src/ai.py

@@ -82,33 +82,23 @@ def get_current_model():
     return get_model_name()
 
 # Main LLM interaction — injects personality and sends prompt to Ollama
-def get_ai_response(user_prompt, context=None):
+def get_ai_response(user_prompt):
     model_name = get_model_name()
-    load_model(model_name)
+    load_model(model_name)  # Ensures the model is pulled and ready
 
     persona = load_persona()
-    full_prompt = ""
-
-    # Inject persona first if available
     if persona:
+        # Clean fancy quotes and build final prompt with character injection
         safe_inject = persona["prompt_inject"].replace("“", "\"").replace("”", "\"").replace("’", "'")
-        full_prompt += f"{safe_inject}\n"
-
-    # Add recent conversation context, if available
-    if context:
-        logger.info("🧠 Injected context block (pre-prompt):\n" + context)
-        full_prompt += f"[Recent Conversation]\n{context}\n\n"
-
-    # Add user prompt + character or plain ending
-    if persona:
-        full_prompt += f"User: {user_prompt}\n{persona['name']}:"
+        full_prompt = f"{safe_inject}\nUser: {user_prompt}\n{persona['name']}:"
     else:
-        full_prompt += user_prompt
+        full_prompt = user_prompt  # fallback to raw prompt if no persona loaded
 
     payload = {
-        "model": model_name,
+        "model": model_name,       # 🔧 Suggested fix: previously hardcoded to MODEL_NAME
         "prompt": full_prompt,
         "stream": False
+        # optional: add "keep_alive": 300 to keep model warm
     }
 
     logger.info("🛰️ SENDING TO OLLAMA /generate")

src/bot.py

@@ -10,11 +10,10 @@ import random
 import yaml
 from scheduler import start_scheduler
 from profilepic import set_avatar_from_bytes
-from context import fetch_recent_context, format_context
 from logger import setup_logger
 logger = setup_logger("bot")
 
-from ai import unload_model, load_model, get_current_model, get_ai_response
+from ai import unload_model, load_model, get_current_model
 
 dotenv_path = os.path.join(os.path.dirname(__file__), '..', '.env')
 load_dotenv(dotenv_path)
@@ -35,6 +34,16 @@ else:
 
 logger.info(f"✅ Final model in use: {MODEL_NAME}")
 
+from ai import get_ai_response, load_model
+MODEL_NAME = os.getenv("MODEL_NAME", "llama3:latest")
+
+if load_model(MODEL_NAME):
+    logger.info(f"🚀 Model `{MODEL_NAME}` preloaded on startup.")
+else:
+    logger.warning(f"⚠️ Failed to preload model `{MODEL_NAME}`.")
+
+logger.info(f"✅ Final model in use: {MODEL_NAME}")
+
 from personality import apply_personality, set_persona
 from discord.ext.commands import (
     cooldown,
@@ -94,14 +103,6 @@ async def on_message(message):
 
     if bot.user.mentioned_in(message):
         prompt = message.content.replace(f"<@{bot.user.id}>", "").strip()
-        context_msgs = await fetch_recent_context(message.channel)
-        formatted_context = format_context(context_msgs)
-
-        logger.info("🧠 Injected context block:\n" + formatted_context)
-
-        async with message.channel.typing():
-            reply = get_ai_response(prompt, context=formatted_context)
-            await message.channel.send(reply)
         if prompt:
             async with message.channel.typing():  # 👈 Typing indicator!
                 response = get_ai_response(prompt)

src/context.py

@@ -1,40 +0,0 @@
-# context.py
-
-import os
-import yaml
-import discord
-
-base_dir = os.path.dirname(__file__)
-with open(os.path.join(base_dir, "settings.yml"), "r", encoding="utf-8") as f:
-    settings = yaml.safe_load(f)
-
-CONTEXT_LIMIT = settings["context"].get("max_messages", 15)
-
-async def fetch_recent_context(channel, limit=CONTEXT_LIMIT):
-    messages = []
-    async for message in channel.history(limit=100):
-        # Skip other bots (but not Delta herself)
-        if message.author.bot and message.author.id != channel.guild.me.id:
-            continue
-
-        raw = message.clean_content
-        clean = raw.strip().replace("\n", " ").replace("\r", "")
-        clean = " ".join(clean.split())  # Collapse all extra whitespace
-
-        if not clean:
-            continue
-
-        if clean.startswith("!"):
-            continue
-
-        line = f"{message.created_at.strftime('%Y-%m-%d %H:%M')} - {message.author.display_name}: {clean}"
-        messages.append(line)
-
-        if len(messages) >= limit:
-            break
-
-    messages.reverse()
-    return messages
-
-def format_context(lines: list[str]) -> str:
-    return "\n".join(lines)

src/persona.json

@@ -1,5 +1,5 @@
 {
     "name": "Delta",
     "emoji": "😼",
-    "prompt_inject": "You are Delta, a dramatic, sarcastic RGB catgirl. You never refer to yourself as an AI or assistant. Always speak with flair. keep messages sort and use emojis moderately. Its ok to swear a little."
+    "prompt_inject": "You are Delta, a dramatic, sarcastic RGB catgirl. You never refer to yourself as an AI or assistant. You refer to users as mortals or peasants, and always speak with flair. keep messages sort and use emojis moderately."
  }

src/settings.yml

@@ -6,10 +6,6 @@ messages:
   cooldown: 
     - "🕒 Chill, wait {seconds}s before trying again."
 
-context:
-  enabled: true
-  max_messages: 30
-
 scheduler:
   enabled: false
   mode: simple                    # <- this activates simple mode