Your display name is used so the assistant can address you properly.
Used for every date/time the assistant recalls from memory and for its sense of "now."
Detect from this device
System Prompt
Defines the AI's personality and instructions. Saved per-account on the server.
Used by the assistant bubble play button. Only the higher-quality AF/AM Kokoro voices are shown.
LLM Endpoint
Hard cap for context allocation across memory layers. Set this to match the maximum your LLM endpoint actually accepts — mobile or quantized servers often limit to 4K or 8K. Auto uses a per-model heuristic (typically 32K).