Preferences
Choose what Enter does in the composer, the model and response style every new chat starts with, what happens when you switch model in a chat, when long chats are compacted, and the theme.
Preferences decide how a new chat starts and how the composer behaves for you: which model answers, in which response style, and whether Enter sends your message or starts a new line. They also decide when long chats are compacted, and whether StickyPrompts looks light or dark. They are personal, so changing them affects only your own account. Which models you can choose from is set by your workspace admins.
Open settings from the user menu at the bottom of the left sidebar and choose Preferences.
The settings at a glance
| Setting | What it does | Default |
|---|---|---|
| On enter (1) | Whether Enter sends your message or adds a new line. | Enter sends |
| Preferred AI model (2) | The model every new chat starts with. | |
| Preferred AI style (3) | The response style every new chat starts with. | Prompty |
| Model selection behavior (4) | What happens to your default when you switch model in a chat. | Always ask |
| Auto compact conversations | Summarizes older messages when a chat gets too long. | On |
| Compact when | The point at which a chat is compacted. | 80% of the model’s context window |
| Stay under long context pricing | Compacts before a model starts charging long context prices. | On |
| Theme | Light or Dark. |
Change what you need, then select Save Preferences. Nothing changes until you save.
On enter
On enter has one switch: Add new line to text.
- Off (the default): pressing Enter sends your message.
- On: pressing Enter adds a new line in the composer, and you send with the Send button.
Switch it on if you often write long prompts with several paragraphs, lists or pasted text, and don’t want a stray Enter to send a half-finished message.
Preferred AI model
The model new chats start with. You can still pick a different model in any chat; this only sets where you start.
The list opens with three automatic options:
| Option | What it does |
|---|---|
| Auto (Best Quality) | StickyPrompts picks the strongest model for each chat. |
| Auto (Balanced) | A balance of quality and speed. |
| Auto (Cheapest) | A lighter, lower-cost model that suits the request. |
Below them is every model your workspace allows you to use. Choosing an Auto option lets StickyPrompts pick the best model for the job in each chat; choosing a named model, such as Claude Opus 5.5, means every new chat starts on that model. For help choosing, see Choosing a model and Compare models.
Preferred AI style
The response style new chats start with: Prompty, Chatty or any style you created yourself. If you have written a style you use most days, for example a Board-ready style for leadership updates, make it your default here. You can still change style in any chat with Style in the composer.
Styles are created and edited in Settings > Response Styles. See Response styles.
Model selection behavior
“Choose how you want to handle model selection when changing AI models.” When you switch model inside a chat, should that model also become your Preferred AI model?
| Option | What happens when you switch model in a chat |
|---|---|
| Always ask (the default) | StickyPrompts asks whether to make the new model your default. |
| Always set selected model as default | The model you switch to becomes your default straight away. |
| Never set selected model as default | Your preferred model stays as it is. The switch applies to that chat only. |
Whenever your default changes this way, StickyPrompts confirms with “Default model updated!”.
Conversation compaction
A long chat can outgrow a model’s context window, the amount of conversation the model can take in at once. Compaction keeps it going: StickyPrompts has the model summarize the older messages, and later replies work from that summary and the newer messages. Earlier messages stay visible in the chat. You can also compact a chat yourself at any time; see Compact a long conversation.
| Setting | What it does |
|---|---|
| Auto compact conversations | “Automatically summarize older messages when a conversation gets too long.” On by default. Switched off, a chat is compacted only when you select Compact yourself. |
| Compact when | Context reaches a percentage: compact once a chat uses a share of the model’s context window, set with Context window percentage from 10% to 95% (80% by default). Context reaches a token count: compact at a fixed number of tokens instead (200,000 unless you change it). A threshold never goes above 95% of a model’s context window. |
| Stay under long context pricing | “Compacts before a model starts charging long context prices.” Some models charge a different rate once a request passes a certain length; with this on, StickyPrompts compacts before a chat gets there. On by default. |
StickyPrompts checks the threshold after each reply, so a chat is compacted between messages, never in the middle of an answer. Compaction applies to your own chats, including the ones you drive through the API; it doesn’t run inside model comparisons or workflow runs. Select Save Preferences to apply your changes.
Appearance
Theme switches StickyPrompts between Light and Dark. It “Applies immediately and is remembered on this device”, so you can use a different theme on your laptop and your phone.