fix: pass Teto system prompt via Vertex system_instruction

cogs/neru_teto_cog.py

@@ -87,15 +87,18 @@ class DmbotTetoCog(commands.Cog):
             "When expressing emotions, never use emojis and instead use text-based emoticons like :) and ;) as well as Japanese-style emoticons, for example >~<, ^~^, >///<, UwU, o.O, and OwO over emojis. \n"
             "Reply to the user in a short, conversational manner, staying in character."
         )
-        contents = [types.Content(role="system", parts=[types.Part(text=system_prompt)])]
+        contents = []
         for msg in messages:
             role = "user" if msg.get("role") == "user" else "model"
-            contents.append(types.Content(role=role, parts=[types.Part(text=msg.get("content", ""))]))
+            contents.append(
+                types.Content(role=role, parts=[types.Part(text=msg.get("content", ""))])
+            )
 
         generation_config = types.GenerateContentConfig(
             temperature=1.0,
             max_output_tokens=2000,
             safety_settings=STANDARD_SAFETY_SETTINGS,
+            system_instruction=types.Content(role="system", parts=[types.Part(text=system_prompt)]),
         )
 
         try: