Chat interaction defaults. Previewed immediately; saved only when you choose Save.
Enter to send
When off, Enter inserts a newline. Ctrl/⌘ + Enter still sends.
Conversation density
Changes vertical breathing room without widening the conversation stage.
AI / Model
Choose the inference model used by this clean-room Chat. The selected model is sent explicitly with every Station request.
Default model
Controls the actual Station inference route for new messages.
Effort preference
Saved as a model preference only until the LALM routing tier consumes it.
Response length
Your preferred answer depth; persistence does not itself enforce generation length.
Routing truth: Only controls connected to live generation are shown here.
You
Durable user profile details stay separate from creative lore. Saving here updates the existing account profile authority; Tier C does not yet inject these fields into LALM inference.
Profile context only; creative identity belongs in User Lore.
§wyrlz
Companion identity is versioned separately from your own profile preferences. Name changes preview in Chat presentation; personality fields are stored profile state and are not yet an inference-persona binding.
Authority: AI-originated companion profile changes remain disabled until Tier D adds proposal + approval semantics.
Lore & Memory
Account-owned continuity records with explicit type, origin, scope, activity state, and version. Tier C stores and manages them; nothing is extracted from conversations or injected into LALM inference automatically yet.
Sign in with durable account storage to load Lore & Memory.
New lore record
Global record
New manual record · authored by USER · confidence 100%
Rapport
Shared vocabulary, callbacks, and interaction conventions live here as inspectable state. Tier E stores and scopes rapport; it does not yet inject rapport into R39 inference.
Sign in with durable account storage to load rapport state.
Current generation only. Reset history remains preserved.
New rapport record
Global rapport
New manual rapport · authored by USER · current generation
Reset semantics: Reset advances the active rapport generation. Older generations remain inspectable when “Show reset history” is enabled; they are not silently destroyed or reused.
Proposal Inbox
AI-originated durable changes land here under your policy. The browser can resolve proposals but cannot forge assistant-authored proposals.
Per-category policy
ASK queues a durable proposal. AUTO_LOW_RISK may apply only trusted LOW-risk, bounded operations and still records audit + revert state. SESSION_ONLY and NEVER create no durable proposal or target mutation.
Open this section while signed in to load proposal state.
Trust boundary: Tier D installs the durable queue, policy, resolver, audit, and trusted server submission helper. The current R39 stream does not yet emit a structured proposal event, and no client endpoint may label arbitrary browser JSON as an assistant proposal.
Appearance
Visual-stage controls that are safe to preview locally.
Theme
Tier A established the Glitch Dragon / Ice Dragon stage as the authoritative theme.
Glass intensity
Lower glass reduces translucent blur surfaces while preserving the theme geometry.
Motion
Motion is state communication, not decoration. System mode respects your OS preference.
Motion mode
Reduced and Off suppress transition/animation timing without changing functional behavior.
Accessibility
Presentation controls stay independent from model behavior.
Text scale
Scales conversation and settings text without changing stored content.
High contrast
Raises foreground and boundary contrast.
Reduce transparency
Replaces translucent glass with more opaque surfaces.
Privacy
This panel reports real storage state. Controls that do not yet have an enforcement owner remain explicitly unavailable.
AccountNot signed in
Settings persistenceSession preview only
Lore & Memory storageSession unavailable
Rapport storageSession unavailable
AI proposal authorityUnavailable without durable account
No silent durability: opening or previewing settings never writes your profile. AI-originated durable changes must pass the configured Tier D policy; direct assistant mutation remains unavailable.
