feat: add job stall detection, database connection limits, configurable local notifications, and secure proxy header handling with entrypoint support.

Dockerfile

@@ -43,6 +43,7 @@ RUN apt-get update && \
     va-driver-all \
     libsqlite3-0 \
     ca-certificates \
+    gosu \
     && if [ "$(dpkg --print-architecture)" = "amd64" ]; then \
     apt-get install -y --no-install-recommends \
     intel-media-va-driver-non-free \
@@ -75,10 +76,16 @@ RUN set -e; \
 COPY --from=builder /app/target/release/alchemist /usr/local/bin/alchemist
 
 # Set environment variables
-ENV LIBVA_DRIVER_NAME=iHD
+# VA-API driver auto-detection: do NOT hardcode LIBVA_DRIVER_NAME here.
+# Users can override via: docker run -e LIBVA_DRIVER_NAME=iHD ...
+# Common values: iHD (Intel ≥ Broadwell), i965 (older Intel), radeonsi (AMD)
 ENV RUST_LOG=info
 ENV ALCHEMIST_CONFIG_PATH=/app/config/config.toml
 ENV ALCHEMIST_DB_PATH=/app/data/alchemist.db
+COPY entrypoint.sh /app/entrypoint.sh
+RUN chmod +x /app/entrypoint.sh
+
 EXPOSE 3000
 
-ENTRYPOINT ["alchemist"]
+ENTRYPOINT ["/app/entrypoint.sh"]
+CMD ["alchemist"]

entrypoint.sh

@@ -0,0 +1,31 @@
+#!/bin/bash
+set -e
+
+PUID=${PUID:-0}
+PGID=${PGID:-0}
+
+if [ "$PUID" -ne 0 ] && [ "$PGID" -ne 0 ]; then
+    echo "Starting Alchemist with UID: $PUID, GID: $PGID"
+    
+    # Create group and user securely if they don't exist
+    if ! getent group alchemist >/dev/null; then
+        groupadd -g "$PGID" alchemist
+    fi
+    if ! getent passwd alchemist >/dev/null; then
+        useradd -u "$PUID" -g "$PGID" -s /bin/bash -m -d /app alchemist
+    fi
+    
+    # Take ownership of app data — skip gracefully for read-only mounts
+    for dir in /app/config /app/data; do
+        if [ -d "$dir" ]; then
+            chown -R alchemist:alchemist "$dir" 2>/dev/null || \
+                echo "Warning: Cannot chown $dir (read-only mount?). Continuing..."
+        fi
+    done
+    
+    # Drop privileges and execute
+    exec gosu alchemist "$@"
+else
+    # Run natively
+    exec "$@"
+fi

src/config.rs

@@ -716,12 +716,23 @@ impl Config {
         Ok(())
     }
 
-    /// Save config to file
+    /// Save config to file atomically (write to temp, then rename).
+    /// This prevents corruption if the process crashes mid-write.
     pub fn save(&self, path: &Path) -> Result<()> {
         let mut config = self.clone();
         config.canonicalize_for_save();
         let content = toml::to_string_pretty(&config)?;
-        std::fs::write(path, content)?;
+
+        let tmp = path.with_extension("toml.tmp");
+        std::fs::write(&tmp, &content)?;
+
+        // Atomic rename: if this fails, the original config is still intact.
+        if let Err(e) = std::fs::rename(&tmp, path) {
+            // Clean up the temp file on rename failure
+            let _ = std::fs::remove_file(&tmp);
+            return Err(e.into());
+        }
+
         Ok(())
     }
 

src/db.rs

@@ -619,7 +619,9 @@ impl Db {
             .journal_mode(SqliteJournalMode::Wal)
             .busy_timeout(Duration::from_secs(5));
 
-        let pool = SqlitePool::connect_with(options).await?;
+        let pool = sqlx::sqlite::SqlitePoolOptions::new()
+            .max_connections(1)
+            .connect_with(options).await?;
         info!(
             target: "startup",
             "Database connection opened in {} ms",

src/main.rs

@@ -463,17 +463,18 @@ async fn run() -> Result<()> {
     // Keep legacy channel for transition compatibility
     let (tx, _rx) = broadcast::channel(100);
 
-    // Initialize Notification Manager
-    let notification_manager = Arc::new(alchemist::notifications::NotificationManager::new(
-        db.as_ref().clone(),
-    ));
-    notification_manager.start_listener(tx.subscribe());
-
     let transcoder = Arc::new(Transcoder::new());
     let hardware_state = hardware::HardwareState::new(Some(hw_info.clone()));
     let hardware_probe_log = Arc::new(RwLock::new(initial_probe_log));
     let config = Arc::new(RwLock::new(config));
 
+    // Initialize Notification Manager (needs config for allow_local_notifications)
+    let notification_manager = Arc::new(alchemist::notifications::NotificationManager::new(
+        db.as_ref().clone(),
+        config.clone(),
+    ));
+    notification_manager.start_listener(tx.subscribe());
+
     let maintenance_db = db.clone();
     let maintenance_config = config.clone();
     tokio::spawn(async move {

src/media/pipeline.rs

@@ -1263,10 +1263,29 @@ impl Pipeline {
 
         if let Ok(file_settings) = self.db.get_file_settings().await {
             if file_settings.delete_source {
+                // Safety: verify the promoted output is intact before destroying the source.
+                // This prevents data loss if the filesystem silently corrupted the output
+                // during rename (e.g., stale NFS/SMB mount, full disk).
+                match std::fs::metadata(context.output_path) {
+                    Ok(m) if m.len() > 0 => {
                         if let Err(e) = std::fs::remove_file(input_path) {
                             tracing::warn!("Failed to delete source {:?}: {}", input_path, e);
                         }
                     }
+                    Ok(_) => {
+                        tracing::error!(
+                            "Job {}: Output file {:?} is empty after promotion — source preserved to prevent data loss",
+                            job_id, context.output_path
+                        );
+                    }
+                    Err(e) => {
+                        tracing::error!(
+                            "Job {}: Cannot verify output {:?} after promotion ({}). Source preserved to prevent data loss",
+                            job_id, context.output_path, e
+                        );
+                    }
+                }
+            }
         }
 
         Ok(())

src/media/processor.rs

@@ -450,44 +450,24 @@ impl Agent {
     }
 
     /// Gracefully shutdown the agent.
-    /// Drains active jobs and waits up to `timeout` for them to complete.
-    /// After timeout, forcefully cancels remaining jobs.
-    pub async fn graceful_shutdown(&self, timeout: std::time::Duration) {
-        info!("Initiating graceful shutdown...");
+    /// Cancels active jobs immediately and returns quickly.
+    pub async fn graceful_shutdown(&self) {
+        info!("Initiating rapid shutdown...");
 
         // Stop accepting new jobs
         self.pause();
-        self.drain();
 
-        // Wait for active jobs to complete (with timeout)
-        let start = std::time::Instant::now();
-        let check_interval = std::time::Duration::from_millis(500);
-
-        while start.elapsed() < timeout {
-            let active = self.orchestrator.active_job_count();
-            if active == 0 {
-                info!("All jobs completed gracefully.");
-                return;
-            }
-            info!(
-                "Waiting for {} active job(s) to complete... ({:.0}s remaining)",
-                active,
-                (timeout - start.elapsed()).as_secs_f64()
-            );
-            tokio::time::sleep(check_interval).await;
-        }
-
-        // Timeout reached - force cancel remaining jobs
+        // Immediately force cancel remaining jobs
         let cancelled = self.orchestrator.cancel_all_jobs();
         if cancelled > 0 {
             tracing::warn!(
-                "Shutdown timeout reached. Forcefully cancelled {} job(s).",
+                "Fast shutdown requested. Forcefully cancelled {} job(s).",
                 cancelled
             );
-            // Give FFmpeg processes a moment to terminate
+            // Give FFmpeg processes a moment to terminate and Tokio to flush DB statuses
             tokio::time::sleep(std::time::Duration::from_secs(2)).await;
         }
 
-        info!("Graceful shutdown complete.");
+        info!("Rapid shutdown complete.");
     }
 }

src/notifications.rs

@@ -1,20 +1,23 @@
+use crate::config::Config;
 use crate::db::{AlchemistEvent, Db, NotificationTarget};
 use reqwest::{Client, Url, redirect::Policy};
 use serde_json::json;
 use std::net::IpAddr;
+use std::sync::Arc;
 use std::time::Duration;
 use tokio::net::lookup_host;
-use tokio::sync::broadcast;
+use tokio::sync::{RwLock, broadcast};
 use tracing::{error, warn};
 
 #[derive(Clone)]
 pub struct NotificationManager {
     db: Db,
+    config: Arc<RwLock<Config>>,
 }
 
 impl NotificationManager {
-    pub fn new(db: Db) -> Self {
-        Self { db }
+    pub fn new(db: Db, config: Arc<RwLock<Config>>) -> Self {
+        Self { db, config }
     }
 
     pub fn start_listener(&self, mut rx: broadcast::Receiver<AlchemistEvent>) {
@@ -111,17 +114,28 @@ impl NotificationManager {
             .ok_or("notification endpoint host is missing")?;
         let port = url.port_or_known_default().ok_or("invalid port")?;
 
-        if host.eq_ignore_ascii_case("localhost") {
+        let allow_local = self.config.read().await.notifications.allow_local_notifications;
+
+        if !allow_local && host.eq_ignore_ascii_case("localhost") {
             return Err("localhost is not allowed as a notification endpoint".into());
         }
 
         let addr = format!("{}:{}", host, port);
         let ips = tokio::time::timeout(Duration::from_secs(3), lookup_host(&addr)).await??;
-        let target_ip = ips
-            .into_iter()
+
+        let target_ip = if allow_local {
+            // When local notifications are allowed, accept any resolved IP
+            ips.into_iter()
+                .map(|a| a.ip())
+                .next()
+                .ok_or("no IP address found for notification endpoint")?
+        } else {
+            // When local notifications are blocked, only use public IPs
+            ips.into_iter()
                 .map(|a| a.ip())
                 .find(|ip| !is_private_ip(*ip))
-            .ok_or("no public IP address found for notification endpoint")?;
+                .ok_or("no public IP address found for notification endpoint")?
+        };
 
         // Pin the request to the validated IP to prevent DNS rebinding
         let client = Client::builder()
@@ -324,7 +338,10 @@ mod tests {
         db_path.push(format!("alchemist_notifications_test_{}.db", token));
 
         let db = Db::new(db_path.to_string_lossy().as_ref()).await?;
-        let manager = NotificationManager::new(db);
+        let mut test_config = crate::config::Config::default();
+        test_config.notifications.allow_local_notifications = true;
+        let config = Arc::new(RwLock::new(test_config));
+        let manager = NotificationManager::new(db, config);
 
         let listener = match TcpListener::bind("127.0.0.1:0").await {
             Ok(listener) => listener,

src/orchestrator.rs

@@ -294,8 +294,9 @@ impl Transcoder {
 
         loop {
             tokio::select! {
-                line_res = reader.next_line() => {
-                    match line_res {
+                line_res_timeout = tokio::time::timeout(tokio::time::Duration::from_secs(600), reader.next_line()) => {
+                    match line_res_timeout {
+                        Ok(line_res) => match line_res {
                             Ok(Some(line)) => {
                                 let line = if line.len() > 4096 {
                                     format!("{}...[truncated]", &line[..4096])
@@ -322,6 +323,19 @@ impl Transcoder {
                                 error!("Error reading FFmpeg stderr: {}", e);
                                 break;
                             }
+                        },
+                        Err(_) => {
+                            error!("Job {:?} stalled: No output from FFmpeg for 10 minutes. Killing process...", job_id);
+                            let _ = child.kill().await;
+                            killed = true;
+                            if let Some(id) = job_id {
+                                match self.cancel_channels.lock() {
+                                    Ok(mut channels) => { channels.remove(&id); }
+                                    Err(e) => { e.into_inner().remove(&id); }
+                                }
+                            }
+                            break;
+                        }
                     }
                 }
                 _ = &mut kill_rx => {

src/server/middleware.rs

@@ -235,6 +235,17 @@ pub(crate) fn get_cookie_value(headers: &axum::http::HeaderMap, name: &str) -> O
 }
 
 pub(crate) fn request_ip(req: &Request) -> Option<IpAddr> {
+    let peer_ip = req
+        .extensions()
+        .get::<ConnectInfo<SocketAddr>>()
+        .map(|info| info.0.ip());
+
+    // Only trust proxy headers (X-Forwarded-For, X-Real-IP) when the direct
+    // TCP peer is a loopback or private IP — i.e., a trusted reverse proxy.
+    // This prevents external attackers from spoofing these headers to bypass
+    // rate limiting.
+    if let Some(peer) = peer_ip {
+        if is_trusted_peer(peer) {
             if let Some(xff) = req.headers().get("X-Forwarded-For") {
                 if let Ok(xff_str) = xff.to_str() {
                     if let Some(ip_str) = xff_str.split(',').next() {
@@ -251,7 +262,18 @@ pub(crate) fn request_ip(req: &Request) -> Option<IpAddr> {
                     }
                 }
             }
-    req.extensions()
-        .get::<ConnectInfo<SocketAddr>>()
-        .map(|info| info.0.ip())
+        }
+    }
+
+    peer_ip
+}
+
+/// Returns true if the peer IP is a loopback or private address,
+/// meaning it is likely a local reverse proxy that can be trusted
+/// to set forwarded headers.
+fn is_trusted_peer(ip: IpAddr) -> bool {
+    match ip {
+        IpAddr::V4(v4) => v4.is_loopback() || v4.is_private() || v4.is_link_local(),
+        IpAddr::V6(v6) => v6.is_loopback() || v6.is_unique_local() || v6.is_unicast_link_local(),
+    }
 }

src/server/mod.rs

@@ -84,6 +84,7 @@ pub struct AppState {
     pub resources_cache: Arc<tokio::sync::Mutex<Option<(serde_json::Value, std::time::Instant)>>>,
     pub(crate) login_rate_limiter: Mutex<HashMap<IpAddr, RateLimitEntry>>,
     pub(crate) global_rate_limiter: Mutex<HashMap<IpAddr, RateLimitEntry>>,
+    pub(crate) sse_connections: Arc<std::sync::atomic::AtomicUsize>,
 }
 
 pub struct RunServerArgs {
@@ -164,6 +165,7 @@ pub async fn run_server(args: RunServerArgs) -> Result<()> {
         resources_cache: Arc::new(tokio::sync::Mutex::new(None)),
         login_rate_limiter: Mutex::new(HashMap::new()),
         global_rate_limiter: Mutex::new(HashMap::new()),
+        sse_connections: Arc::new(std::sync::atomic::AtomicUsize::new(0)),
     });
 
     // Clone agent for shutdown handler before moving state into router
@@ -265,9 +267,9 @@ pub async fn run_server(args: RunServerArgs) -> Result<()> {
             }
         }
 
-        // Give active jobs up to 5 minutes to complete
+        // Forceful immediate shutdown of active jobs
         shutdown_agent
-            .graceful_shutdown(std::time::Duration::from_secs(300))
+            .graceful_shutdown()
             .await;
     })
     .await

src/server/sse.rs

@@ -171,9 +171,40 @@ pub(crate) fn sse_unified_stream(
     ])
 }
 
+/// Maximum concurrent SSE connections to prevent resource exhaustion.
+const MAX_SSE_CONNECTIONS: usize = 50;
+
+/// RAII guard that decrements the SSE connection counter on drop.
+struct SseConnectionGuard(Arc<std::sync::atomic::AtomicUsize>);
+
+impl Drop for SseConnectionGuard {
+    fn drop(&mut self) {
+        self.0.fetch_sub(1, std::sync::atomic::Ordering::SeqCst);
+    }
+}
+
 pub(crate) async fn sse_handler(
     State(state): State<Arc<AppState>>,
-) -> Sse<impl Stream<Item = std::result::Result<AxumEvent, Infallible>>> {
+) -> std::result::Result<
+    Sse<impl Stream<Item = std::result::Result<AxumEvent, Infallible>>>,
+    axum::http::StatusCode,
+> {
+    use std::sync::atomic::Ordering;
+
+    // Enforce connection limit
+    let current = state.sse_connections.fetch_add(1, Ordering::SeqCst);
+    if current >= MAX_SSE_CONNECTIONS {
+        state.sse_connections.fetch_sub(1, Ordering::SeqCst);
+        warn!(
+            "SSE connection limit reached ({}/{}). Rejecting new connection.",
+            current, MAX_SSE_CONNECTIONS
+        );
+        return Err(axum::http::StatusCode::TOO_MANY_REQUESTS);
+    }
+
+    // RAII guard to decrement the counter when the stream is dropped
+    let guard = Arc::new(SseConnectionGuard(state.sse_connections.clone()));
+
     // Subscribe to all channels
     let job_rx = state.event_channels.jobs.subscribe();
     let config_rx = state.event_channels.config.subscribe();
@@ -182,10 +213,13 @@ pub(crate) async fn sse_handler(
     // Create unified stream from new typed channels
     let unified_stream = sse_unified_stream(job_rx, config_rx, system_rx);
 
-    let stream = unified_stream.map(|message| match message {
+    let stream = unified_stream.map(move |message| {
+        let _guard = guard.clone(); // keep the guard alive as long as the stream lives
+        match message {
             Ok(message) => Ok(message.into()),
             Err(never) => match never {},
+        }
     });
 
-    Sse::new(stream).keep_alive(axum::response::sse::KeepAlive::default())
+    Ok(Sse::new(stream).keep_alive(axum::response::sse::KeepAlive::default()))
 }

src/server/tests.rs

@@ -107,6 +107,7 @@ where
         telemetry_runtime_id: "test-runtime".to_string(),
         notification_manager: Arc::new(crate::notifications::NotificationManager::new(
             db.as_ref().clone(),
+            config.clone(),
         )),
         sys: Mutex::new(sys),
         file_watcher,
@@ -118,6 +119,7 @@ where
         resources_cache: Arc::new(tokio::sync::Mutex::new(None)),
         login_rate_limiter: Mutex::new(HashMap::new()),
         global_rate_limiter: Mutex::new(HashMap::new()),
+        sse_connections: Arc::new(std::sync::atomic::AtomicUsize::new(0)),
     });
 
     Ok((state.clone(), app_router(state), config_path, db_path))

src/system/fs_browser.rs

@@ -266,25 +266,20 @@ fn preview_blocking(request: FsPreviewRequest) -> Result<FsPreviewResponse> {
             let exists = canonical.exists();
             let readable = exists && canonical.is_dir() && std::fs::read_dir(&canonical).is_ok();
 
-            let media_files = if readable {
-                scanner
-                    .scan_with_recursion(vec![(canonical.clone(), true)])
-                    .len()
-            } else {
-                0
-            };
-            total_media_files += media_files;
-
-            let sample_files = if readable {
-                scanner
-                    .scan_with_recursion(vec![(canonical.clone(), true)])
-                    .into_iter()
-                    .take(5)
-                    .map(|media| media.path.to_string_lossy().to_string())
-                    .collect::<Vec<_>>()
+            // Scan once and reuse results for both count and samples
+            let scan_results = if readable {
+                scanner.scan_with_recursion(vec![(canonical.clone(), true)])
             } else {
                 Vec::new()
             };
+            let media_files = scan_results.len();
+            total_media_files += media_files;
+
+            let sample_files = scan_results
+                .into_iter()
+                .take(5)
+                .map(|media| media.path.to_string_lossy().to_string())
+                .collect::<Vec<_>>();
 
             let mut dir_warnings = directory_warnings(&canonical, readable);
             if readable && media_files == 0 {

web/src/components/Dashboard.tsx

@@ -12,6 +12,7 @@ import { apiJson, isApiError } from "../lib/api";
 import { useSharedStats } from "../lib/statsStore";
 import { showToast } from "../lib/toast";
 import ResourceMonitor from "./ResourceMonitor";
+import { withErrorBoundary } from "./ErrorBoundary";
 
 interface Job {
     id: number;
@@ -76,7 +77,7 @@ function StatCard({ label, value, icon: Icon, colorClass }: StatCardProps) {
     );
 }
 
-export default function Dashboard() {
+function Dashboard() {
     const [jobs, setJobs] = useState<Job[]>([]);
     const [jobsLoading, setJobsLoading] = useState(true);
     const [bundle, setBundle] = useState<SettingsBundleResponse | null>(null);
@@ -366,3 +367,5 @@ export default function Dashboard() {
         </div>
     );
 }
+
+export default withErrorBoundary(Dashboard, "Dashboard");

web/src/components/ErrorBoundary.tsx

@@ -0,0 +1,69 @@
+import React, { Component, type ReactNode } from "react";
+import { AlertCircle } from "lucide-react";
+
+interface Props {
+  children: ReactNode;
+  fallback?: ReactNode;
+  moduleName?: string;
+}
+
+interface State {
+  hasError: boolean;
+  errorMessage: string;
+}
+
+export class ErrorBoundary extends Component<Props, State> {
+  public state: State = {
+    hasError: false,
+    errorMessage: "",
+  };
+
+  public static getDerivedStateFromError(error: Error): State {
+    return { hasError: true, errorMessage: error.message };
+  }
+
+  public componentDidCatch(error: Error, errorInfo: React.ErrorInfo) {
+    console.error("Uncaught error in ErrorBoundary:", error, errorInfo);
+  }
+
+  public render() {
+    if (this.state.hasError) {
+      if (this.props.fallback) {
+        return this.props.fallback;
+      }
+      return (
+        <div className="flex flex-col items-center justify-center p-8 bg-helios-background border border-helios-red/50 rounded-lg shadow-sm text-center w-full min-h-[300px]">
+          <AlertCircle className="w-12 h-12 text-helios-red mb-4" />
+          <h2 className="text-xl font-bold text-white mb-2">Something went wrong</h2>
+          <p className="text-helios-text/70 mb-4 max-w-md">
+            The {this.props.moduleName || "component"} encountered an unexpected error and could not be displayed.
+          </p>
+          <div className="text-xs text-helios-red/80 font-mono bg-helios-red/10 p-4 rounded w-full overflow-auto max-w-lg mb-6 text-left break-words max-h-32">
+            {this.state.errorMessage}
+          </div>
+          <button
+            onClick={() => window.location.reload()}
+            className="px-6 py-2 bg-helios-orange hover:bg-helios-orange/80 text-white font-medium rounded transition"
+          >
+            Reload Page
+          </button>
+        </div>
+      );
+    }
+
+    return this.props.children;
+  }
+}
+
+export const withErrorBoundary = <P extends object>(
+  WrappedComponent: React.ComponentType<P>,
+  moduleName?: string
+) => {
+  return function WithErrorBoundary(props: P) {
+    return (
+      <ErrorBoundary moduleName={moduleName}>
+        <WrappedComponent {...props} />
+      </ErrorBoundary>
+    );
+  };
+};

web/src/components/JobManager.tsx

@@ -11,6 +11,7 @@ import ConfirmDialog from "./ui/ConfirmDialog";
 import { clsx, type ClassValue } from "clsx";
 import { twMerge } from "tailwind-merge";
 import { motion, AnimatePresence } from "framer-motion";
+import { withErrorBoundary } from "./ErrorBoundary";
 
 function cn(...inputs: ClassValue[]) {
     return twMerge(clsx(inputs));
@@ -364,7 +365,7 @@ const SORT_OPTIONS: Array<{ value: SortField; label: string }> = [
     { value: "size", label: "File Size" },
 ];
 
-export default function JobManager() {
+function JobManager() {
     const [jobs, setJobs] = useState<Job[]>([]);
     const [loading, setLoading] = useState(true);
     const [selected, setSelected] = useState<Set<number>>(new Set());
@@ -1769,3 +1770,5 @@ export default function JobManager() {
         </div>
     );
 }
+
+export default withErrorBoundary(JobManager, "Job Management");

web/src/pages/500.astro

@@ -0,0 +1,38 @@
+---
+import Layout from "../layouts/Layout.astro";
+import { AlertTriangle } from "lucide-react";
+
+interface Props {
+  error: unknown;
+}
+
+const { error } = Astro.props;
+---
+
+<Layout title="Alchemist | Server Error">
+    <div class="min-h-screen bg-helios-background flex items-center justify-center p-6 pb-24">
+        <div class="max-w-md w-full flex flex-col items-center text-center">
+            <div class="w-16 h-16 rounded-full bg-helios-red/10 flex items-center justify-center mb-6">
+                <AlertTriangle className="w-8 h-8 text-helios-red" />
+            </div>
+            
+            <h1 class="text-3xl font-bold text-white mb-3">500 Server Error</h1>
+            <p class="text-helios-text mb-8">
+                Alchemist encountered an internal error. Please check the backend logs.
+            </p>
+            
+            {error instanceof Error ? (
+              <div class="bg-black/50 border border-white/5 rounded-lg p-4 mb-8 w-full overflow-auto text-left">
+                  <p class="text-helios-red/90 font-mono text-sm break-words">{error.message}</p>
+              </div>
+            ) : null}
+
+            <a
+                href="/"
+                class="px-6 py-2.5 bg-helios-orange hover:bg-helios-orange/90 text-white font-medium rounded-md transition-colors"
+            >
+                Return to Dashboard
+            </a>
+        </div>
+    </div>
+</Layout>