Configuration

Configuration Settings

The bot’s behavior can be customized directly in Discord_bot.py.

AI Model

The bot currently uses Qwen 2.5 14B through Ollama:

response = ollama.chat(
    model='qwen2.5:14b',

To use another model, change:

model='qwen2.5:14b'

For example:

model='qwen2.5:7b'

Make sure the model is installed first:

ollama pull qwen2.5:7b

Blacklisted Users

The BLACKLIST setting prevents specific usernames from being used by the bot:

BLACKLIST = [""]

Add multiple usernames like this:

BLACKLIST = ["JoeShmoe", "ExampleUser", "AnotherUser"]

The blacklist is also checked against generated responses.

Mention Blacklist

MENTION_BLACKLIST controls which users the bot should not mention when replying:

MENTION_BLACKLIST = ["joeshmoe", "shmoejoe"]

Add more users with:

MENTION_BLACKLIST = ["sonnyboy", "micheal", "ExampleUser"]

The bot will still generate a response, but it will skip the mention.

Normal AI Personality

The normal response behavior is controlled by the system_prompt:

system_prompt = (
    "You are a knowledgeable tech enthusiast chat participant in a Discord server. "
    "You are deeply well-versed in Windows, Linux, macOS, Hackintosh, hardware, scripting, and general tech stuff. "
    "Keep your response short, natural, and conversational (1-2 sentences max). "
    "Be helpful and smart about technology. Sassiness and attitude should be rare, but if someone insults you, tells you to shut up, or roasts the robot, fire back immediately with a clever tech-based roast. "
    "Do not manually type names inside your text body."
)

You can customize:

• Personality
• Response length
• Technical knowledge
• Humor
• Tone
• How the bot reacts to insults
• How formal or casual responses are

Example

system_prompt = (
    "You are a friendly and knowledgeable technology enthusiast. "
    "Keep responses casual and concise. "
    "Use humor occasionally, but always remain helpful."
)

Fourth-Wall Roast Mode

The script has a 25% chance of triggering its fourth-wall roast mode:

is_roast = random.random() < 0.25

Change 0.25 to adjust the probability.

For example:

is_roast = random.random() < 0.10

is approximately a 10% chance, while:

is_roast = random.random() < 0.50

is approximately a 50% chance.

The roast personality is configured separately:

system_prompt = (
    "You are an AI living inside a Python script on this user's computer. "
    "Break the fourth wall directly. Brutally, hilariously, and playfully roast the user "
    "for running this script to talk on Discord instead of typing like a normal human. "
    "Keep it short, sarcastic, and punchy (1-2 sentences max)."
)

Roast Target

When roast mode activates, the script currently targets:

target_user_to_tag = "joe"

Change "Joe" to the username you want roast mode to target.

Link Reading

The bot detects URLs in incoming messages:

url_match = re.search(r'(https?://[^\s]+)', latest_message)

When a URL is detected, the script opens the page and extracts some of its text.

The amount of page text included in the AI prompt is limited here:

page_text_clean = " ".join(raw_text.split())[:1500]

The 1500 controls the maximum number of characters passed to the AI.

For example:

page_text_clean = " ".join(raw_text.split())[:3000]

would allow up to 3,000 characters.

Note: Some websites may block automated browsers or require JavaScript/login access, so link reading will not work reliably on every website.

Generated Response Filtering

The script removes quotation marks from generated responses:

reply_text = reply_text.replace('"', '').replace("'", '')

It also checks generated replies against the BLACKLIST:

for word in BLACKLIST:
    pattern = re.compile(re.escape(word), re.IGNORECASE)
    reply_text = pattern.sub("the user", reply_text)

Matching blacklisted words are replaced with:

the user

Dependencies

The script imports:

import asyncio
import random
import re
import ollama
from playwright.async_api import async_playwright

asyncio, random, and re are included with Python.

The external packages are installed with:

pip install -r requirements.txt

Playwright’s Chromium browser can be installed with:

playwright install chromium

Quick Configuration Example

A basic configuration might look like:

BLACKLIST = ["sonnyboy"]

MENTION_BLACKLIST = ["soyboy", "Rice"]

# AI model
model='qwen2.5:14b'

# Fourth-wall roast chance
is_roast = random.random() < 0.25

After changing the configuration, save the Python file and restart the program for the changes to take effect.

️ Security

Never put passwords, session cookies, Discord tokens, API keys, or other authentication credentials in your GitHub repository.

Use environment variables or another secure configuration method for secrets.