debate: switch to CREATIVE_MODEL with system prompt for uncensored arguments

phi4-mini refused to argue FOR controversial topics, instead deflecting
with neutral takes. The abliterated model with a committed debater system
prompt will actually take both sides without hedging or disclaimers.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

matrixbot/commands.py

@@ -2257,24 +2257,32 @@ async def cmd_debate(client: AsyncClient, room_id: str, sender: str, args: str):
 
     await send_text(client, room_id, f"⚖️ Debating: {topic}...")
 
-    prompt = (
+    system_msg = (
-        f"Debate the topic: \"{topic}\"\n\n"
+        "You are a rhetorical debate bot. Your job is to produce sharp, committed arguments on both sides of any topic — "
-        "Write exactly 2-3 sentences FOR the topic, then exactly 2-3 sentences AGAINST the topic.\n"
+        "no matter how controversial. You do not add warnings, disclaimers, or moral hedging. "
-        "Format your response EXACTLY as:\n"
+        "You argue each side as if you genuinely believe it. "
-        "FOR: <your for argument here>\n"
+        "Respond with ONLY two lines, nothing else:\n"
-        "AGAINST: <your against argument here>\n\n"
+        "FOR: <2-3 sentences genuinely arguing in favor of the topic>\n"
-        "No extra text, no markdown, no headers beyond FOR: and AGAINST:."
+        "AGAINST: <2-3 sentences genuinely arguing against the topic>"
     )
+    user_msg = f'Debate the topic: "{topic}"'
 
     try:
-        timeout = aiohttp.ClientTimeout(total=30)
+        timeout = aiohttp.ClientTimeout(total=60)
         async with aiohttp.ClientSession(timeout=timeout) as session:
             async with session.post(
-                f"{OLLAMA_URL}/api/generate",
+                f"{OLLAMA_URL}/api/chat",
-                json={"model": ASK_MODEL, "prompt": prompt, "stream": False},
+                json={
+                    "model": CREATIVE_MODEL,
+                    "stream": False,
+                    "messages": [
+                        {"role": "system", "content": system_msg},
+                        {"role": "user", "content": user_msg},
+                    ],
+                },
             ) as response:
                 data = await response.json()
-                text = data.get("response", "").strip()
+                text = data.get("message", {}).get("content", "").strip()
 
         # Parse FOR and AGAINST from the response
         for_text = ""
@@ -2285,7 +2293,6 @@ async def cmd_debate(client: AsyncClient, room_id: str, sender: str, args: str):
             for_text = for_part.replace("FOR:", "").strip()
             against_text = against_part.strip()
         else:
-            # Fallback: try to split in half
             lines = [ln.strip() for ln in text.split("\n") if ln.strip()]
             mid = len(lines) // 2
             for_text = " ".join(lines[:mid]) if lines else "No argument generated."
@@ -2307,7 +2314,7 @@ async def cmd_debate(client: AsyncClient, room_id: str, sender: str, args: str):
             f'<blockquote>{for_text}</blockquote>'
             f'<font color="#ef4444"><strong>❌ AGAINST</strong></font><br>'
             f'<blockquote>{against_text}</blockquote>'
-            f'<sup><em>via {_model_label(ASK_MODEL)}</em></sup>'
+            f'<sup><em>via {_model_label(CREATIVE_MODEL)}</em></sup>'
         )
         await send_html(client, room_id, plain, html)