TempurAI

Frictionless capture from anywhere to your LLMs.

TempurAI is a macOS menu bar tool that removes the back-and-forth friction when moving context (text, screenshots, files) from PDFs, decks, websites, and other apps into web-based LLMs like ChatGPT. Think of it as “the capture surface” that IDEs like Cursor give you natively—now available system-wide for everything else you work with.

• Always-available menu bar popover
• Drag & drop anything (text, images, files)
• Organize into lightweight “Collections” before sending
• Copy everything in one go and paste into ChatGPT/Claude/etc.
• Zero cloud dependency; everything stored locally

---

Why TempurAI?

LLMs are incredible once you get your context in. The problem is getting it in. Moving between PDF viewers, browsers, slide decks, and tools creates tedious copy/paste cycles and fragments your thought process. TempurAI compresses that workflow into one focused flow:

1. Capture context fast (drag/drop or paste)
2. Curate it once (reorder, prune)
3. Send it to your LLM of choice in a single paste

You stay in flow. Your model gets better prompts. Everyone wins.

---

Features

• Menu Bar Popover

  • Click the 🍣 icon to open the capture surface instantly.
  • Prominent Close button to dismiss quickly.

• Drag & Drop Capture

  • Drop text, images, or files directly into the popover.
  • Text files load as text; image files load as images.
  • Mixed content supported.

• Paste Box

  • Stage text with ⌘V, then press “Save text” (or ⌘↩) to commit to the active collection.

• Collections

  • Create and switch between collections to group related context.
  • Simple selection sheet with “Use Existing” or “Create New”.

• Quick Reorder

  • Hover an item to reveal Up/Down buttons for precise ordering.
  • Items at the top/bottom intelligently disable movement in those directions.

• Copy All

  • Consolidates the active collection into one paste ready for ChatGPT/Claude/etc.
  • Images currently appear in the combined clipboard as placeholders like “[Image 1]” to preserve order and context.

• Local and Private

  • Your data stays on-device (UserDefaults with JSON encoding for assets).
  • No network calls or accounts required.

---

Install & Run

Requirements:

• macOS 13+ (Ventura or newer recommended)
• Swift toolchain (Xcode 15+ recommended)

Build and run:

• Using SwiftPM:
  • swift run
• Using Xcode:
  • Open Package.swift in Xcode and run the tempur target.

You’ll see a 🍣 icon appear in the macOS menu bar. Click it to open the TempurAI popover.

Building for Release

To build a signed and notarized DMG for distribution:

1. Set up your Apple Developer credentials (as environment variables):

   export NOTARIZATION_ACCOUNT="[email protected]"
   export NOTARIZATION_PASSWORD="your-app-specific-password"
   export APPLE_TEAM_ID="your-team-id"
   export DEVELOPER_ID_APPLICATION="your-cert-hash"

2. Run the build script:

   ./build-release.sh

3. Find your DMG: The signed and notarized DMG will be created in the dist/ directory.

Security Note: Never commit your Apple Developer credentials to version control. The build script uses environment variables to keep your credentials secure and out of the repository.

---

Usage

• Open:

  • Click the 🍣 menu bar icon to toggle the popover.

• Choose a Collection:

  • First run will ask you to create or pick a collection.
  • Switch collections later via the “Collections” header menu.

• Capture:

  • Drag & drop text, images, or files into the “Collect to this board” card.
  • Or paste text into the editor and press “Save text” (or ⌘↩).

• Curate:

  • Hover saved items and use ↑/↓ to reorder.
  • Delete items via the trash button.
  • “Clear” removes all items in the active collection.

• Send:

  • Use “Copy All” to place everything (in order) onto your clipboard.
  • Paste into ChatGPT/Claude or any LLM UI.

• Close:

  • Click the Close button in the header to dismiss the popover.

Tips:

• Consider one collection per task (e.g., “Quarterly Review”, “Vendor Research”, “Paper Summary”).
• Reorder strategically—front-load the most important context to guide the model.

---

Keyboard & Interactions

• Paste staged text: ⌘V (into the text box)
• Save staged text: ⌘↩
• Hover actions on items: Trash, Move Up, Move Down
• Close the popover: Close button in header

---

Roadmap

• One-click send to ChatGPT/Claude (open model UI, auto-focus, paste)
• Global hotkey to toggle the popover
• Browser Share Extension (Safari/Chrome) and Services integration
• OCR for PDFs/screenshots to extract text
• Rich “Copy All” with Markdown, image captions, and source attribution
• Multi-select operations (batch delete, batch move)
• Encrypted sync across devices
• Smarter deduplication and snippet tagging

Have a feature request? Please open an issue.

---

Architecture Overview

High level:

• Status Bar + Popover: Always-available capture surface
• SwiftUI views render the UI and handle user interactions
• ObservableObject store manages state and persistence

Key components:

• AppDelegate + StatusBarController

  • Creates the NSStatusItem (🍣) and shows an NSPopover with the SwiftUI content.
  • Handles Close requests via a NotificationCenter event.

• CollectorView (SwiftUI)

  • Main UI: header, capture card, saved items card, onboarding.
  • Presents collection management sheet.

• CollectionStore (ObservableObject)

  • Manages collections, activeCollectionID, and items.
  • Persists via UserDefaults (keys: tempur.collections, tempur.activeCollectionID).
  • Encodes images as PNG, stored alongside text content.

• Models

  • ScrapCollection: id, name, createdAt, items.
  • CollectedItem: id, timestamp, content.
  • CollectedContent: .text(String) | .image(NSImage)

• Notable Views

  • DropZoneView: animated drop target for files/images/text.
  • PasteBoxView: staged text capture with save action.
  • CollectedItemsList, CollectedItemRow: list and item UI, reordering controls.
  • CollectionSelectionView: sheet to switch/create collections.

---

Privacy

• Local-first by design. No network calls.
• Data persists on-device in UserDefaults as JSON blobs (images stored as PNG).
• You can delete items or clear a collection at any time.

---

Troubleshooting

• I don’t see the 🍣 icon:

  • Ensure the app built successfully and is running.
  • macOS may group menu extras—adjust your menu bar items or restart the app.

• Drag & drop isn’t doing anything:

  • Confirm you’re dropping onto the highlighted drop zone.
  • Try a simple text file or image first to sanity-check.

• Reordering isn’t reflected:

  • You must hover the item to reveal the up/down controls; items at the top/bottom can’t move past boundaries.
  • Use “Copy All” after reordering; the order is preserved in the clipboard.

• Copy All doesn’t include image pixels:

  • Current “Copy All” consolidates text and uses [Image N] placeholders for images to preserve context and order. Full rich content export is on the roadmap.

---

Contributing

• Clone the repo, run with swift run.
• Submit issues with clear reproduction steps and environment details.
• PRs welcome—please keep changes focused and include rationale.

---

License

MIT License

Copyright (c) 2025 setrf

This project is licensed under the MIT License. See LICENSE for details.

Screenshots (Optional)

• Collections sheet with selection and create-new
• Capture card with drop zone and paste box
• Saved items with hover controls (reorder, delete)

Add screenshots to showcase real workflows (PDF + web article + deck → “Copy All” → paste into ChatGPT).

---

Credits

Inspired by the ergonomics of IDE-integrated LLMs (e.g., Cursor) and designed to bring that “frictionless context handoff” to the rest of your desktop workflow.