Updated Gemma chat-template usage

app.py

@@ -192,9 +192,9 @@ def generate_recording_agent_reply(chat_history: List[Tuple[str, str]]) -> str:
     )
     pipe = get_textgen_pipeline()
     tokenizer = pipe.tokenizer
+    combined_prompt = system_prompt + "\n\n" + user_prompt
     messages = [
-        {"role": "system", "content": system_prompt},
-        {"role": "user", "content": user_prompt},
+        {"role": "user", "content": combined_prompt},
     ]
     prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
     gen = pipe(
@@ -231,9 +231,9 @@ def scoring_agent_infer(chat_history: List[Tuple[str, str]], features: Dict[str,
     )
     pipe = get_textgen_pipeline()
     tokenizer = pipe.tokenizer
+    combined_prompt = system_prompt + "\n\n" + user_prompt
     messages = [
-        {"role": "system", "content": system_prompt},
-        {"role": "user", "content": user_prompt},
+        {"role": "user", "content": combined_prompt},
     ]
     prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
     gen = pipe(