🧠 God Bot — Local AI Discord Platform

God Bot is a local, Ollama-powered Discord AI platform with:

• ✅ Deterministic tools for finance, fitness, nutrition, math, Wild Rift builds
• ✅ A Discord chat bot with streaming replies
• ✅ A web dashboard (live logs, model switcher, memory viewer, plugins, agents, voice toggle)
• ✅ Long-term memory (SQLite + JSON) and vector memory
• ✅ A plugin system with sandboxed tools (temporarily running in non-sandboxed mode)
• ✅ A scheduler for background tasks (heartbeats, plugin reload, memory cleanup, daily reports)

It's built to be your personal "AI OS" running on your own hardware, using your models.

---

🔧 Features

💬 Discord AI Assistant

• Streams responses from your local Ollama model (default: dolphin-llama3:latest)
• Chat in channels or DMs
• Image description for attachments
• Personality layer on top of LLM output
• Long-term memory of past conversations (per-user)

🧮 Deterministic Tool System (Zero-Hallucination for Numbers)

Located in deterministic/:

• math_tools.py — arithmetic, percentages, tips, basic conversions
• finance_tools.py — runway, FI age, FIRE-style calculations (more to come)
• fitness_tools.py — 1RM & strength calculations (Epley, etc.)
• nutrition_tools.py — calories/macros helpers
• wildrift_tools.py — Wild Rift build loader + deterministic build lookup

These tools are pure Python. When a message matches a deterministic pattern (e.g. specific finance/bench/Wild Rift queries), God Bot:

1. Runs the Python tool first
2. Returns the exact numeric answer
3. Only uses the LLM for explanation or non-numeric reasoning

No hallucinations for math/finance/1RM once a tool is in place.

⚔️ Wild Rift Build System

• Champion builds stored as Markdown in data/wild_rift_builds/
  • e.g. garen.md, vi.md, shyvana.md, jinx.md, leona.md, nasus.md
• Deterministic loader in deterministic/wildrift_tools.py
• Triggers when user asks for a Wild Rift build or mentions specific champs

Easy to extend: just drop a new .md file and the system can pick it up.

📊 Finance Tools

• Runway calculator (months/years left at your current burn)
• FI age / time to FI (baseline)
• Designed to be extended with:
  • Coast FI, Lean FI
  • Inflation-adjusted millionaire timelines
  • Drawdown simulators
  • Net worth projections

All of this lives in deterministic/finance_tools.py.

🍗 Fitness & Nutrition

• 1RM estimation tools in deterministic/fitness_tools.py
• Basic macros/TDEE helpers in deterministic/nutrition_tools.py
• Extensible so you can add:
  • Strength standard checks
  • Program generators
  • Volume recommendations
  • Macro presets

🧠 Memory System

• godbot/core/memory.py — JSON + SQLite (MemoryDB) for long-term chat history
• memory.json — per-user "facts" (stable preferences, jobs, etc.)
• long_memory.db — long-term conversation logs
• godbot/core/vector_memory.py — vector memory for semantic search

The bot uses a combination of:

• recent chat snippets
• long-term memory
• optional vector search results

…to build context for each LLM call.

🕒 Scheduler

Core scheduler lives in godbot/core/scheduler.py.

Tasks are registered in main.py (or via scheduled task modules) and can include:

• Heartbeat logs
• Plugin auto-reload checks
• Memory cleanup (dedup + trimming)
• Daily reports to a Discord channel

Runs in the background as an async loop.

🧩 Plugins

The plugins/ directory contains:

• plugin_manager.py — manages plugin loading, behavior injections, tools
• Example plugin: example_search_tool/
• Friendly behavior mode: friendly_mode/

Plugins can:

• Inject behavior into prompts (e.g., tone, style)
• Provide tool functions callable by the LLM
• Be hot-reloaded (via plugin manager and scheduler)

Note: A temporary non-sandboxed safe_exec is in place during refactor; a stronger sandbox will be restored in a later phase.

🌐 Dashboard

dashboard.py + dashboard/ui/ provide a small web UI:

• Live logs
• Model list + switcher (via Ollama tags)
• Memory viewer/editor
• Plugin viewer
• Agent list
• Voice agent toggle

Runs at:

http://localhost:5000

when the bot is running.

---

🏗️ Architecture Overview

High-level layout:

flowchart LR
    subgraph Discord
        U[User] -->|messages| B[God Bot Client]
    end

    subgraph Core
        LLM[LLM Wrapper<br/>godbot/core/llm.py]
        MEM[MemoryDB + JSON<br/>godbot/core/memory.py]
        VM[Vector Memory<br/>godbot/core/vector_memory.py]
        SCHED[Scheduler<br/>godbot/core/scheduler.py]
    end

    subgraph Deterministic
        FIN[Finance Tools]
        FIT[Fitness Tools]
        NUT[Nutrition Tools]
        MATH[Math Tools]
        WR[Wild Rift Tools]
    end

    subgraph Plugins
        PM[Plugin Manager]
        PL[Plugins Folder]
    end

    subgraph Dashboard
        FLASK[dashboard.py]
        UI[dashboard/ui]
    end

    B --> LLM
    B --> MEM
    B --> VM
    B --> SCHED
    B --> PM
    PM --> PL

    B --> FIN
    B --> FIT
    B --> NUT
    B --> MATH
    B --> WR

    B --> FLASK
    FLASK --> UI

---

📁 Project Structure

Current key layout:

ai_discord_bot/
├── main.py                 # Entry point (Discord bot bootstrap)
├── agents.py               # Agent manager
├── audio.py                # Voice agent
├── committee_agent.py      # Committee-style multi-agent logic
├── dashboard.py            # Flask backend for web UI
├── godbot/
│   ├── core/
│   │   ├── llm.py          # Central Ollama streaming wrapper
│   │   ├── memory.py       # MemoryDB + JSON memory logic
│   │   ├── vector_memory.py# Vector memory search
│   │   └── scheduler.py    # Core scheduler
│   └── discord/
│       └── bot.py          # MyClient definition & factory
├── deterministic/
│   ├── finance_tools.py
│   ├── fitness_tools.py
│   ├── math_tools.py
│   ├── nutrition_tools.py
│   └── wildrift_tools.py
├── data/
│   └── wild_rift_builds/
│       ├── garen.md
│       ├── vi.md
│       ├── shyvana.md
│       ├── leona.md
│       ├── nasus.md
│       └── jinx.md
├── plugins/
│   ├── plugin_manager.py
│   ├── example_search_tool/
│   └── friendly_mode/
├── dashboard/
│   └── ui/
│       ├── index.html
│       ├── app.js
│       └── style.css
├── scheduled_tasks/        # (if present) memory cleanup, plugin reload, etc.
├── tests/                  # basic test scaffolding
├── requirements.txt
├── pyproject.toml
└── README.md               # (this file)

---

🚀 Getting Started

1. Requirements

• Python 3.10+
• A working Ollama install with at least one model (e.g. dolphin-llama3:latest)
• A Discord bot token

2. Clone and set up

git clone https://github.com/IStayLurkin/ai_discord_bot.git
cd ai_discord_bot

# Optional: create a virtual environment
python -m venv .venv
# Windows:
.venv\Scripts\activate
# Linux/macOS:
source .venv/bin/activate

pip install -r requirements.txt

3. Create your .env

In the project root (ai_discord_bot/):

DISCORD_TOKEN=your_discord_bot_token_here
OLLAMA_URL=http://localhost:11434
OLLAMA_MODEL=dolphin-llama3:latest

4. Make sure Ollama is running

ollama serve
ollama list
# ensure the model in OLLAMA_MODEL is available

5. Run the bot

python main.py

You should see logs like:

Discord slash commands synced.
Dashboard running on http://localhost:5000
Logged in as God Bot#1234
[Scheduler] Started

---

💡 Usage

Discord

Mention the bot or talk in a channel it can read.

Use slash commands like /ask, /research, /committee (depending on your current setup).

It will automatically:

• Use deterministic tools for math/finance/WR when applicable.
• Fall back to LLM for general reasoning.

Dashboard

Once the bot is running:

Open http://localhost:5000 in your browser.

View logs, switch models, inspect memory, and more.

---

🔧 Extending God Bot

Add a new deterministic tool

1. Pick a module in deterministic/ or create a new one.
2. Use the registry system in deterministic/registry.py (if present) or integrate with the existing pattern in main.py.
3. Write pure Python logic for the tool.
4. Add pattern detection in the deterministic layer so the tool fires before the LLM.

Add a new Wild Rift champion build

1. Create a new file in data/wild_rift_builds/, e.g. darius.md
2. Follow the build format used in garen.md / vi.md
3. Update wildrift_tools.py if needed to include any special logic

Add a plugin

1. Create a new folder under plugins/
2. Add a manifest.json and tool.py (see example_search_tool for reference)
3. The plugin manager will scan and load it on startup (or via auto-reload).

---

📜 License

MIT License

---

⭐ Acknowledgements

Built for high-performance local AI with Ollama, using Deterministic-First architecture for reliability.