Skip to content
Download for Mac

Sparks

Sparks is a direct prompt interface inside the universal overlay. Highlight text in any application, press a hotkey, type an instruction, and get a streamed response — without leaving the app you are working in.

The same overlay also provides access to Flows — pre-configured execution pipelines. Sparks and Flows are independent features that share the overlay as a launch surface. This page covers Sparks.


The overlay is a floating panel triggered by a global hotkey. It has two tabs:

  • Spark — Type a freeform instruction and get a streamed response. This is Sparks.
  • Flows — Browse and execute your saved Flows. See Flows UI.

The two tabs are independent. Sparks handles ad-hoc prompts; Flows runs pre-configured pipelines. They share the overlay window, the clipboard capture, and the output action system — but they are separate features with different behaviors.

PlatformDefault Hotkey
macOSCmd+Option+Space
Windows / LinuxCtrl+Alt+Space

The hotkey registers at the OS level — it captures input regardless of which application has focus. Configure the hotkey in Settings → Keyboard Shortcuts.

Set which tab opens by default in Settings → Sparks & Flows → Default Mode.


When you trigger the overlay, QARK captures your current text selection automatically:

  1. Highlight text in any application — an email paragraph, a code block, a browser article.
  2. Press the global hotkey.
  3. The highlighted text appears in a collapsible context block above the input field.

No manual pasting required. The captured text is sent as context alongside your instruction. If no text is selected, QARK falls back to clipboard contents.


Type your instruction in the input field and press Enter. The AI streams a response directly in the overlay.

  • Model — The overlay uses your configured Sparks model. Change it from the model selector in the input toolbar.
  • Thinking — Toggle the thinking/reasoning step from the brain icon in the overlay header. When enabled, the reasoning process displays in a collapsible block above the response.
  • Tools — Type @ in the input field to mention and invoke any configured tool, exactly as in the main composer. Tool results stream inline.
  • Attachments — Drag a file or image onto the overlay, or click the attachment icon. Attached images render as thumbnails in the context block.

After the response finishes streaming, action buttons appear below the result:

ActionBehavior
Copy to ClipboardPlaces the full response on your clipboard
Paste at CursorInserts the response at the cursor position in the source application
Insert Before SelectionPlaces the response immediately before the highlighted text
Insert After SelectionPlaces the response immediately after the highlighted text
Open in ChatTransfers the conversation to the main QARK window

Set your preferred default action in Settings → Sparks & Flows → Default Action. When a default is set, pressing Enter after the response finishes executes that action immediately.


While the AI is still streaming, an Open in main chat button (external link icon) appears in the overlay header. Click it to transfer the conversation to the main QARK window — the response continues streaming there without interruption.

This is useful when a Spark response turns out to be longer or more complex than expected and you want the full conversation interface to continue iterating, attach documents, or use additional tools.

  1. You send a prompt in the Spark tab. The AI begins streaming.
  2. The Open in main chat button appears in the overlay header once the conversation is created.
  3. Click the button. The main QARK window comes to the front and loads the conversation.
  4. If the AI is mid-stream, the main window picks up the stream seamlessly — no content is lost, no restart occurs. The backend never stopped generating; the main window adopts the active stream.
  5. The overlay closes and resets.

The conversation is saved to the Sparks system group. You can continue the exchange as a full conversation — ask follow-ups, attach files, enable tools, or switch models.


By default, the overlay vanishes after you execute an output action. To keep it visible while you work:

  • Click the pin icon in the overlay header to lock the overlay on screen.
  • A pinned overlay stays visible across app switches and remains interactive.
  • Click the pin icon again to unpin, returning to the default vanish-on-action behavior.

Pinned mode is useful when you are iterating — refining a prompt, comparing outputs, or working through a multi-step transformation.

When Settings → Sparks & Flows → Dismiss on Blur is enabled, the overlay automatically closes when it loses focus. This has no effect when the overlay is pinned or while a response is streaming.


  1. Select a dense paragraph in an email you are reading.
  2. Press Cmd+Option+Space. The overlay appears with the selected paragraph loaded as context.
  3. Type: “Rewrite this to be concise and professional, keeping all action items.”
  4. The AI processes your instruction and streams the rewritten text into the overlay.
  5. Click Paste at Cursor (or press Enter if it is your default action). The rewritten text inserts at the cursor position in the email.
  6. The overlay closes. You are back in the email app with the improved text in place.

Every Spark exchange is automatically saved to the Sparks system group in your conversation sidebar:

  • Every prompt, context, and response is preserved — nothing is lost.
  • Revisit any past Spark from the sidebar and continue it as a full conversation.
  • The Sparks group uses amber styling and cannot be renamed or deleted.

SettingLocationDescription
Global HotkeySettings → Keyboard ShortcutsSystem-wide key combination that opens the overlay
Default ModeSettings → Sparks & FlowsWhether the overlay opens on the Spark tab or the Flows tab
Default ActionSettings → Sparks & FlowsWhich output action executes on Enter after a response
Dismiss on BlurSettings → Sparks & FlowsWhether the overlay closes when it loses focus
Pin StateOverlay headerToggle to keep the overlay visible across interactions
Thinking ToggleOverlay headerEnable or disable the AI reasoning step