assistant: reply 'images not supported' and skip images

The cog's vision_model defaulted to the text-only llama3.2:3b, so any image in
the message/reply was sent to a non-vision model, causing Ollama 400 'invalid
image input'. Now the bot tells the user images aren't supported and continues
with text-only, preserving normal chat.

assistant/assistant.py

@@ -138,6 +138,16 @@ class Assistant(commands.Cog):
         if not content and not image_urls:
             return
 
+        # Vision is not supported by the configured text-only model. If the user
+        # sent image(s), tell them once and skip the images, continuing with any
+        # text so normal chat keeps working (avoids Ollama 400 "invalid image
+        # input" from sending images to a non-vision model).
+        if image_urls:
+            await message.reply("Images not supported at this time.")
+            image_urls = []
+            if not content:
+                return
+
         if question_mode and not content.rstrip().endswith("?"):
             if is_ping:
                 await message.reply("Question mode is on - messages must end with **?**.")