fix: use router.huggingface.co for quiz (api-inference deprecated)

backend/quiz_service.py

@@ -1,20 +1,15 @@
 """
 Quiz generation service.
-Retrieves chunks from Supabase, sends to HF Inference API, saves artifact.
+Retrieves chunks from Supabase, sends to LLM via Hugging Face router, saves artifact.
 """
 
 import os
 import json
 import re
-import requests
-from backend.db import supabase
-from backend.artifacts_service import create_artifact
-from huggingface_hub import InferenceClient
-
 
-HF_TOKEN = os.environ.get("HF_TOKEN", "")
-HF_MODEL = "meta-llama/Llama-3.2-3B-Instruct"
-HF_API_URL = f"https://api-inference.huggingface.co/models/{HF_MODEL}"
+from backend.artifacts_service import create_artifact
+from backend.db import supabase
+from backend.llm_client import DEFAULT_MODEL, get_llm_client
 
 # Retrieval 
 def _get_chunks_for_notebook(notebook_id: str, limit: int = 15) -> list[str]:
@@ -60,15 +55,15 @@ Format your response as a JSON array only, no preamble:
 Context:
 {context}"""
 
-# HF Inference Call 
+# LLM Call (uses router.huggingface.co via llm_client)
 def _call_hf(prompt: str) -> str:
-    client = InferenceClient(token=HF_TOKEN)
-    response = client.chat_completion(
-        model=HF_MODEL,
+    client = get_llm_client()
+    response = client.chat.completions.create(
+        model=DEFAULT_MODEL,
         messages=[{"role": "user", "content": prompt}],
-        max_tokens=3000,  # ← increase this
+        max_tokens=3000,
     )
-    return response.choices[0].message.content
+    return response.choices[0].message.content or ""
 
 # Main Function 
 def generate_quiz(notebook_id: str, source_type: str = "all", source_id: str = None) -> dict: