BigDPP

A C++ Discord operations bot for server maintenance, moderation, diagnostics, announcements, logging, and administrative automation.

Project status

BigDPP is in Phase 1 — Discord Foundation. The native foundation, initial DPP integration, local LLM flow, and explicit administrator commands are implemented locally; live Discord verification awaits test credentials.

Direction

The optional local AI flow is implemented for /ask; enable it only when a loopback Ollama service and a dedicated AI channel are available.

BigDPP targets C++20, DPP/D++, CMake, nlohmann-json, and Ollama for optional local AI. PostgreSQL/libpqxx, spdlog, and Docker remain later roadmap items. It is designed as a production-oriented modular monolith for a real Discord server, with explicit Discord commands kept separate from the prompt-only local AI.

Local LLM

BigDPP uses a local Ollama runtime for /ask; it does not send prompts to a hosted LLM. The model is configurable through BIGDPP_LLM_MODEL and the example configuration uses llama3.2. When enabled, BigDPP calls Ollama's loopback POST /api/chat endpoint asynchronously with a non-streaming request. The model receives only the member's question and a fixed assistant instruction; it cannot read Discord state, access credentials, or execute server-management commands.

Enable it with BIGDPP_LLM_ENABLED=true, download the selected model with ollama pull <model>, and use BIGDPP_LLM_AUTOSTART to control whether BigDPP starts ollama serve automatically. See the configuration reference for all LLM settings.

Example /ask response from the local Discord assistant.

Branding

• branding/github-social-preview.png — GitHub repository social preview, 1774×887
• branding/Banner.png — original project banner
• branding/Logo.png — project logo
• branding/ai-screenshot.png — local AI Discord example

Documentation

• Start here: codebase tour
• Architecture
• Development guide
• Editing and extension guide
• Configuration reference
• Command reference
• Discord application setup
• Security and operations baseline
• Local LLM decision
• Administrator command decision
• Delivery roadmap
• Repository working instructions

If you are new to the project, read the codebase tour first, follow the development guide to build and test, then use the editing guide before adding a command or module.

Implemented

• C++20 CMake executable and reusable core target
• environment-backed configuration with validation
• pinned vcpkg manifest
• DPP 10.1.5 Gateway adapter using the Guilds and Guild Members intents
• /ping and /status commands
• /ask for questions answered by a local Ollama model in the ai-agent channel
• automatic discovery or creation of the configured AI text channel
• explicit owner/Administrator server-management commands with confirmation and role hierarchy checks
• additive /server-setup baseline for categories, channels, roles, and restricted #mod-log
• process and #mod-log audit records for administrative actions
• development-guild or global bulk command registration
• secure ignored .env fallback with process-environment precedence
• offline CTest configuration and administrator-command coverage
• project scope, architectural boundaries, phased roadmap, and development/security policies

Planned

• live test-guild verification of Phase 1 and the administrative command surface
• remaining container and CI verification
• PostgreSQL persistence and migrations
• Read-only server diagnostics
• PostgreSQL-backed warnings, durable audit history, announcements, audit events, and advanced maintenance in later phases

See the roadmap for phase acceptance criteria. Planned features are intentionally not presented as available commands.

Installation and build on Windows

Prerequisites

• Windows 10 or newer
• Visual Studio Build Tools with the Desktop C++ workload and Windows SDK
• CMake 3.25 or newer
• Ninja
• vcpkg
• A Discord bot application and token
• Ollama only if /ask is needed

DPP and nlohmann-json are installed automatically through the checked-in vcpkg manifest. Node.js is not required.

Configure and build

If vcpkg is not installed, bootstrap it once:

git clone https://github.com/microsoft/vcpkg.git C:\tools\vcpkg
& C:\tools\vcpkg\bootstrap-vcpkg.bat

Open a new terminal, then activate the Visual Studio developer environment:

cmd.exe /k VsDevCmd.bat -arch=x64 -host_arch=x64

In that developer shell, set VCPKG_ROOT to your vcpkg installation and run:

$env:VCPKG_ROOT = "C:\path\to\vcpkg"
cmake --preset windows-debug
cmake --build --preset windows-debug
ctest --preset windows-debug

The first configure may build the manifest dependencies and take several minutes. The generated executable is out\build\windows-debug\bigdpp.exe.

Configure Discord credentials

From the repository root, create the ignored settings file:

Copy-Item .env.example .env

Set DISCORD_TOKEN and, during development, DEVELOPMENT_GUILD_ID in .env. Guild registration is fast; without that value, commands are registered globally and may take longer to appear. Complete the Discord application setup in Discord Application Setup before starting the bot.

Optional local LLM setup

Install Ollama, download the configured model, and set these values in .env:

BIGDPP_LLM_ENABLED=true
BIGDPP_LLM_AUTOSTART=true
BIGDPP_LLM_BASE_URL=http://127.0.0.1:11434
BIGDPP_LLM_MODEL=llama3.2
BIGDPP_AI_CHANNEL_NAME=ai-agent

ollama pull llama3.2

When BIGDPP_LLM_AUTOSTART=true, BigDPP starts ollama serve itself. Set it to false when Ollama is managed separately.

Start BigDPP

.\out\build\windows-debug\bigdpp.exe

Stop the development process with Ctrl+C. Graceful shutdown is planned for production hardening. Never commit .env or expose the bot token in logs.

Commands

Command	Status	Purpose
/ping	Implemented	Verify that BigDPP responds
/status	Implemented	Show the bot's online identity
/ask question:<text>	Implemented	Ask the configured local LLM; available only in #ai-agent by default

Administrative commands are available only to the guild owner or a member with Discord Administrator. Destructive commands require confirm:true, and target members and roles must be below BigDPP's highest role. See the command reference for parameters, required bot permissions, and the exact confirmation behavior. The command groups are:

• /server-info, /server-edit, /server-setup
• /channel-create, /channel-edit, /channel-delete, /channel-lock, /channel-unlock, /channel-access
• /role-create, /role-edit, /role-delete, /role-add, /role-remove
• /member-warn, /member-timeout, /member-untimeout, /member-kick, /member-ban, /member-unban
• /messages-purge, /audit-log

/create-role remains as a compatibility alias for /role-create. Validation responses are ephemeral; deferred mutation responses are completed asynchronously. /ask is chat-only and cannot invoke these commands or mutate the server.

The local LLM is prompt-only: it cannot inspect Discord state or invoke administrator commands. See the command reference for the available server-management operations and their safety checks.