Switch generateTitle to use Hugging Face Llama model

Replaces OpenAI GPT-3.5-turbo with Hugging Face's meta-llama/Llama-3.1-8B-Instruct model for the generateTitle function. Updates the import and API call accordingly.

app/actions.ts

@@ -1,6 +1,7 @@
 "use server";
 
 import { openai } from "@/lib/openai-client";
+import { hf } from "@/lib/hf-client";
 import { z } from "zod";
 import { startMcpSandbox } from "@/lib/mcp-sandbox";
 
@@ -48,8 +49,8 @@ export async function generateTitle(messages: any[]): Promise<string> {
     content: Array.isArray(m.content) ? m.content.join("\n") : m.content,
   }));
 
-  const response = await openai.chat.completions.create({
-    model: "gpt-3.5-turbo",
+  const response = await hf.chat.completions.create({
+    model: "meta-llama/Llama-3.1-8B-Instruct",
     max_tokens: 30,
     temperature: 0.4,
     messages: [