fix: pass LiteLLM credentials when flushing model cache

When refreshing LiteLLM models, the cache flush was not receiving the
required apiKey and baseUrl parameters, causing an 'Invalid URL' error.

This fix:
- Updates flushModels() to accept optional GetModelsOptions parameter
- Passes LiteLLM credentials when flushing cache in both places:
  1. When explicitly refreshing via flushRouterModels message
  2. When requesting models with new credentials

Fixes #9682

src/api/providers/fetchers/modelCache.ts

@@ -273,13 +273,20 @@ export async function initializeModelCacheRefresh(): Promise<void> {
  *
  * @param router - The router to flush models for.
  * @param refresh - If true, immediately fetch fresh data from API
+ * @param options - Optional provider options (e.g., apiKey, baseUrl) needed for certain providers
  */
-export const flushModels = async (router: RouterName, refresh: boolean = false): Promise<void> => {
+export const flushModels = async (
+	router: RouterName,
+	refresh: boolean = false,
+	options?: GetModelsOptions,
+): Promise<void> => {
 	if (refresh) {
 		// Don't delete memory cache - let refreshModels atomically replace it
 		// This prevents a race condition where getModels() might be called
 		// before refresh completes, avoiding a gap in cache availability
-		refreshModels({ provider: router } as GetModelsOptions).catch((error) => {
+		// Use provided options if available, otherwise fallback to minimal options
+		const refreshOptions = options || ({ provider: router } as GetModelsOptions)
+		refreshModels(refreshOptions).catch((error) => {
 			console.error(`[flushModels] Refresh failed for ${router}:`, error)
 		})
 	} else {

src/core/webview/webviewMessageHandler.ts

@@ -876,7 +876,12 @@ export const webviewMessageHandler = async (
 				// If explicit credentials are provided in message.values (from Refresh Models button),
 				// flush the cache first to ensure we fetch fresh data with the new credentials
 				if (message?.values?.litellmApiKey || message?.values?.litellmBaseUrl) {
-					await flushModels("litellm", true)
+					const litellmFlushOptions: GetModelsOptions = {
+						provider: "litellm",
+						apiKey: litellmApiKey,
+						baseUrl: litellmBaseUrl,
+					}
+					await flushModels("litellm", true, litellmFlushOptions)
 				}
 
 				candidates.push({