GPTDiscord

GPTDiscord is an open-source Discord bot, written in Python and released under the MIT License, that brings OpenAI models into Discord servers. Its README describes ChatGPT-style conversations, DALL-E image generation, AI moderation, custom indexes over your own documents, and internet-connected chat. The latest listed release is v12.4.0.

What It Is

GPTDiscord is a self-hosted bot that connects a Discord server to OpenAI's APIs. Users interact through slash commands such as /gpt converse, /dalle draw, /index chat, /internet chat and /code chat. The README says it aims for feature parity with ChatGPT web.

Conversation, Search and Code Execution

Conversations can be long-term, support per-user and per-channel instructions, and allow redoing or editing messages. Multi-modal conversations accept images and can draw images with DALL-E. Internet-connected chat uses Google, Wolfram Alpha and a link crawler. A code interpreter mode provides an isolated execution environment, powered by E2B according to the README, where packages can be installed.

Knowledge Indexes and Moderation

Users can build indexes from files, PDFs, text, websites or Discord channel content, then query them or chat with them. Indexes can be loaded per user or per server, and multiple indexes can be composed together. The README says it is migrating from Pinecone to Qdrant as its vector database. AI moderation can automatically moderate a server, and DeepL translation is supported.

Setup and Administration

The repository provides detailed guides for installation, a DigitalOcean droplet deployment, and obtaining an OpenAI token. It is also distributed through PyPI and Docker. Admin commands let operators adjust model parameters, track token usage, enable a low usage mode, and send debug output to a channel.