persistent auth in SQLite, API chat/logs, agent completion via channel

- Auth: move from state.json to SQLite authed_chats table, with memory cache
- Remove Persistent/state.json, all state now in noc.db
- HTTP API: POST /api/chat (end-to-end LLM), GET /api/logs (failed API requests)
- API logging: store raw request/response for 400 errors in api_log table
- Agent completion: spawn_agent sends LifeEvent::AgentDone via channel,
  life loop picks up with full conversation context and responds
- Config structs: derive Clone for HTTP server
- System prompt: instruct LLM not to add timestamps
- Makefile: rsync without --delete to preserve VPS-only tools

src/life.rs

@@ -19,6 +19,15 @@ const DIARY_SCHEDULE: &str = "cron:0 55 22 * * *";
 pub enum LifeEvent {
     /// Force-fire a specific timer by ID.
     FireTimer(i64),
+    /// A sub-agent completed — feed result back through LLM.
+    AgentDone {
+        id: String,
+        chat_id: i64,
+        session_id: String,
+        task: String,
+        output: String,
+        exit_code: Option<i32>,
+    },
 }
 
 pub async fn life_loop(
@@ -54,6 +63,46 @@ pub async fn life_loop(
                             warn!(timer_id = id, "force-fire: timer not found");
                         }
                     }
+                    LifeEvent::AgentDone { id, chat_id: cid, session_id, task, output, exit_code } => {
+                        info!(agent = %id, session = %session_id, "agent done, notifying");
+                        let preview = crate::display::truncate_at_char_boundary(&output, 3000);
+                        let notification = format!(
+                            "[子代理 '{id}' 完成 (exit={exit_code:?})]\n任务: {task}\n输出:\n{preview}"
+                        );
+
+                        // load conversation context so LLM knows what was discussed
+                        let conv = state.load_conv(&session_id).await;
+                        let persona = state.get_config("persona").await.unwrap_or_default();
+                        let memory_slots = state.get_memory_slots().await;
+                        let inner = state.get_inner_state().await;
+
+                        let system = crate::stream::build_system_prompt(
+                            &conv.summary, &persona, &memory_slots, &inner,
+                        );
+
+                        let mut messages = vec![system];
+                        // include recent conversation history
+                        messages.extend(conv.messages.iter().cloned());
+                        // append the agent completion as a new user message
+                        messages.push(serde_json::json!({"role": "user", "content": notification}));
+
+                        if let BackendConfig::OpenAI { ref endpoint, ref model, ref api_key } = config.backend {
+                            let chat_id_tg = ChatId(cid);
+                            let sid = format!("agent-{id}");
+                            let mut tg_output;
+                            let mut buf_output;
+                            let out: &mut dyn crate::output::Output = if cid == 0 {
+                                buf_output = BufferOutput::new();
+                                &mut buf_output
+                            } else {
+                                tg_output = TelegramOutput::new(bot.clone(), chat_id_tg, true);
+                                &mut tg_output
+                            };
+                            let _ = run_openai_with_tools(
+                                endpoint, model, api_key, messages, out, &state, &sid, &config, cid,
+                            ).await;
+                        }
+                    }
                 }
             }
         }