Settings

Tune how LUMEN works

This is a front-end demo. Responses are canned and simulated — no LLM is called. To make it real, connect your provider in one place: streamCompletion() in src/lib/llm.ts. Everything below is wired and persisted to your browser.

Default model

Used for new conversations. Each model trades depth for speed differently.

LUMEN Pro200K ctxBest balance of depth and speed. The default for most work.LUMEN AirFastFastest, lightest. Great for quick drafts and chat.LUMEN ReasonDeepDeliberate, step-by-step reasoning for hard problems.

System prompt

The standing instructions LUMEN follows in every chat. Set the tone, persona, and constraints.

Appearance

LUMEN remembers your choice across visits.

Dark theme

Switch between the dark and light skins.

Conversation data

Your chats are stored only in this browser (localStorage).