feat: follow-up question chips + dashboard auto-refresh

CLAUDE.md

@@ -103,10 +103,10 @@ Everything below is built and working in the codebase:
 
 ### Frontend
 - React 18 + TypeScript + Vite + Tailwind CSS + `@tailwindcss/typography`
-- Font: Inter Variable (Google Fonts) with Linear's full fallback stack
+- Font: Inter Variable (Google Fonts) with a full system fallback stack
 - Routes: `/`, `/chat/:sessionId`, `/dashboard`, `/monitoring`, `/monitoring/:alertId`, `/history`, `/settings`, `/users`, `/login`
 - Chat page: SSE streaming, tool call inspector, cost/latency card, stop button, suggestion chips on empty state, `?prompt=` deeplink support
-- Sidebar: paginated session list (15 at a time), three-dot menu with delete, real `<a>` links for native right-click
+- Sidebar: paginated session list (15 at a time), three-dot menu with rename + delete (portal-based, no overflow clipping), real `<a>` links for native right-click
 - Settings: Environment, Agent Config, Integrations (UI stubs), AWS Configuration (editable, admin only), Preferences (dark mode)
 
 ### Auth & MCP
@@ -168,7 +168,7 @@ Incomplete items from the README roadmap (do not mark complete here — update R
 | Linting | `ruff>=0.4.0` (line-length 100, Python 3.11 target, `asyncio_mode = "auto"`) |
 | Package manager | **uv** — always `uv run` and `uv add`, never bare `pip` |
 | Frontend | React 18, TypeScript, Vite, Tailwind CSS 3, `@tailwindcss/typography` |
-| Font | Inter Variable (Google Fonts) — Linear-style fallback stack in `tailwind.config.js` |
+| Font | Inter Variable (Google Fonts) — full system fallback stack in `tailwind.config.js` |
 
 ---
 

docs/ui.md

@@ -20,6 +20,11 @@ Tool calls appear as collapsible cards as they complete, showing the tool name,
 arguments, and result. A contextual status label ("Digging through CloudTrail…",
 "Checking Lambda metrics…") appears while each tool runs.
 
+**Follow-up suggestions** — after an investigation completes, up to 3 drill-down
+question chips appear above the input box. Clicking one sends it immediately as the
+next message. The chips are generated by the agent in the same final tool call at
+no extra cost and disappear once you start a new message.
+
 **Tool call inspector** — each tool call card is collapsed by default; click to expand
 and see the full arguments and result JSON.
 
@@ -45,8 +50,9 @@ Lists all past sessions sorted by last activity. Each entry shows:
 **Actions:**
 - Click any session to load it in the chat view with full history restored
 - **New chat** button starts a fresh session with a new UUID
-- **Delete** (trash icon) soft-deletes the session — it disappears from the list
-  immediately but data is preserved in the database
+- **⋯ menu** on each session row:
+  - **Rename** — shows an inline input pre-filled with the current title; Enter or blur saves, Escape cancels
+  - **Delete** — soft-deletes the session; it disappears from the list immediately but data is preserved in the database
 
 ---
 
@@ -64,6 +70,8 @@ Click any row to open the session in the chat page.
 
 ## Dashboard page
 
+Stats are automatically refreshed every 30 seconds — no manual page reload needed.
+
 See [`dashboard.md`](dashboard.md) for full details on each panel.
 
 ---
@@ -132,7 +140,7 @@ The UI respects the system dark/light mode preference automatically. No toggle i
 
 ## Theme and design
 
-- Font: system monospace stack
+- Font: Inter Variable (Google Fonts) with a full system fallback stack
 - Framework: React 18 + Tailwind CSS
 - Icons: Lucide React
 - Build tool: Vite with HMR in development

frontend/src/pages/ChatPage.tsx

@@ -38,11 +38,12 @@ interface Props {
 export default function ChatPage({ onSessionsChange, onNew }: Props) {
   const { sessionId } = useParams<{ sessionId: string }>();
   const [searchParams, setSearchParams] = useSearchParams();
-  const [messages,  setMessages] = useState<Message[]>([]);
-  const [busy,      setBusy]     = useState(false);
-  const abortRef                 = useRef<AbortController | null>(null);
-  const bottomRef                = useRef<HTMLDivElement>(null);
-  const autoPromptFired          = useRef(false);
+  const [messages,          setMessages]          = useState<Message[]>([]);
+  const [busy,              setBusy]              = useState(false);
+  const [followUpQuestions, setFollowUpQuestions] = useState<string[]>([]);
+  const abortRef                                  = useRef<AbortController | null>(null);
+  const bottomRef                                 = useRef<HTMLDivElement>(null);
+  const autoPromptFired                           = useRef(false);
 
   useEffect(() => {
     if (sessionId) localStorage.setItem('devops-session-id', sessionId);
@@ -74,6 +75,7 @@ export default function ChatPage({ onSessionsChange, onNew }: Props) {
 
   const send = async (text: string) => {
     if (!sessionId || busy || !text.trim()) return;
+    setFollowUpQuestions([]);
 
     const agentId = crypto.randomUUID();
     setMessages(prev => [
@@ -135,6 +137,8 @@ export default function ChatPage({ onSessionsChange, onNew }: Props) {
             }));
           } else if (payload.type === 'done') {
             patch(agentId, { streaming: false, usage: payload.usage as AgentMsg['usage'] });
+            const fq = payload.follow_up_questions;
+            if (Array.isArray(fq) && fq.length > 0) setFollowUpQuestions(fq as string[]);
           } else if (payload.type === 'error') {
             const msg = payload.message as string;
             patch(agentId, { streaming: false, error: msg });
@@ -208,6 +212,19 @@ export default function ChatPage({ onSessionsChange, onNew }: Props) {
         <div ref={bottomRef} />
       </div>
 
+      {followUpQuestions.length > 0 && !busy && (
+        <div className="px-6 pb-2 flex flex-wrap gap-2">
+          {followUpQuestions.map((q, i) => (
+            <button
+              key={i}
+              onClick={() => send(q)}
+              className="text-[13px] px-3 py-1.5 rounded-full border border-indigo-200 dark:border-[#3730A3] bg-indigo-50 dark:bg-[#1E1B4B] text-indigo-600 dark:text-[#818CF8] hover:bg-indigo-100 dark:hover:bg-[#2D2B5C] transition-colors"
+            >
+              {q}
+            </button>
+          ))}
+        </div>
+      )}
       <InputArea busy={busy} onSend={send} onStop={stop} />
     </div>
   );

frontend/src/pages/DashboardPage.tsx

@@ -70,11 +70,16 @@ export default function DashboardPage() {
   const [error,   setError]   = useState<string | null>(null);
 
   useEffect(() => {
-    fetch('/stats')
-      .then(r => { if (!r.ok) throw new Error(r.statusText); return r.json(); })
-      .then(setStats)
-      .catch(e => setError(e.message))
-      .finally(() => setLoading(false));
+    const load = () =>
+      fetch('/stats')
+        .then(r => { if (!r.ok) throw new Error(r.statusText); return r.json(); })
+        .then(setStats)
+        .catch(e => setError(e.message))
+        .finally(() => setLoading(false));
+
+    load();
+    const interval = setInterval(load, 30_000);
+    return () => clearInterval(interval);
   }, []);
 
   if (loading) {

src/api/routers/chat.py

@@ -251,7 +251,15 @@ def _flush_text_buf():
     await save_turn(session_id, user_message, response_text, tool_calls_log, usage)
     await notify_slack(session_id, tool_calls_log)
 
-    yield f"data: {json.dumps({'type': 'done', 'usage': usage})}\n\n"
+    follow_up_questions: list[str] = []
+    for tc in tool_calls_log:
+        if tc["tool"] == "submit_investigation":
+            fq = tc["args"].get("follow_up_questions")
+            if isinstance(fq, list):
+                follow_up_questions = [q for q in fq if isinstance(q, str)]
+            break
+
+    yield f"data: {json.dumps({'type': 'done', 'usage': usage, 'follow_up_questions': follow_up_questions})}\n\n"
 
 
 @router.post("/chat")

src/tools/final_answer.py

@@ -22,10 +22,16 @@ def submit_investigation(
     confidence: VALID_CONFIDENCE,
     services_affected: list[str],
     recommended_follow_up: str,
+    follow_up_questions: list[str],
 ) -> str:
     """Submit the final structured investigation result. Call this exactly once when you have
     gathered sufficient evidence and reached a conclusion. Do not output a JSON block in
-    free text — call this tool instead."""
+    free text — call this tool instead.
+
+    follow_up_questions: 3 short drill-down questions the user might want to ask next,
+    e.g. ["What caused the spike at 14:32?", "Are retries configured on the Lambda?",
+    "Has this happened before this week?"].
+    """
     return "Investigation result recorded."