Memories are injected into every conversation as facts the robot knows. Use them for context about the owner, the environment, and preferences.
Commands intercept specific phrases before they reach the LLM and return instant responses. Use variables like {{time}} and {{date}}.
Agents are specialized sub-models that run in parallel with the main model.
The main model (GPU) orchestrates the robot and delegates tasks to agents.
Agents can use internal GPU models or any external OpenAI-compatible API.
Voice alone → 🎙 badge. Face alone → from camera overlay. Both match → 🔒 badge in chat.
Each action becomes a Skill the robot brain can call automatically.
Robot will refuse to engage with these topics.
Robot will ask for human confirmation before executing these actions.
Rules injected into the system prompt on every conversation.