Fix: Move n_ctx parameter to model setup!

utils.py

@@ -35,7 +35,7 @@ else:
 
 if in_memory_llm is None and USE_HTTP_SERVER is False:
     print("Loading model into memory. If you didn't want this, set the USE_HTTP_SERVER environment variable to 'true'.")
-    in_memory_llm = Llama(model_path=LLM_MODEL_PATH)
+    in_memory_llm = Llama(model_path=LLM_MODEL_PATH, n_ctx=4096)
 
 def llm_streaming(
     prompt: str, pydantic_model_class, return_pydantic_object=False
@@ -117,7 +117,6 @@ def llm_stream_sans_network(
 
     stream = in_memory_llm(
         prompt,
-        n_ctx=4096,
         max_tokens=1000,
         temperature=0.7,
         grammar=grammar,