fix: align all UI model documentation strings with Tencent Hy3 + Gemini 3.1 Flash backend

RAG_FULL_APPLICATION_FRONTEND/src/pages/AdvancedRagPage.jsx

@@ -12,7 +12,7 @@ export default function AdvancedRagPage() {
   const [graphData, setGraphData] = useState(null);
   const [graphLoading, setGraphLoading] = useState(false);
   const [inputText, setInputText] = useState(
-    "Artificial Intelligence and Machine Learning models rely on Vector Embeddings and Supabase pgvector storage. HyDE query expansion uses Large Language Models like Qwen and GLM-4.7-Flash to generate synthetic document responses for reciprocal rank fusion retrieval."
+    "Artificial Intelligence and Machine Learning models rely on Vector Embeddings and Supabase pgvector storage. HyDE query expansion uses Large Language Models like Tencent Hy3 and Gemini 3.1 Flash to generate synthetic document responses for reciprocal rank fusion retrieval."
   );
 
   // Guardrails State

RAG_FULL_APPLICATION_FRONTEND/src/pages/FundamentalsPage.jsx

@@ -7,8 +7,8 @@ export default function FundamentalsPage() {
     { title: '1. Knowledge Ingestion', desc: 'Parsing PDF, TXT, DOCX, CSV, PNG documents and tokenizing text into discrete chunks.' },
     { title: '2. Vector Indexing', desc: 'Generating 1024-dim dense embeddings using bge-m3 and indexing in Supabase pgvector & local BM25.' },
     { title: '3. Strategic Retrieval', desc: 'Executing Hybrid Search (BM25 + Vector RRF), HyDE, Re-ranking, or Agentic RAG to fetch context chunks.' },
-    { title: '4. Augmented Generation', desc: 'Dispatching formatted context prompts to Qwen (Primary) or GLM-4.7-Flash (Backup Failover).' },
-    { title: '5. Quality Evaluation (RAGAS)', desc: 'Evaluating Faithfulness, Relevancy, Precision, and Recall using GLM-4.7-Flash as LLM Judge.' }
+    { title: '4. Augmented Generation', desc: 'Dispatching formatted context prompts to Tencent Hy3 (Primary) with Gemini 3.1 Flash backup.' },
+    { title: '5. Quality Evaluation (RAGAS)', desc: 'Evaluating Faithfulness, Relevancy, Precision, and Recall using Gemini 3.1 Flash as LLM Judge.' }
   ];
 
   return (