Update app.py

app.py

@@ -14,9 +14,9 @@ load_dotenv()
 Settings.llm = HuggingFaceInferenceAPI(
     model_name="google/gemma-1.1-7b-it",
     tokenizer_name="google/gemma-1.1-7b-it",
-    context_window=3900,
+    context_window=3000,
     token=os.getenv("HF_TOKEN"),
-    max_new_tokens=1000,
+    max_new_tokens=512,
     generate_kwargs={"temperature": 0.1},
 )
 Settings.embed_model = HuggingFaceEmbedding(