Capsize-Games/airunner

Offline inference engine for art, real-time voice conversations, LLM powered chatbots and automated workflows

What it solves

AI Runner provides a private, local-first environment for interacting with AI companions and generating AI art. It eliminates the need for internet connections, API keys, or subscriptions by running everything on the user's own hardware, ensuring data privacy and offline availability.

How it works

The system operates via a headless daemon (FastAPI server) that orchestrates various AI runtimes. It uses llama.cpp and whisper.cpp as native sidecars for LLM and STT capabilities, and Python-based runtimes for image generation (SDXL, Z-Image Turbo) and TTS. A desktop GUI client connects to this daemon to provide a user interface for chatting and art creation.

Who it’s for

Users who want a highly customizable, private AI companion with long-term memory and a layered art canvas for AI-assisted drawing and image generation, all running locally.

Highlights

  • Customizable AI Companion: Create a named companion with a specific personality, voice, and mood that evolves through conversation.
  • Layered Art Canvas: A multi-layer drawing surface that supports sketching, painting, and image-to-image generation with inpainting and filters.
  • Local-First Privacy: No telemetry and no data leaves the machine by default.
  • Multimodal Capabilities: Integrated Text-to-Speech (TTS) and Speech-to-Text (STT) for voice conversations.
  • RAG-Powered Memory: The companion builds long-term memory of the user across sessions using Retrieval-Augmented Generation.
  • Environmental Awareness: The companion is aware of the current time, date, and local weather.
  • Model Management: Built-in downloaders for HuggingFace and CivitAI models.

Related

  • Project
    ThinkInAIXYZ/deepchat

    An open-source, local-first AI agent desktop client that integrates multiple LLM providers, MCP/ACP protocols, and remote control capabilities for messaging apps.

  • Project
    HKUDS/nanobot

    An ultra-lightweight, self-hosted personal AI agent framework that combines long-term memory, MCP integrations, and scheduled automation across WebUI, terminal, and various chat apps.

  • Project
    oobabooga/textgen

    A private, open-source desktop app for running local LLMs with support for multiple backends, multimodal input, tool-calling, and local fine-tuning.

  • Project
    RunanywhereAI/runanywhere-sdks

    A cross-platform SDK that enables developers to run LLMs, vision, speech, and image generation models locally on phones, browsers, and desktops using a unified API.

  • Project
    agiresearch/AIOS

    AIOS is an open‑source “operating system” for LLM‑based agents. The kernel (this repo) manages LLM instances, memory, storage, and tool sandboxing, exposing a syscall‑style API via the Cerebrum SDK. It supports many model back‑ends (Ollama, vLLM, HuggingFace, Deepseek), integrates external tools, and offers several deployment modes—from local single‑machine setups to remote, personal, and virtualized kernels. Installation is via a Python 3.10/3.11 environment, configuration is done in `config.yaml`, and agents can be run through a web UI or a natural‑language terminal. The project is research‑grade (multiple conference papers) and actively maintained.