Favicon of GLaDOS

GLaDOS

A local AI voice assistant with a Portal-inspired voice, camera observations and persistent memory, using llama.cpp, Ollama or compatible cloud APIs.

GLaDOS is a local AI voice assistant for people who want a conversational companion modeled on the personality core from Valve's Portal games. It combines a custom voice and dry humor with awareness of its surroundings.

  • Voice and animated avatar: Talk through a local microphone or browser and hear replies in GLaDOS's voice, with an animated browser character that responds to spoken emotion directions.
  • Camera observations: The assistant describes scenes and recent changes, adjusting how often it observes the camera feed according to motion.
  • Persistent memory: It saves useful facts and conversation summaries, then retrieves relevant memories for later conversations.
  • Proactive replies: It can speak in response to background observations, while giving user interactions priority; Quiet mode pauses replies and background activity until a wake request.
  • Tools and research: MCP connections support services such as Home Assistant, while native tools manage preferences, memory and tasks; requested internet research uses Exa, an external service.

The browser console provides microphone and camera controls, settings, and views of saved memory and background activity. A terminal interface is also available.

GLaDOS is open source under the MIT license. The default profile sends audio directly to Gemma 4 E4B through a local llama.cpp server and is tuned for CUDA hardware. An alternative profile uses Parakeet transcription with Ollama or an OpenAI-compatible API. Choosing a cloud API sends conversation requests to that provider. Home Assistant integration requires your own server and credentials.

Similar to GLaDOS