Personal AI assistant 'Lyra' proposes layered design: personality, memory, skills

Lyra is a personal project designed differently from typical chatbots. Most quick follow a simple flow: User → Interface → AI API → Response.

Lyra instead follows a longer chain: User → Interface → Personality → Memory → Skills → Tools → Projects. The goal isn't just answering questions, but building an assistant that grows with the user over time.

The system is split into separate pieces: a core system that handles communication and decision flow, an AI layer that connects to different AI providers and models, a that stores and retrieves useful information, a skills (plugin) layer that lets new abilities be added without rewriting the whole system, an interface layer meant to eventually support both chat and voice, and a system that manages settings and . The itself is layered, starting with that holds the current conversation.

Key points

  • Lyra follows a longer flow: User → Interface → Personality → Memory → Skills → Tools → Projects
  • It aims for a more personalized, growing assistant rather than a generic chatbot
  • A skills/plugin layer lets new abilities be added without rewriting the whole system
  • Memory is designed in layers, starting with for current conversations
  • It's still a design in progress seeking architecture feedback, not a finished build
Read original