_task(task: &str) -> String {
+    // Task-based model selection
+    match task.to_lowercase().as_str() {
+        t if t.contains("chat") || t.contains("tutor") || t.contains("conversation") => {
+            ModelType::Chat.get_model()
+        }
+        t if t.contains("quiz") || t.contains("question") || t.contains("assessment") => {
+            ModelType::Complex.get_model()
+        }
+        t if t.contains("course") || t.contains("curriculum") || t.contains("syllabus") => {
+            ModelType::Advanced.get_model()
+        }
+        t if t.contains("feedback") || t.contains("recommendation") || t.contains("analysis") => {
+            ModelType::Complex.get_model()
+        }
+        t if t.contains("transcript") || t.contains("summary") => {
+            ModelType::Chat.get_model()
+        }
+        _ => ModelType::Chat.get_model(),
+    }
+}
+
 /// Create a reqwest client that accepts invalid certificates (for dev with self-signed certs)
 fn create_insecure_client() -> Result<reqwest::Client, AiError> {
     reqwest::Client::builder()