Add tool execution guardrails with confirmation popup (#438)

docs/guides/agent-ui.mdx

@@ -10,17 +10,6 @@ GAIA Agent UI is a desktop interface for running AI agents **100% locally** on y
   **Ready to install?** See the [Quickstart](/quickstart#agent-ui-fastest) for installation instructions.
 </Info>
 
-<Warning>
-  **Tested Configuration:** The Agent UI has been tested exclusively on **AMD Ryzen AI MAX+ 395** processors running the **Qwen3-Coder-30B-A3B-Instruct-GGUF** model via Lemonade Server. Other hardware or model combinations may work but are not officially verified.
-
-  If you encounter issues on a different configuration, please [open a GitHub issue](https://github.com/amd/gaia/issues/new) and include:
-  - Your processor model (e.g., Ryzen AI 9 HX 370, Ryzen AI MAX+ 395)
-  - RAM and available memory
-  - The LLM model you are using
-  - Operating system and version
-  - Steps to reproduce the issue
-</Warning>
-
 ---
 
 ## What You Can Do
@@ -85,11 +74,7 @@ See the [Agent UI MCP Server guide](/guides/mcp/agent-ui) for setup instructions
 
   <Accordion title="Port 4200 already in use">
     ```bash
-    # npm CLI
-    gaia-ui --port 8080
-
-    # Python CLI
-    gaia --ui-port 8080
+    gaia --ui --ui-port 8080
     ```
   </Accordion>
 

docs/sdk/sdks/agent-ui.mdx

@@ -19,10 +19,6 @@ from gaia.ui.models import SystemStatus, ChatRequest, SessionResponse, DocumentR
 
 **See also:** [User Guide](/guides/agent-ui) | [Agent SDK](/sdk/sdks/chat) | [API Specification](/spec/agent-ui-server)
 
-<Warning>
-  **Tested Configuration:** The Agent UI has been tested on **AMD Ryzen AI MAX+ 395** with **Qwen3-Coder-30B-A3B-Instruct-GGUF**. Other configurations are not officially verified. See the [User Guide](/guides/agent-ui) for full details and how to report issues on other hardware.
-</Warning>
-
 ---
 
 ## Overview
@@ -350,7 +346,6 @@ class MessageResponse(BaseModel):
     content: str
     created_at: str
     rag_sources: Optional[List[SourceInfo]] = None
-    agent_steps: Optional[List[AgentStepResponse]] = None
 
 class MessageListResponse(BaseModel):
     messages: List[MessageResponse]
@@ -370,7 +365,6 @@ class DocumentResponse(BaseModel):
     indexed_at: str
     last_accessed_at: Optional[str] = None
     sessions_using: int = 0
-    indexing_status: str = "complete"  # pending | indexing | complete | failed | cancelled | missing
 
 class DocumentListResponse(BaseModel):
     documents: List[DocumentResponse]
@@ -865,8 +859,8 @@ from gaia.rag.sdk import RAGSDK, RAGConfig
 
 config = RAGConfig()
 rag = RAGSDK(config)
-result = rag.index_document(filepath)
-chunk_count = result.get("num_chunks", 0)
+result = rag.index_file(filepath)
+chunk_count = result.get("chunk_count", 0)
 ```
 
 ---
@@ -879,16 +873,16 @@ GAIA Agent UI is also available as an npm package for quick installation:
 npm install -g @amd-gaia/agent-ui
 ```
 
-This provides the `gaia-ui` CLI command:
+This provides the `gaia` CLI command:
 
 ```bash
-gaia-ui              # Start Python backend + open browser
-gaia-ui --serve      # Serve frontend only (Node.js static server)
-gaia-ui --port 8080  # Custom port
-gaia-ui --version    # Show version
+gaia              # Start Python backend + open browser
+gaia --serve      # Serve frontend only (Node.js static server)
+gaia --port 8080  # Custom port
+gaia --version    # Show version
 ```
 
-On first run, `gaia-ui` automatically installs the Python backend (uv, Python 3.12, amd-gaia) if not already present. On subsequent runs, it auto-updates if the version doesn't match.
+On first run, `gaia` automatically installs the Python backend (uv, Python 3.12, amd-gaia) if not already present.
 
 ### Package Contents
 

src/gaia/agents/base/agent.py

@@ -31,6 +31,11 @@
 CHUNK_TRUNCATION_THRESHOLD = 5000
 CHUNK_TRUNCATION_SIZE = 2500
 
+# Tools that require explicit user confirmation before execution.
+# Adding a tool name here causes _execute_tool() to call
+# console.confirm_tool_execution() and block until the user responds.
+TOOLS_REQUIRING_CONFIRMATION = {"run_shell_command"}
+
 
 class Agent(abc.ABC):
     """
@@ -1148,6 +1153,16 @@ def _execute_tool(self, tool_name: str, tool_args: Dict[str, Any]) -> Any:
                 logger.error(f"Tool '{tool_name}' not found in registry")
                 return {"status": "error", "error": f"Tool '{tool_name}' not found"}
 
+        # Guardrail: require explicit user confirmation for high-risk tools.
+        # The SSEOutputHandler overrides this to block until the frontend
+        # responds; the default implementation auto-approves (CLI path).
+        if tool_name in TOOLS_REQUIRING_CONFIRMATION:
+            if not self.console.confirm_tool_execution(tool_name, tool_args):
+                return {
+                    "status": "denied",
+                    "error": f"Tool '{tool_name}' was denied by the user.",
+                }
+
         tool = _TOOL_REGISTRY[tool_name]["function"]
         sig = inspect.signature(tool)
 

src/gaia/agents/base/console.py

@@ -205,6 +205,14 @@ def print_header(self, text: str):  # pylint: disable=unused-argument
         """Print header. Optional - default no-op."""
         ...
 
+    def confirm_tool_execution(
+        self,
+        tool_name: str,  # pylint: disable=unused-argument
+        tool_args: Dict[str, Any],  # pylint: disable=unused-argument
+    ) -> bool:
+        """Request user confirmation before executing a tool. Returns True to proceed."""
+        return True
+
     def print_separator(self, length: int = 50):  # pylint: disable=unused-argument
         """Print separator. Optional - default no-op."""
         ...

src/gaia/apps/webui/src/components/ChatView.tsx

@@ -5,6 +5,8 @@ import { useEffect, useRef, useCallback, useState } from 'react';
 import { Edit3, Paperclip, Download, Send, Upload, MessageSquare, Square, ArrowDown, Lock, FileText, FolderSearch, CheckCircle2, X, Link } from 'lucide-react';
 import { MessageBubble } from './MessageBubble';
 import { useChatStore } from '../stores/chatStore';
+import { useNotificationStore, ALWAYS_ALLOW_TOOLS_KEY } from '../stores/notificationStore';
+import type { GaiaNotification } from '../types/agent';
 import * as api from '../services/api';
 import { log } from '../utils/logger';
 import { getSessionHash } from '../utils/format';
@@ -129,6 +131,8 @@ export function ChatView({ sessionId }: ChatViewProps) {
         systemStatus,
     } = useChatStore();
 
+    const { addNotification } = useNotificationStore();
+
     const session = sessions.find((s) => s.id === sessionId);
     const [input, setInput] = useState('');
     const [editingTitle, setEditingTitle] = useState(false);
@@ -598,6 +602,43 @@ export function ChatView({ sessionId }: ChatViewProps) {
                 }
             },
             onAgentEvent: (event) => {
+                // ── Tool confirmation popup ──────────────────────────────
+                if (event.type === 'tool_confirm') {
+                    if (!event.confirm_id) {
+                        console.error('[ChatView] tool_confirm event missing confirm_id, ignoring');
+                        return;
+                    }
+                    const toolName = event.tool || '';
+                    const alwaysAllowed: string[] = JSON.parse(
+                        localStorage.getItem(ALWAYS_ALLOW_TOOLS_KEY) || '[]'
+                    );
+                    if (alwaysAllowed.includes(toolName)) {
+                        // Auto-approve without showing the modal
+                        api.confirmToolExecution(sessionId, event.confirm_id, 'allow', false).catch(
+                            (err) => console.error('[ChatView] auto-confirm failed:', err)
+                        );
+                        return;
+                    }
+                    // Show the PermissionPrompt modal via notificationStore
+                    const notification: GaiaNotification = {
+                        id: event.confirm_id,
+                        type: 'permission_request',
+                        agentId: 'chat',
+                        agentName: 'GAIA',
+                        title: `Allow ${toolName}?`,
+                        message: `The agent wants to execute: ${toolName}`,
+                        timestamp: Date.now(),
+                        read: false,
+                        dismissed: false,
+                        priority: 'high',
+                        tool: toolName,
+                        toolArgs: event.args as Record<string, unknown> | undefined,
+                        timeoutSeconds: event.timeout_seconds ?? 60,
+                    };
+                    addNotification(notification);
+                    return;
+                }
+
                 // Tool completion updates the last TOOL step (not just the last step,
                 // since thinking/status events may have been interleaved during execution)
                 if (event.type === 'tool_end') {

src/gaia/apps/webui/src/services/api.ts

@@ -143,7 +143,7 @@ export interface StreamCallbacks {
 /** Agent event types that represent activity rather than content. */
 const AGENT_EVENT_TYPES = new Set([
     'status', 'step', 'thinking', 'plan',
-    'tool_start', 'tool_end', 'tool_result', 'tool_args', 'agent_error',
+    'tool_start', 'tool_end', 'tool_result', 'tool_args', 'tool_confirm', 'agent_error',
 ]);
 
 export function sendMessageStream(
@@ -277,6 +277,18 @@ export function sendMessageStream(
     return controller;
 }
 
+// -- Tool Confirmation ---------------------------------------------------------
+
+/** Resolve a pending tool execution confirmation (Allow or Deny). */
+export async function confirmToolExecution(
+    sessionId: string,
+    confirmId: string,
+    action: 'allow' | 'deny',
+    remember: boolean,
+): Promise<void> {
+    return apiFetch('POST', '/chat/confirm', { session_id: sessionId, confirm_id: confirmId, action, remember });
+}
+
 // -- Documents -----------------------------------------------------------------
 
 export async function listDocuments(): Promise<{ documents: Document[]; total: number; total_size_bytes: number; total_chunks: number }> {

src/gaia/apps/webui/src/stores/notificationStore.ts

@@ -11,12 +11,17 @@
 
 import { create } from 'zustand';
 import type { GaiaNotification, NotificationType } from '../types/agent';
+import { confirmToolExecution } from '../services/api';
+import { useChatStore } from './chatStore';
 
 // ── Constants ────────────────────────────────────────────────────────────
 
 /** Maximum notifications kept in the center to prevent unbounded growth. */
 const MAX_NOTIFICATIONS = 500;
 
+/** localStorage key for the "always allow" tool list. */
+export const ALWAYS_ALLOW_TOOLS_KEY = 'gaia_always_allow_tools';
+
 // ── State Interface ──────────────────────────────────────────────────────
 
 interface NotificationState {
@@ -78,18 +83,41 @@ export const useNotificationStore = create<NotificationState>((set, get) => ({
   setTypeFilter: (type) => set({ typeFilter: type }),
 
   respondToPermission: async (id, action, remember) => {
-    const api = window.gaiaAPI;
-    if (api) {
+    const electronApi = window.gaiaAPI;
+    if (electronApi) {
+      // Electron path: route via IPC
       try {
-        await api.notification.respondPermission(id, action, remember);
+        await electronApi.notification.respondPermission(id, action, remember);
       } catch (err) {
         console.error('[notificationStore] Failed to send permission response via IPC:', err);
         // Don't update local state — the agent didn't receive the response.
         // The permission prompt remains actionable so the user can retry.
         return;
       }
+    } else {
+      // Web path: route via HTTP to /api/chat/confirm
+      const sessionId = useChatStore.getState().currentSessionId;
+      if (sessionId) {
+        try {
+          await confirmToolExecution(sessionId, id, action, remember);
+        } catch (err) {
+          console.error('[notificationStore] Failed to send permission response via HTTP:', err);
+          return;
+        }
+      }
+    }
+    // Persist "always allow" preference in localStorage
+    if (action === 'allow' && remember) {
+      const notification = get().notifications.find((n) => n.id === id);
+      if (notification?.tool) {
+        const existing: string[] = JSON.parse(localStorage.getItem(ALWAYS_ALLOW_TOOLS_KEY) || '[]');
+        if (!existing.includes(notification.tool)) {
+          existing.push(notification.tool);
+          localStorage.setItem(ALWAYS_ALLOW_TOOLS_KEY, JSON.stringify(existing));
+        }
+      }
     }
-    // Update local state only after IPC succeeds (or if no IPC is available)
+    // Update local state after response is delivered
     set((state) => ({
       notifications: state.notifications.map((n) =>
         n.id === id

src/gaia/apps/webui/src/styles/index.css

@@ -348,6 +348,23 @@ textarea:focus-visible {
     line-height: 1.3;
 }
 
+/* ── Beta Badge ─────────────────────────────────────────────────── */
+
+.beta-badge {
+    display: inline-block;
+    font-size: 10px;
+    font-weight: 800;
+    font-family: var(--font-mono);
+    text-transform: uppercase;
+    letter-spacing: 1px;
+    padding: 2px 6px;
+    border-radius: 3px;
+    background: var(--accent-yellow);
+    color: #111;
+    vertical-align: middle;
+    line-height: 1.3;
+}
+
 /* ── Modal Base ──────────────────────────────────────────────────── */
 
 .modal-overlay {

src/gaia/apps/webui/src/types/index.ts

@@ -197,19 +197,20 @@ export interface AgentStep {
 
 /** Extended SSE event types for agent communication. */
 export type StreamEventType =
-    | 'chunk'       // Text content chunk
-    | 'done'        // Stream complete
-    | 'error'       // Error
-    | 'status'      // Agent state change
-    | 'step'        // Step progress
-    | 'thinking'    // Agent reasoning
-    | 'plan'        // Agent plan
-    | 'tool_start'  // Tool execution started
-    | 'tool_end'    // Tool execution completed
-    | 'tool_result' // Tool result summary
-    | 'tool_args'   // Tool arguments detail
-    | 'answer'      // Final answer from agent
-    | 'agent_error';// Agent-level error (non-fatal)
+    | 'chunk'        // Text content chunk
+    | 'done'         // Stream complete
+    | 'error'        // Error
+    | 'status'       // Agent state change
+    | 'step'         // Step progress
+    | 'thinking'     // Agent reasoning
+    | 'plan'         // Agent plan
+    | 'tool_start'   // Tool execution started
+    | 'tool_end'     // Tool execution completed
+    | 'tool_result'  // Tool result summary
+    | 'tool_args'    // Tool arguments detail
+    | 'tool_confirm' // Tool requires user confirmation (blocking)
+    | 'answer'       // Final answer from agent
+    | 'agent_error'; // Agent-level error (non-fatal)
 
 export interface StreamEvent {
     type: StreamEventType;
@@ -243,6 +244,10 @@ export interface StreamEvent {
         duration_seconds?: number;
         truncated?: boolean;
     };
+    /** Confirmation ID (for tool_confirm events). */
+    confirm_id?: string;
+    /** Timeout in seconds (for tool_confirm events). */
+    timeout_seconds?: number;
     /** Structured result data (for tool_result with search results, file lists, etc.). */
     result_data?: {
         type: string;