Implementamos stream de la respuesta.

app.py

@@ -36,7 +36,7 @@ def initLLM():
     """
     llm = ChatGroq(
         model="mixtral-8x7b-32768",
-        temperature=0,
+        temperature=0.7,
         max_tokens=None,
         timeout=None,
         max_retries=2,
@@ -51,12 +51,14 @@ def respond(message, history):
     response = ""
 
     try:
-        response = llm_chain.invoke({"input": message})
+        response = llm_chain.stream({"input": message})
     except:
         raise gradio.Error("Se ha producido un error al interactuar con el modelo LLM.", duratio=5)
 
-    print(response)
-    return response.content
+    partial_message = ""
+    for chunk in response:
+        partial_message = partial_message + chunk.dict()['content']
+        yield partial_message