Inkstone Logo
Inkstone
DocsSettings

Settings & Configuration

The Room With All the Dials

The Room With All the Dials

Every app has a settings page. Most of them feel like filing your taxes. Inkstone's is organized into sections that actually correspond to things you'll want to touch, and the defaults are chosen so you can ignore most of them until you have a reason not to.

Here's what each one does, and more importantly, when you'd bother changing it.

API Configuration

The handshake between Inkstone and your AI

This is the handshake between Inkstone and whatever AI service is generating your responses. Three fields, one button.

API Base URLThe endpoint address of your provider. Quick-select chips let you swap providers without typing anything.
API KeyYour secret key, stored on-device in platform-native secure storage. Never sent anywhere except to the provider you've configured.
Model NameWhich AI model you're talking to. Tap “Fetch” to pull a searchable list, or type a model name directly if you already know what you want.
Test Connection

The Test Connection button does exactly what it says. Use it every time you change something here. It takes two seconds and saves you from staring at an error message mid-conversation wondering what went wrong.

Generation Parameters

The dials that shape how the AI writes

These are the dials that shape how the AI writes. Most of them have sensible defaults. All of them reward understanding.

Range0.0 to 2.0
Default0.8
0.00.51.01.52.0
FocusedBalancedCreativeWildChaos
Default: 0.8

Temperature controls how much the AI surprises itself. At low values (0.3 to 0.5), it picks the most probable word almost every time, producing focused, consistent, slightly predictable prose. Crank it higher (0.8 to 1.2) and the writing gets more expressive, more willing to take a left turn, more alive. Push it past 1.5 and you're in experimental territory where sentences start doing things nobody asked for.

For roleplay, the sweet spot is 0.8 to 1.0. Enough personality to feel like a real scene partner, enough coherence to stay in character.

Context Window

The AI's short-term memory

Range2,048 to 200,000 tokens
Default32,768

This is the AI's short-term memory. It determines how much conversation history gets sent with each message. A bigger window means the AI remembers more of your story, your character details, the thing you mentioned forty messages ago. It also means each message costs more of your API quota.

4K8K16K32Kdefault64K128K200K

The default of 32K is a solid middle ground. If you're running long, intricate stories and your model supports it, go bigger. If you're on a budget, go smaller and let the AI forget gracefully.

Context Summary

Keep the story going, even when memory runs out

Experimental Feature

This is marked experimental because it's a genuine trade-off: you gain longevity at the cost of granularity. Fine details will blur. The broad strokes survive.

Long conversations eventually outgrow even generous context windows. This feature handles that by summarizing older messages to free up space, keeping your most recent exchanges intact and replacing the rest with a compressed version of what happened.

Enable for New ChatsActivates summaries automatically for newly created chats. Default: Off
Summarize EveryHow many messages accumulate before triggering a summary. Default: 40 (range: 10 to 100)
Recent Messages to KeepHow many of the most recent messages stay in full, unsummarized context. Default: 20 (range: 5 to 50)
Max Summary TokensCaps the length of the generated summary. Default: 1,024 (range: 256 to 4,096)
Summary ModelUse a different model for summaries. Handy if your main model is expensive and you'd rather summarize with something cheaper. Leave empty to use your main model.
Custom Summary PromptWrite your own summarization instructions. The built-in prompt is optimized for roleplay, but if you have specific needs, this is yours to shape.

Ghostwriting

You supply the intent, the AI supplies the polish

Experimental Feature

A collaboration tool. Whether it's useful depends entirely on how you like to play.

Sometimes you know what you want your character to say but you'd rather not write it at publication quality yourself. Ghostwriting adds a rewrite button beside the chat composer. Type a rough draft of your message, tap the button, and the AI rewrites it in your character's voice before sending.

Enable GhostwritingShows the rewrite button beside the composer. Default: Off
Ghostwriting ModelUse a different model for rewrites, same logic as the summary model.
Ghostwriting PromptCustomize the rewriting instructions. The built-in prompt is tuned for roleplay ghostwriting, but you can override it.

Display Settings

How things look and feel

Auto-ScrollKeeps the chat pinned to the bottom when a new message arrives. Turn it off if you prefer to scroll manually. Default: On
Confirm Before DeleteShows a confirmation dialog before you delete chats, characters, or other items. Turning this off is living dangerously, but it's your call. Default: On
Reduce AnimationsSimplifies theme background effects for better performance. If your device is older or you just prefer a calmer screen, this helps. Default: Off
Font SizeScales text from 75% to 125% with a live preview, so you can see exactly what you're getting before you commit.

Data Management

Your data, your rules

Creates a ZIP file containing your characters (as PNG cards), lorebooks, system prompts, personas, and settings. You can optionally include chat histories, though be warned: that can make the file quite large if you've been at this a while.

One thing to know: this export is a safety net, not a migration tool. You can't automatically restore it into Inkstone. It's for keeping a backup of your work or manually moving individual pieces.

API Keys

API keys are not included in the export for security reasons. You'll need to re-enter them if you set up on a new device.

About

Default Values at a Glance

The full cheat sheet

SettingDefault
Temperature0.8
Top-P0.95
Max Response Tokens1,280
Context Window32,768
StreamingOn
Show ThinkingOn
Auto-ScrollOn
Confirm DeleteOn
Show Token CountOff
SummaryOff
GhostwritingOff
Debug ModeOff
Start Here

These defaults are chosen so you can start chatting immediately without touching anything. Come back and adjust once you know what you want more (or less) of.

Use to navigate between pages