ai/cmd/ai-app/main.go

@@ -202,6 +202,9 @@ func build(pool *pgxpool.Pool, cfg config, prov provider.Provider, workerID stri
 	return wiring{mux: mux, queue: q, worker: worker, quotaStore: quotaStore}
 }
 
+// Muss ueber dem Provider-Timeout (120s) liegen, sonst kappt der Server langsame Inferenz-Antworten.
+const serverWriteTimeout = 130 * time.Second
+
 func main() {
 	cfg, err := loadConfig(os.Getenv)
 	if err != nil {
@@ -226,7 +229,7 @@ func main() {
 		Handler:           w.mux,
 		ReadHeaderTimeout: 5 * time.Second,
 		ReadTimeout:       15 * time.Second,
-		WriteTimeout:      15 * time.Second,
+		WriteTimeout:      serverWriteTimeout,
 		IdleTimeout:       60 * time.Second,
 	}
 

ai/cmd/ai-app/main_test.go

@@ -87,6 +87,12 @@ func TestLoadConfig_Defaults(t *testing.T) {
 	}
 }
 
+func TestServerWriteTimeoutCoversProviderTimeout(t *testing.T) {
+	if serverWriteTimeout <= 120*time.Second {
+		t.Fatalf("serverWriteTimeout = %v, muss ueber dem Provider-Timeout von 120s liegen", serverWriteTimeout)
+	}
+}
+
 func newTestPool(t *testing.T) *pgxpool.Pool {
 	t.Helper()
 	dsn := os.Getenv("TEST_TENANT_DSN")