fix: fix auth provider caching

docker-compose.yaml

@@ -211,9 +211,13 @@ services:
   ui:
     image: ${REGISTRY:-dograhai}/dograh-ui:latest
     environment:
+      # Bind the Next.js standalone server to all interfaces
+      HOSTNAME: "0.0.0.0"
+
       # Server-side URL (SSR, internal Docker network)
       BACKEND_URL: "${BACKEND_URL:-http://api:8000}"
       NODE_ENV: "oss"
+
       # Flag to enable/ disable posthog
       ENABLE_TELEMETRY: "${ENABLE_TELEMETRY:-true}"
 
@@ -229,7 +233,7 @@ services:
       test:
         [
           "CMD-SHELL",
-          "wget --no-verbose --tries=1 --spider http://localhost:3010 || exit 1",
+          "wget --no-verbose --tries=1 --spider http://127.0.0.1:3010 || exit 1",
         ]
       interval: 30s
       timeout: 10s

ui/src/lib/auth/config.ts

@@ -49,11 +49,14 @@ async function resolveAuthConfig(): Promise<ResolvedAuthConfig> {
       return cachedConfig;
     }
   } catch {
-    // Backend not reachable — fall back to local
+    // Backend not reachable — fall through without caching so we retry next request.
   }
 
-  cachedConfig = { authProvider: "local", stackConfig: null };
-  return cachedConfig;
+  // Unknown (backend unreachable). Return the local fallback for THIS request but
+  // do NOT cache it: caching here would pin the entire UI to local auth until a
+  // container restart if the first resolution loses the startup race with the api
+  // service. Leaving it uncached means the next request retries and self-heals.
+  return { authProvider: "local", stackConfig: null };
 }
 
 /**

ui/src/middleware.ts

@@ -20,15 +20,23 @@ async function fetchAuthProvider(): Promise<string> {
     const res = await fetch(`${backendUrl}/api/v1/health`);
     if (res.ok) {
       const data = await res.json();
+      // Only cache a DEFINITIVE answer from the backend. Never cache a failure:
+      // this is a module-scoped cache with no TTL, so a single early request
+      // during container startup (before the api service is reachable) would
+      // otherwise poison it to 'local' for the life of the worker — redirecting
+      // every Stack user to the local /auth/login form even though the backend
+      // reports `stack`.
       cachedAuthProvider = (data.auth_provider as string) || 'local';
       return cachedAuthProvider;
     }
   } catch {
-    // Backend not reachable — fall back to local
+    // Backend not reachable — fall through without caching so we retry next request.
   }
 
-  cachedAuthProvider = 'local';
-  return cachedAuthProvider;
+  // Provider unknown (backend unreachable). Return a non-'local' sentinel so the
+  // middleware does NOT guard/redirect: assuming 'local' here would bounce Stack
+  // users to /auth/login. Deliberately not cached — the next request retries.
+  return 'unknown';
 }
 
 export async function middleware(request: NextRequest) {