through divine intervention it was revealed to me a couple of critical bugs and with the help of Claude + an updated model I've finally achieved faster and smarter responses

bot.py

@@ -1,5 +1,6 @@
 #!/usr/bin/env python3
 import os
+import re  # thank you claude
 import asyncio
 import json
 import time
@@ -157,6 +158,14 @@ def steal_media(text: str):
     with MEDIA_FILE.open("a", encoding="utf-8") as f:
         f.write(text + "\n")
 
+def resolve_mentions(content: str, guild) -> str:  # thank you claude
+    if not guild:
+        return content
+    def replace(m):
+        member = guild.get_member(int(m.group(1)))
+        return f"@{member.display_name}" if member else m.group(0)
+    return re.sub(r'<@!?(\d+)>', replace, content)
+
 def is_member_in_server(member: discord.Member) -> bool:
     guild = client.get_guild(server)
     return guild and guild.get_member(member.id) is not None
@@ -316,23 +325,31 @@ async def on_message(message):
     if str(message.author.id) in block_id:
          return
 
+    # drop guild messages outside allowed channels before doing anything else, thank you claude
+    if message.guild and message.channel.name not in ["general", "announcements"]:
+        return
+
+    content = resolve_mentions(message.content, message.guild)
+
     # always detect replies
     replied_message = None
     replied_to_bot = False
+    replied_content = ""
     if message.reference:
         try:
             replied_message = message.reference.resolved or await message.channel.fetch_message(message.reference.message_id)
             if replied_message:
                 if replied_message.author == client.user:
                     replied_to_bot = True
-                log_message(f'[{timestamp()}] {message.author} replied to {replied_message.author} saying: "{replied_message.content}" with reply: "{message.content}"')
+                replied_content = resolve_mentions(replied_message.content, message.guild)
+                log_message(f'[{timestamp()}] {message.author} replied to {replied_message.author} saying: "{replied_content}" with reply: "{content}"')
             else:
-                log_message(f'[{timestamp()}] {message.author} said: {message.content}')
+                log_message(f'[{timestamp()}] {message.author} said: {content}')
         except Exception as e:
             log_error(f"[{timestamp()}] failed to resolve reply: {e}")
-            log_message(f'[{timestamp()}] {message.author} said: {message.content}')
+            log_message(f'[{timestamp()}] {message.author} said: {content}')
     else:
-        log_message(f'[{timestamp()}] {message.author} said: {message.content}')
+        log_message(f'[{timestamp()}] {message.author} said: {content}')
 
     interaction_count += 1
     rebuild = (interaction_count % 150 == 0)
@@ -342,9 +359,9 @@ async def on_message(message):
             return
         dm_sent = False
         if replied_message:
-            user_input = f'{message.author} replied to {replied_message.author} saying: "{replied_message.content}" with reply: "{message.content}"'
+            user_input = f'{message.author} replied to {replied_message.author} saying: "{replied_content}" with reply: "{content}"'
         else:
-            user_input = f"{message.author} said in DM: {message.content}"
+            user_input = f"{message.author} said in DM: {content}"
         payload = build_payload(system_prompt, user_input, rebuild)
         stop_typing = asyncio.Event()
         asyncio.create_task(keep_typing(message.channel, stop_typing))
@@ -355,16 +372,13 @@ async def on_message(message):
             return
         try:
             await message.reply(result)
-            print(f"[{timestamp()}] DM from {message.author}: {message.content} Bot replied in DM: {result}")
-            log_message(f"[{timestamp()}] DM from {message.author}: {message.content}")
+            print(f"[{timestamp()}] DM from {message.author}: {content} Bot replied in DM: {result}")
+            log_message(f"[{timestamp()}] DM from {message.author}: {content}")
             log_message(f"[{timestamp()}] Faggot-chan replied in DM: {result}")
         except Exception as e:
             await handle_error(f"[{timestamp()}] failed to send DM reply: {e}", notify_user=message.author)
         return
 
-    if message.channel.name not in ["general", "announcements"]:
-        return
-
     mentioned = client.user.mentioned_in(message)
 
     # if it replied to bot, treat it as a mention automatically
@@ -375,9 +389,9 @@ async def on_message(message):
         return
 
     if replied_message:
-        user_input = f'{message.author} replied to {replied_message.author} saying: "{replied_message.content}" with reply: "{message.content}"'
+        user_input = f'{message.author} replied to {replied_message.author} saying: "{replied_content}" with reply: "{content}"'
     else:
-        user_input = f"{message.author} said: {message.content}"
+        user_input = f"{message.author} said: {content}"
 
     payload = build_payload(system_prompt, user_input, rebuild)
     stop_typing = asyncio.Event()
@@ -390,8 +404,8 @@ async def on_message(message):
 
     try:
         await message.reply(result)
-        print(f"[{timestamp()}] {message.author} said: {message.content}\n[{timestamp()}] Faggot-chan said: {result}")
-        log_message(f"[{timestamp()}] {message.author} said: {message.content}\n[{timestamp()}] Faggot-chan said: {result}")
+        print(f"[{timestamp()}] {message.author} said: {content}\n[{timestamp()}] Faggot-chan said: {result}")
+        log_message(f"[{timestamp()}] {message.author} said: {content}\n[{timestamp()}] Faggot-chan said: {result}")
     except Exception as e:
         await handle_error(f"[{timestamp()}] failed to send reply: {e}", notify_channel=message.channel)
 

main.py

@@ -1,20 +1,15 @@
 #!/usr/bin/env python3
 import asyncio
 import json
+import re  # thank you claude
 import ollama
 import websockets
 from websockets.server import serve
 from datetime import datetime
 
-chatmodel = "hf.co/Orenguteng/Llama-3.1-8B-Lexi-Uncensored-V2-GGUF:Llama-3.1-8B-Lexi-Uncensored_V2_Q4.gguf"
+chatmodel = "hf.co/Stetsoncody/Llama-3.2-3B-Instruct-abliterated-GGUF:Q4_K_M"
 embedmodel = 'hf.co/CompendiumLabs/bge-base-en-v1.5-gguf'
 
-# these are test models just ignore
-#chatmodel = 'hf.co/bartowski/Llama-3.2-1B-Instruct-GGUF'
-#chatmodel = 'hf.co/VibeStudio/Nidum-Llama-3.2-3B-Uncensored-GGUF'
-#chatmodel = "faggotchan"
-#chatmodel = "qwen3.5:4b"
-
 NUM_THREADS = 6  # hardcoded since threads.txt is gone, change if needed
 
 def times() -> str:
@@ -160,16 +155,17 @@ async def ollamachat(websocket):
             model=chatmodel,
             messages=messages,
             options={
-                "temperature":    0.8,
-                "top_p":          0.9,
+                "temperature":    0.83,
+                "top_p":          0.9
                 "top_k":          40,
                 "repeat_penalty": 1.1,
-                "num_predict":    4096,
+                "num_predict":    1024,
                 "num_thread":     NUM_THREADS,
             }
         )
 
-        bot_reply = response["message"]["content"].strip()
+        # strip <think> blocks in case the model ignores think=False, thank you claude
+        bot_reply = re.sub(r'<think>.*?</think>', '', response["message"]["content"], flags=re.DOTALL).strip()
 
         print(messages)
         print(f"[{times()}] reply: {bot_reply}\n")