more updates

app.py

@@ -40,7 +40,7 @@ def generate_llm_response(text, model_id="ccibeekeoc42/Llama3.1-8b-base-SFT-2024
         chat_completion = client.chat.completions.create(
             model="tgi",
             messages=[
-                {"role": "system", "content": "You are a BRIEF AND DIRECT assistant. A part of a speech pipeline so keep your responses short, fluent, and straight to the point. Avoid markdown in responses."},
+                {"role": "system", "content": "You are a very BRIEF AND DIRECT assistant. As part of a speech pipeline so keep your responses short (under 60 words), fluent, and straight to the point. Avoid markdown or digits in responses."},
                 {"role": "user", "content": text}
             ],
             top_p=0.3,