Product Introduction
- AIRI is a self-hosted, open-source AI companion platform designed to create and manage virtual characters ("containers of souls") capable of real-time interaction and integration with gaming environments like Minecraft and Factorio.
- Its core value lies in enabling full user ownership of AI-driven virtual beings while achieving Neuro-sama-level interactivity through multimodal capabilities, including voice synthesis, game automation, and adaptive personality modeling.
Main Features
- AIRI supports real-time voice chat with sub-500ms latency using WebRTC and custom speech synthesis models trained on anime-style vocal patterns for authentic character interactions.
- The platform integrates directly with Minecraft Java Edition 1.20+ and Factorio 2.0 through dedicated API bridges, enabling AI-controlled avatars to perform in-game tasks using reinforcement learning models.
- All character data and interaction logs are stored locally in SQLite or PostgreSQL databases with optional end-to-end encryption, ensuring compliance with GDPR and CCPA standards for user privacy.
Problems Solved
- AIRI addresses the lack of privacy-focused, customizable AI companions that can operate without reliance on third-party cloud services or data monetization practices.
- It primarily serves anime/VTuber enthusiasts, indie game developers, and AI researchers seeking ethical alternatives to commercial platforms like Grok Companion.
- Typical use cases include creating persistent AI personas for live-streaming assistive roles, automating repetitive tasks in sandbox games, and conducting experiments in neurosymbolic AI architectures.
Unique Advantages
- Unlike cloud-based competitors, AIRI implements a Kubernetes-native architecture that scales from Raspberry Pi clusters to enterprise-grade GPU servers while maintaining offline functionality.
- The platform introduces a "Soul Container" modular system using ONNX-format neural networks that allows hot-swapping personality modules without service interruption.
- Competitive advantages include MIT-licensed source code, native integration with Ollama and TensorRT inference engines, and community-driven LoRA training pipelines for character customization.
Frequently Asked Questions (FAQ)
- What hardware is required to run AIRI locally? AIRI requires a x86_64 or ARMv8 CPU with AVX2 support, 8GB RAM minimum, and optional NVIDIA CUDA 11.8 drivers for GPU acceleration, deployable via Docker or bare-metal Kubernetes.
- Can I import existing AI models into AIRI? The platform supports PyTorch 2.0+ and Hugging Face Transformers models through its adapter system, with pre-converted checkpoints available for models like Paimon 7B and Neuro-sama v3.
- How does game integration work? AIRI uses modified Fabric/Forge mods for Minecraft and Factorio API hooks to establish direct memory access for AI agents, bypassing screen scraping for faster reaction times below 200ms.