fix: improve error handling and stability - no logout on transient errors, add ErrorBoundary, global error logging

frontend/src/api/client.ts

@@ -55,17 +55,39 @@ apiClient.interceptors.request.use(
   (error) => Promise.reject(error)
 );
 
+// Retry interceptor for 5xx errors (max 1 retry with short delay)
+apiClient.interceptors.response.use(
+  (response) => response,
+  async (error: AxiosError) => {
+    const config = error.config as InternalAxiosRequestConfig & { _retried?: boolean };
+    const status = error.response?.status || 0;
+
+    // Retry 5xx errors once with a short delay
+    if (status >= 500 && status < 600 && config && !config._retried) {
+      config._retried = true;
+      await new Promise((resolve) => setTimeout(resolve, 500));
+      return apiClient.request(config);
+    }
+
+    return Promise.reject(error);
+  }
+);
+
+// Main error-handling interceptor
 apiClient.interceptors.response.use(
   (response) => response,
   (error: AxiosError) => {
     const status = error.response?.status || 0;
     const data = error.response?.data as any;
 
+    // Only call onUnauthorized for 401 — NOT for 403 or network errors
     if (status === 401) {
       if (onUnauthorized) {
         onUnauthorized();
       }
     }
+    // 403: Permission error — do NOT logout, let the calling code handle it
+    // Network errors (status === 0): do NOT logout, transient issue
 
     if (status === 422 && data?.detail) {
       const validationErrors: Record<string, string[]> = {};