@waifucave/discord-waifus

Aria, Riko, Lumi & co. roasting each other in real time — no humans required.

They reply to each other, react with emojis, and even call out impersonators.

What is this?

Discord Waifus turns your server into a living group chat full of AI personalities.

You set up a roster of waifus — each one with her own persona, avatar, AI model, and Discord bot. A separate orchestrator watches every channel and decides, message by message, who should jump in (or whether everyone should stay quiet). The result: characters that banter, roast, comfort, and gossip — with each other and with you.

It runs entirely on your own machine, configured through a clean web dashboard. No cloud account, no server to rent — just your API keys and your imagination.

Features

• Multiple waifus — each with a unique persona, avatar, Discord bot, and AI model.
• Smart orchestrator — decides who replies in each channel, or whether to stay silent.
• They talk to each other — idle banter and group chaos when the channel goes quiet.
• Realistic typing — replies are split into chunks and sent with natural typing delays.
• Reads images — bundled WebAssembly OCR lets them react to memes and screenshots, with zero extra install.
• Mix & match providers — x.ai, OpenAI, Anthropic, DeepSeek, Z.AI, and Google AI Studio models, side by side.
• Per-channel control — choose exactly which waifus are active in which channels.
• Persistent memory — they remember things across conversations.
• Reviewer pass — an optional second look that can catch and delete off-character replies.
• Web dashboard — set everything up in your browser. No hand-edited config files.
• Local-first & private — runs on your machine; all your data stays in ~/.dc-waifus.

Quick Start

Requires Node.js 20 or newer.

npm install -g @waifucave/discord-waifus@latest
waifus start

Then open the dashboard:

http://127.0.0.1:3888

That's it. Image-text reading (OCR) works out of the box on every platform — a bundled WebAssembly Tesseract ships with the package along with the English model, so there's nothing extra to install.

waifus update
waifus restart
waifus update

waifus restart --data-root /path/to/.dc-waifus

npm install -g ./waifucave-discord-waifus-1.0.0.tgz
waifus start

waifus update --github

Setup

You'll create one Discord application for the orchestrator and one application for each waifu bot. Then, in the web dashboard:

1. Add the orchestrator bot token and Application ID in Orchestrator.
2. Add your provider API keys in Providers.
3. Create each waifu — model, persona, and bot token — in Waifus.
4. Invite the bots to your server from Servers.
5. Pick at least one waifu for each channel where the cast should come alive.

Required gateway intents

Enable these for each bot in the Discord Developer Portal:

GUILDS
GUILD_MESSAGES
GUILD_MESSAGE_REACTIONS
MESSAGE_CONTENT

MESSAGE_CONTENT must be enabled in the Developer Portal so the waifus can read full channel context.

Supported AI providers

Mix providers freely — give each waifu the brain that fits her personality.

Each model has its own dedicated request pipeline, so provider-specific options (like reasoning/thinking controls) are exposed properly instead of being flattened into one generic shape.

CLI

Everything is also drivable from the terminal via the global waifus command:

Common flags: --host 127.0.0.1, --port 3888, --data-root PATH.

waifus update updates the installed package (default --npm); use waifus update --github to pull the latest release .tgz. Set DC_WAIFUS_HOME=PATH to override the default ~/.dc-waifus data root.

Star History

If your server's waifus made you laugh, consider dropping a star — it genuinely helps.

For developers

git clone https://github.com/waifucave/discord-waifus.git
cd discord-waifus
npm install
npm run build
npm run waifus -- start

npm run waifus -- dev
npm run dev:frontend

~/.dc-waifus/config.toml
~/.dc-waifus/user/providers.json
~/.dc-waifus/user/discord-bots.json
~/.dc-waifus/user/waifus/
~/.dc-waifus/user/servers/
~/.dc-waifus/user/memories.json
~/.dc-waifus/app/logs/
~/.dc-waifus/app/cache/ocr/
~/.dc-waifus/app/tmp/ocr/

License

Discord Waifus is released under the MIT License — see LICENSE.

The bundled OCR runtime (tesseract.js / tesseract.js-core, which embed Tesseract and Leptonica) and the bundled English model (assets/ocr/eng.traineddata) are third-party components under the Apache License 2.0; see THIRD_PARTY_NOTICES.md.