feat(api): add pagination to messages endpoint and improve chat API robustness

- Add pagination support to /api/conversations/[id]/messages with limit and before query params, returning hasMore and cursor for efficient message fetching
- Enhance /api/ollama/chat by adding non-blocking Ollama availability check and conditionally disabling debug logging in production环境下Featured

app/api/ollama/models/route.ts

@@ -10,6 +10,10 @@ import {
 
 export const dynamic = "force-dynamic";
 
+// Cache leve em memória para reduzir chamadas repetidas em curto intervalo
+let __modelsCache: { ts: number; payload: any } | null = null;
+const CACHE_TTL_MS = 8000; // 8s
+
 function normalizeModel(model: any): ModelInfo {
   const name: string = model?.model || model?.name || "unknown";
   const sizeBytes: number = model?.size || 0;
@@ -30,6 +34,11 @@ function normalizeModel(model: any): ModelInfo {
 
 export async function GET() {
   try {
+    // Tentar retorno via cache recente
+    if (__modelsCache && Date.now() - __modelsCache.ts < CACHE_TTL_MS) {
+      return NextResponse.json(__modelsCache.payload);
+    }
+
     let models: any[] = [];
     // Tentar garantir disponibilidade rapidamente (não bloquear longamente a UI)
     await ensureOllamaAvailable({ timeoutMs: 3000 }).catch(() => {});
@@ -39,8 +48,12 @@ export async function GET() {
       models = await listModelsViaHttp();
     }
     const normalized = models.map((m) => normalizeModel(m));
-    return NextResponse.json({ offline: false, models: normalized });
+    const payload = { offline: false, models: normalized };
+    __modelsCache = { ts: Date.now(), payload };
+    return NextResponse.json(payload);
   } catch (e) {
-    return NextResponse.json({ offline: true, models: mockModels });
+    const payload = { offline: true, models: mockModels };
+    __modelsCache = { ts: Date.now(), payload };
+    return NextResponse.json(payload);
   }
 }