Johnson-Durui/Companion-Space

What it solves

Companion Space is a local-first application designed for "companion learning." It allows users to create dedicated study spaces for different subjects, import their own materials, and interact with virtual characters to help them organize, review, and memorize information.

How it works

The system operates primarily on the user's own computer to ensure privacy, with data and credentials stored locally. It uses a Retrieval-Augmented Generation (RAG) approach to allow users to ask questions based on imported documents, providing answers with citations. For interaction, it supports text and real-time voice communication, including the ability to interrupt the AI. Visually, it integrates 3D VRM characters (and 2D backups) that provide feedback through lip-syncing, emotions, gaze, and movement. The AI backend is flexible, supporting mock services for immediate testing, as well as OpenAI-compatible APIs and Ollama.

Who it’s for

Students, lifelong learners, and fans of anime-style virtual characters who want a private, interactive, and visually engaging way to manage their study materials and review content.

Highlights

• Local-First Privacy: Data and credentials stay on the user's machine.
• RAG-Powered Study: Import personal documents for grounded AI conversations with citations.
• Multimodal Interaction: Supports real-time voice, text, and 3D VRM character animations.
• Flexible AI Backend: Compatible with Ollama and OpenAI-compatible providers.
• Customizable Spaces: Organize materials and conversations by course, project, or interest.