新增: Phase2 阶段收尾（Sprint 1-20）

重构：删 5 零引用 crate（df-evolve/plugin/stages/task/traceability）+ 清死模块、ai.rs 拆 11 子 module、ai.ts 拆 6 composable、i18n 拆目录
功能：知识库全栈（df-project/scan + CRUD + 时间线 + 前端）、Settings 拆分、appSettings KV 迁移、模型池、LLM 并发 Semaphore
修复：审批持久化根治、ConditionEngine 默认拒绝、NodeRegistry unimplemented 清除、promote 补偿删除、工具结果截断 50KB、路径校验防 symlink 逃逸
文档：B-03 人工审批设计、决策记录三分档、规格契约自检、经验记录、todo 看板、PROGRESS 更新

详见 PROGRESS.md。src-tauri/儿童每日打卡应用/ 与本项目无关，已排除。

src-tauri/src/commands/ai/title.rs

@@ -0,0 +1,140 @@
+//! 对话标题生成
+
+use std::sync::Arc;
+
+use tauri::{AppHandle, Emitter};
+use tokio::sync::Mutex;
+
+use df_ai::provider::{ChatMessage, CompletionRequest, LlmProvider, MessageRole};
+use df_storage::crud::AiConversationRepo;
+use df_storage::db::Database;
+use df_storage::models::AiProviderRecord;
+
+use crate::state::LlmConcurrency;
+
+use super::AiSession;
+
+/// 对话完成后按需生成智能标题（仅 title 为空时触发一次，不覆盖用户改名）
+///
+/// - 已有 title（用户改名或已生成）→ 跳过
+/// - 否则调 LLM 非流式总结生成 ≤15 字标题；LLM 失败回退 extract_title 截断
+/// - 完成后 emit ai-conversation-changed 通知前端侧栏刷新标题
+pub(crate) async fn ensure_conversation_title(
+    provider_config: &AiProviderRecord,
+    db: &Arc<Database>,
+    conv_id: &str,
+    app_handle: &AppHandle,
+    session_arc: &Arc<Mutex<AiSession>>,
+    llm_concurrency: LlmConcurrency,
+) {
+    let conv_repo = AiConversationRepo::new(db);
+
+    // 已有标题（用户改名或已生成）→ 不覆盖
+    if let Ok(Some(rec)) = conv_repo.get_by_id(conv_id).await {
+        if rec.title.is_some() {
+            return;
+        }
+    }
+
+    // 取对话文本（仅 user/assistant，跳过 tool 噪音），取前 6 条供 LLM 总结
+    let (summary_msgs, all_msgs) = {
+        let session = session_arc.lock().await;
+        let summary: Vec<ChatMessage> = session.messages.iter()
+            .filter(|m| matches!(m.role, MessageRole::User | MessageRole::Assistant))
+            .take(6)
+            .map(|m| ChatMessage {
+                role: m.role.clone(),
+                content: m.content.clone(),
+                tool_call_id: None,
+                tool_calls: None,
+                model: None,
+            })
+            .collect();
+        (summary, session.messages.all_messages_clone())
+    };
+    if summary_msgs.is_empty() {
+        return;
+    }
+
+    // 标题生成是独立一次 LLM 调用，自建 provider（便于后台 spawn，不借主 loop 的 &dyn LlmProvider）
+    let provider: Box<dyn LlmProvider> = df_ai::build_provider(
+        &provider_config.provider_type,
+        &provider_config.base_url,
+        &provider_config.api_key,
+        &provider_config.default_model,
+    );
+    let title = match generate_title_via_llm(&*provider, &provider_config.default_model, summary_msgs, &llm_concurrency).await {
+        Some(t) => t,
+        None => extract_title(&all_msgs).unwrap_or_else(|| "新对话".to_string()),
+    };
+
+    let _ = conv_repo.update_field(conv_id, "title", &title).await;
+    let _ = app_handle.emit("ai-conversation-changed", ());
+}
+
+/// 后台生成对话标题（不阻塞主流程；失败有 extract_title 兜底）
+pub(crate) fn spawn_ensure_title(
+    provider_config: &AiProviderRecord,
+    db: &Arc<Database>,
+    conv_id: &str,
+    app_handle: &AppHandle,
+    session_arc: &Arc<Mutex<AiSession>>,
+    llm_concurrency: &LlmConcurrency,
+) {
+    let provider_config = provider_config.clone();
+    let db = db.clone();
+    let conv_id = conv_id.to_string();
+    let app_handle = app_handle.clone();
+    let session_arc = session_arc.clone();
+    let llm_concurrency = llm_concurrency.clone();
+    tauri::async_runtime::spawn(async move {
+        ensure_conversation_title(&provider_config, &db, &conv_id, &app_handle, &session_arc, llm_concurrency).await;
+    });
+}
+
+/// 调 LLM 非流式生成对话标题
+async fn generate_title_via_llm(
+    provider: &dyn LlmProvider,
+    model: &str,
+    msgs: Vec<ChatMessage>,
+    llm_concurrency: &LlmConcurrency,
+) -> Option<String> {
+    let mut prompt = vec![ChatMessage::system(
+        "你是标题生成器。根据用户与助手的对话，生成一个简短中文标题。要求：不超过15字，纯文本，不加引号不加书名号不加标点不加 emoji，只输出标题本身，不要任何前缀或解释。"
+    )];
+    prompt.extend(msgs);
+    let request = CompletionRequest {
+        model: model.to_string(),
+        messages: prompt,
+        temperature: Some(0.3),
+        max_tokens: Some(30),
+        stream: false,
+        tools: None,
+        tool_choice: None,
+    };
+    // LLM 并发限流（标题生成属独立调用，纳入双层 Semaphore）
+    let _global_permit = llm_concurrency.acquire_global().await;
+    let _per_conv_permit = llm_concurrency.acquire_per_conv().await;
+    let resp = provider.complete(request).await.ok()?;
+    Some(clean_title(&resp.text))
+}
+
+/// 清理 LLM 返回的标题：去首尾引号/书名号/空白/末尾标点，截 15 字，空则兜底"新对话"
+fn clean_title(raw: &str) -> String {
+    let t = raw.trim()
+        .trim_matches(|c: char| matches!(c, '"' | '\'' | '「' | '」' | '《' | '》' | ' ' | '\n' | '\r'));
+    let t = t.trim_end_matches(|c: char| matches!(c, '。' | '.' | ',' | '，' | '！' | '!' | '？' | '?' | ':' | '：'));
+    let cleaned: String = t.chars().take(15).collect();
+    if cleaned.is_empty() { "新对话".to_string() } else { cleaned }
+}
+
+/// 从消息历史中提取对话标题（取第一条用户消息前 30 字）
+pub(crate) fn extract_title(messages: &[ChatMessage]) -> Option<String> {
+    messages.iter()
+        .find(|m| matches!(m.role, MessageRole::User))
+        .map(|m| {
+            let mut chars = m.content.chars();
+            let t: String = chars.by_ref().take(30).collect();
+            if chars.next().is_some() { format!("{}...", t) } else { t }
+        })
+}