assistant: add local Ollama backend for the humor chatbot

Make the API base URL configurable (new global config base_url, default to the
local Ollama OpenAI-compatible endpoint http://127.0.0.1:11434/v1 on gpu1).
Local endpoints (127.0.0.1/localhost) need no API key: the on_message, _chat,
and models guards allow a missing key when local, and the Authorization header
is only sent when a key is set. Default guild model is now
llama3.2:3b-instruct-q4_K_M with a humor-focused system prompt. Adds
[p]assistant baseurl to switch endpoints (omit to reset to local), shows the
backend in settings, and raises the request timeout to 120s for local
inference. GreenPT remains usable as a fallback by setting the base URL back to
it and providing a key.

Note: Llama-3.2-3B-Instruct is text-only; the vision path needs a
vision-capable model (e.g. a local llava/llama3.2-vision) or GreenPT.

assistant/assistant.py

@@ -35,7 +35,11 @@ class APIKeyButton(discord.ui.View):
 class Assistant(commands.Cog):
     """AI chatbot assistant. Set a chat channel or get responses when the bot is pinged."""
 
-    BASE_URL = "https://api.greenpt.ai/v1"
+    # Default backend: the local Ollama server (OpenAI-compatible) on gpu1.
+    # Override per deployment with [p]assistant baseurl. GreenPT
+    # (https://api.greenpt.ai/v1) remains usable as a fallback by setting the
+    # base URL back to it and providing an API key.
+    DEFAULT_BASE_URL = "http://127.0.0.1:11434/v1"
 
     def __init__(self, bot: Red):
         self.bot = bot
@@ -44,17 +48,27 @@ class Assistant(commands.Cog):
         self.config.register_guild(
             enabled=True,
             chat_channel=None,
-            model="gemma4",
+            model="llama3.2:3b-instruct-q4_K_M",
-            vision_model="gemma4",
+            vision_model="llama3.2:3b-instruct-q4_K_M",
             temperature=0.7,
-            system_prompt="You are a helpful assistant.",
+            system_prompt=(
+                "You are a witty, quick-humored chatbot in a Discord server. "
+                "Keep replies short, punchy, and funny. Favor clever, playful "
+                "banter over long explanations. Stay good-natured; never be "
+                "genuinely mean or offensive."
+            ),
             question_mode=False,
             max_tokens=None,
             reasoning_effort=None,
         )
-        self.config.register_global(api_key=None)
+        self.config.register_global(api_key=None, base_url=self.DEFAULT_BASE_URL)
         self.history: dict[int, list[dict]] = {}
 
+    @staticmethod
+    def _is_local(base_url: str) -> bool:
+        """A local endpoint (Ollama) does not require an API key."""
+        return "127.0.0.1" in base_url or "localhost" in base_url
+
     async def cog_load(self):
         self.session = aiohttp.ClientSession()
 
@@ -79,8 +93,10 @@ class Assistant(commands.Cog):
         if not enabled:
             return
 
+        # A local (Ollama) endpoint needs no API key; a remote one does.
+        base_url = await self.config.base_url()
         api_key = await self.config.api_key()
-        if not api_key:
+        if not self._is_local(base_url) and not api_key:
             return
 
         chat_channel_id = await self.config.guild(guild).chat_channel()
@@ -124,7 +140,7 @@ class Assistant(commands.Cog):
 
         if question_mode and not content.rstrip().endswith("?"):
             if is_ping:
-                await message.reply("Question mode is on — messages must end with **?**.")
+                await message.reply("Question mode is on - messages must end with **?**.")
             return
 
         async with message.channel.typing():
@@ -150,8 +166,11 @@ class Assistant(commands.Cog):
     # ---------------------------------------------------------------------------
 
     async def _chat(self, guild: discord.Guild, channel_id: int, user_msg: str, image_urls: list[str] | None = None) -> str | None:
+        if not self.session:
+            return None
+        base_url = await self.config.base_url()
         api_key = await self.config.api_key()
-        if not api_key or not self.session:
+        if not self._is_local(base_url) and not api_key:
             return None
 
         model = await self.config.guild(guild).model()
@@ -196,15 +215,16 @@ class Assistant(commands.Cog):
         if reasoning_effort:
             payload["reasoning_effort"] = reasoning_effort
 
+        headers = {"Content-Type": "application/json"}
+        if api_key:
+            headers["Authorization"] = f"Bearer {api_key}"
+
         try:
             async with self.session.post(
-                f"{self.BASE_URL}/chat/completions",
+                f"{base_url}/chat/completions",
                 json=payload,
-                headers={
+                headers=headers,
-                    "Authorization": f"Bearer {api_key}",
+                timeout=aiohttp.ClientTimeout(total=120),
-                    "Content-Type": "application/json",
-                },
-                timeout=aiohttp.ClientTimeout(total=60),
             ) as resp:
                 if resp.status != 200:
                     body = await resp.text()
@@ -246,6 +266,23 @@ class Assistant(commands.Cog):
         view = APIKeyButton(self.config)
         await ctx.send("Click below to set the API key:", view=view)
 
+    @assistant.command(name="baseurl")
+    @checks.is_owner()
+    async def assistant_baseurl(self, ctx: commands.Context, url: str | None = None):
+        """Set the API base URL. Omit to reset to the local Ollama endpoint.
+
+        Examples:
+          [p]assistant baseurl                            (reset to local Ollama)
+          [p]assistant baseurl https://api.greenpt.ai/v1  (use GreenPT)
+        A local endpoint (127.0.0.1/localhost) needs no API key.
+        """
+        if url is None:
+            url = self.DEFAULT_BASE_URL
+        url = url.rstrip("/")
+        await self.config.base_url.set(url)
+        note = " (local, no API key needed)" if self._is_local(url) else " (remote, API key required)"
+        await ctx.send(f"Base URL set to `{url}`{note}.")
+
     @assistant.command(name="toggle")
     async def assistant_toggle(self, ctx: commands.Context):
         """Toggle the assistant on or off for this server."""
@@ -266,7 +303,7 @@ class Assistant(commands.Cog):
 
     @assistant.command(name="model")
     async def assistant_model(self, ctx: commands.Context, model: str):
-        """Set the model (e.g. gemma4, green-l, green-r)."""
+        """Set the model. Local (Ollama): llama3.2:3b-instruct-q4_K_M. Remote (GreenPT): gemma4, green-l."""
         await self.config.guild(ctx.guild).model.set(model)
         await ctx.send(f"Model set to **{model}**.")
 
@@ -330,16 +367,21 @@ class Assistant(commands.Cog):
             "qwen3-embedding-8b",
         }
 
+        base_url = await self.config.base_url()
         api_key = await self.config.api_key()
-        if not api_key or not self.session:
+        if not self.session or (not self._is_local(base_url) and not api_key):
             await ctx.send("API key not set. Use `[p]assistant setkey` first.")
             return
 
+        headers = {}
+        if api_key:
+            headers["Authorization"] = f"Bearer {api_key}"
+
         # Fetch live model list from API
         try:
             async with self.session.get(
-                f"{self.BASE_URL}/models",
+                f"{base_url}/models",
-                headers={"Authorization": f"Bearer {api_key}"},
+                headers=headers,
                 timeout=aiohttp.ClientTimeout(total=15),
             ) as resp:
                 if resp.status != 200:
@@ -376,10 +418,10 @@ class Assistant(commands.Cog):
             marker = " ◀" if model_id == current_model else ""
             if pricing:
                 lines.append(
-                    f"`{model_id}` — €{pricing['input']:.2f} in / €{pricing['output']:.2f} out{marker}"
+                    f"`{model_id}` - €{pricing['input']:.2f} in / €{pricing['output']:.2f} out{marker}"
                 )
             else:
-                lines.append(f"`{model_id}` — pricing unknown{marker}")
+                lines.append(f"`{model_id}` - pricing unknown{marker}")
 
         embed = discord.Embed(
             title="Available Models (cheapest first)",
@@ -420,7 +462,7 @@ class Assistant(commands.Cog):
         valid_levels = ("none", "minimal", "low", "medium", "high")
         if level is None:
             await self.config.guild(ctx.guild).reasoning_effort.set(None)
-            await ctx.send("Reasoning effort cleared (model default — thinking enabled).")
+            await ctx.send("Reasoning effort cleared (model default - thinking enabled).")
         elif level.lower() not in valid_levels:
             await ctx.send(f"Invalid level. Choose from: {', '.join(valid_levels)}")
         else:
@@ -473,6 +515,7 @@ class Assistant(commands.Cog):
         q_mode = await cfg.question_mode()
         max_tokens = await cfg.max_tokens()
         reasoning_effort = await cfg.reasoning_effort()
+        base_url = await self.config.base_url()
         has_key = await self.config.api_key() is not None
 
         channel_str = f"<#{chat_channel}>" if chat_channel else "None"
@@ -481,6 +524,7 @@ class Assistant(commands.Cog):
         embed.add_field(name="Enabled", value=str(enabled), inline=True)
         embed.add_field(name="Chat Channel", value=channel_str, inline=True)
         embed.add_field(name="API Key", value="Set" if has_key else "Not set", inline=True)
+        embed.add_field(name="Backend", value=base_url, inline=False)
         embed.add_field(name="Model", value=model, inline=True)
         embed.add_field(name="Vision Model", value=vision_model, inline=True)
         embed.add_field(name="Temperature", value=str(temp), inline=True)