v0.2.0-llm: LLM integration, core utilities, and persistence

- Rust backend: LLM abstraction with OpenAI-compatible + Ollama API clients
- Tauri commands: generate, generate_stream, get/set LLM config
- LLM config: api_url, api_key, model, temperature, max_tokens, top_p
- Works with Ollama (localhost:11434), LM Studio, any OpenAI-compatible API
- Streaming infrastructure via Tauri Channels (LlmEvent)
- tauri-plugin-store, tauri-plugin-fs added
- Frontend: NPC Generator with race/class/alignment selection + AI generation
- Frontend: Dice Roller with notation parsing (2d6+3), presets, history
- Frontend: Session Logger with note-taking + AI summarization
- Frontend: Settings panel for LLM configuration (URL, key, model, temperature)
- App shell: left rail nav with active states, settings toggle
- All glassmorphism cards with gold glow hover states
- Builds clean: tsc, vite build, cargo check, tauri build

src-tauri/src/llm/mod.rs

@@ -0,0 +1,108 @@
+use serde::{Deserialize, Serialize};
+use std::sync::Mutex;
+use tauri::ipc::Channel;
+
+// ─── LLM Event (for streaming) ───────────────────────────────
+
+#[derive(Clone, Serialize, Deserialize)]
+#[serde(tag = "type", content = "data", rename_all = "camelCase")]
+pub enum LlmEvent {
+    Token(String),
+    Done(String),   // full text
+    Error(String),
+}
+
+// ─── App State ────────────────────────────────────────────────
+
+pub struct AppState {
+    pub config: Mutex<LlmConfig>,
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct LlmConfig {
+    pub api_url: String,
+    pub api_key: String,
+    pub model: String,
+    pub temperature: f32,
+    pub max_tokens: u32,
+    pub top_p: f32,
+}
+
+impl Default for LlmConfig {
+    fn default() -> Self {
+        Self {
+            // Default to Ollama local server; also works with LM Studio, llama.cpp server, or OpenAI
+            api_url: "http://localhost:11434".to_string(),
+            api_key: String::new(),
+            model: "llama3.2".to_string(),
+            temperature: 0.7,
+            max_tokens: 512,
+            top_p: 0.9,
+        }
+    }
+}
+
+// ─── Generation Request ───────────────────────────────────────
+
+#[derive(Debug, Deserialize)]
+pub struct GenerateRequest {
+    pub prompt: String,
+    pub system: Option<String>,
+    pub temperature: Option<f32>,
+    pub max_tokens: Option<u32>,
+}
+
+// ─── OpenAI-Compatible Chat Response ─────────────────────────
+
+#[derive(Debug, Deserialize)]
+pub struct ChatResponse {
+    pub choices: Vec<ChatChoice>,
+}
+
+#[derive(Debug, Deserialize)]
+pub struct ChatChoice {
+    pub message: ChatMessage,
+}
+
+#[derive(Debug, Serialize, Deserialize)]
+pub struct ChatMessage {
+    pub role: String,
+    pub content: String,
+}
+
+// ─── Ollama Generate Response ─────────────────────────────────
+
+#[derive(Debug, Deserialize)]
+pub struct OllamaGenerateResponse {
+    pub response: String,
+    pub done: bool,
+}
+
+#[derive(Debug, Deserialize)]
+pub struct OllamaChatResponse {
+    pub message: ChatMessage,
+    pub done: bool,
+}
+
+// ─── Helper: detect if we're talking to Ollama ───────────────
+
+pub fn is_ollama(url: &str) -> bool {
+    url.contains("localhost:11434") || url.contains("127.0.0.1:11434")
+}
+
+// ─── Build system + user messages from a prompt ──────────────
+
+pub fn build_messages(system: Option<&str>, prompt: &str) -> Vec<ChatMessage> {
+    let mut messages = Vec::new();
+    if let Some(sys) = system {
+        messages.push(ChatMessage {
+            role: "system".to_string(),
+            content: sys.to_string(),
+        });
+    }
+    messages.push(ChatMessage {
+        role: "user".to_string(),
+        content: prompt.to_string(),
+    });
+    messages
+}