infinigpt-matrix

InfiniGPT is a powerful AI chatbot for the Matrix chat protocol that can roleplay as almost anything you can imagine. It supports multiple providers — OpenAI, xAI, Google, Mistral, Anthropic — and optional local models via Ollama and LM Studio. Each user has a separate conversation history per room, with dynamic personalities and admin controls for models and resets.

Also available for IRC: https://github.com/h1ddenpr0cess20/infinigpt-irc
Ollama-only version at https://github.com/h1ddenpr0cess20/ollamarama-matrix

Documentation

• Overview
• Getting Started
• Configuration
• Commands
• Tools & MCP
• Images Directory
• Docker
• CLI Reference
• Operations & E2E
• Architecture
• Ollama Setup
• LM Studio Setup
• Development
• Migration
• Legacy Map
• Security
• Not a Companion
• AI Output Disclaimer

Features

• Dynamic personalities with quick switching
• Per‑user history, isolated per room and user
• Collaborative mode to talk across histories
• Multi‑provider LLMs with unified model selection
• Tool calling (builtin and MCP) for actions and lookups
• Admin controls for model switching and global resets

Installation

From source (installs CLI):

• pip install .
• Or use pipx: pipx install .

From source without installing the package:

• pip install -r requirements.txt
• Run with: python -m infinigpt --config config.json

After installation, use the infinigpt-matrix command.

Quick Start

1. Create a Matrix account for the bot and note the server URL, username, and password.
2. Choose providers. For cloud providers, set API keys via env or config.json. For local models, install Ollama or LM Studio
3. Create config.json. See Configuration for full schema
4. Run:

• Installed command: infinigpt-matrix --config config.json
• As module: python -m infinigpt --config config.json

Usage

Common commands (see Commands for the full list):

Command	Description	Example
.ai <message> or BotName: <message>	Chat with the AI	.ai Hello there!
.x <user> <message>	Continue another user's conversation	.x Alice What did we discuss?
.persona <text>	Change your personality	.persona helpful librarian
.custom <prompt>	Use a custom system prompt	.custom You are a coding expert
.reset / .stock	Clear history (default/stock prompt)	.reset
.mymodel [name]	Show/change personal model	.mymodel gpt-4o-mini
`.model [name	reset]` (admin)	Show/change model
.clear (admin)	Reset globally for all users	.clear
.help	Show inline help	.help

Encryption Support

• Works in encrypted Matrix rooms using matrix-nio[e2e] with device verification.
• Requires libolm available to Python for E2E. If unavailable, you can run without E2E; see Operations and Verification.
• Persist the store/ directory to retain device keys and encryption state.

Community & Policies

• Code of Conduct: Code of Conduct
• Contributing: Contributing
• Security Policy: Security Policy
• Security Guide: Security Guide

License

AGPL‑3.0 — see License for details.