fix(flags, providers): update default model and provider; adjust rate limit for no-login usage

lib/flags.ts

@@ -11,17 +11,17 @@ export const DISABLE_AUTH = EMBED_MODE || (process.env.NEXT_PUBLIC_DISABLE_AUTH
 // Choose a responsive general-purpose coding model from the existing list
 export const DEFAULT_MODEL = process.env.NEXT_PUBLIC_DEFAULT_MODEL
   || process.env.DEFAULT_MODEL
-  || "moonshotai/Kimi-K2-Instruct-0905"; // fast on Groq if configured
+  || "deepseek-ai/DeepSeek-V3.1"; // budget-strong default via Novita
 
 export const DEFAULT_PROVIDER = process.env.NEXT_PUBLIC_DEFAULT_PROVIDER
   || process.env.DEFAULT_PROVIDER
-  || "groq"; // fallback; will be validated against model/providers list
+  || "novita"; // validated against model/providers list
 
 // Anonymous/IP rate limit for no-login usage
 export const MAX_REQUESTS_PER_IP_ENV = Number(
   process.env.NEXT_PUBLIC_MAX_REQUESTS_PER_IP
   || process.env.MAX_REQUESTS_PER_IP
-  || 2
+  || 3
 );
 
 // Hard cap on model output tokens to control cost in demo/embed mode

lib/providers.ts

@@ -77,7 +77,7 @@ export const MODELS = [
     label: "DeepSeek V3.1",
     providers: ["fireworks-ai", "novita"],
     isNew: true,
-    autoProvider: "fireworks-ai",
+    autoProvider: "novita",
   },
   {
     value: "moonshotai/Kimi-K2-Instruct-0905",