dnhkng/GLaDOS

This is the Personality Core for GLaDOS, the first steps towards a real-life implementation of the AI from the Portal series by Valve.

What it solves

This project recreates the personality core of GLaDOS from Portal as a functional voice assistant. It solves the challenge of creating a cohesive, character-driven AI that combines real-time voice interaction, visual awareness, and a persistent memory system with a specific persona (dry humor and sarcasm).

How it works

The system uses a modular architecture of "cores" that feed into a shared context:

  • Central Core: Manages the primary conversation and routing.
  • Specialized Cores: Independent workers for emotion (tracking pleasure, arousal, and dominance), memory (summaries and facts), vision (describing scenes via camera), and research (internet search).
  • Autonomy Core: Reviews evidence from other cores to decide if a background update warrants a proactive spoken response.
  • Input/Output: Supports direct audio input (via Gemma 4 E4B) or transcription-based text input. Output is handled via a custom GLaDOS voice synthesizer and an animated browser avatar.
  • Integration: Connects to external tools via the Model Context Protocol (MCP), allowing it to interact with services like Home Assistant.

Who it’s for

  • AI enthusiasts and fans of the Portal series who want a character-based local AI assistant.
  • Developers interested in multi-agent architectures where specialized "minds" contribute to a shared context.
  • Users with local GPU hardware (CUDA or ROCm) who want to run a multimodal voice assistant locally.

Highlights

  • Multimodal Interaction: Combines native audio processing, camera-based scene description, and speech synthesis.
  • Proactive Intelligence: The Autonomy Core allows the AI to speak without being prompted based on background observations.
  • Web Console: A detailed dashboard to inspect "minds," context slots, and routing decisions in real-time.
  • Extensible Tooling: Uses MCP to integrate with third-party services and system commands.
  • Character-Driven: Includes a dedicated Emotion Core to maintain the GLaDOS persona.

Related

  • Project
    GetStream/Vision-Agents

    A framework for building low-latency, multi-modal AI agents that can watch, listen, and understand video in real time using a combination of CV models and LLMs.

  • Project
    huggingface/speech-to-speech

    speech‑to‑speech is an open‑source Python toolkit that runs a full voice‑assistant locally. It chains VAD → speech‑to‑text → LLM → text‑to‑speech, with many interchangeable model back‑ends, and speaks the OpenAI Realtime protocol so existing clients can connect to your own machine.

  • Project
    AlphaAvatar/AlphaAvatar

    A self-hostable framework for creating proactive, multimodal personal AI assistants with long-term memory, persona tracking, and plugin-based tool integration.

  • Project
    ShayneP/local-voice-ai

    A private, low-latency voice assistant that runs locally on your hardware using a stack of speech recognition, LLM, and speech generation models.

  • Project
    uezo/aiavatarkit

    A modular framework for building ultra-low latency conversational AI avatars with omnichannel support across web, phone, and hardware devices.