Skip to content

Latest commit

 

History

History
160 lines (110 loc) · 6.52 KB

File metadata and controls

160 lines (110 loc) · 6.52 KB



Vapor
The Prompting & Context OS for AI power users.

Vapor is a macOS floating app that turns your voice, screenshots, and browser research into context-rich prompts, and sends them directly into any AI chat. Capture context, compose, and inject without leaving your workflow.

Screenshots

The floating prompt window - hold Fn to dictate, Vapor listens and transcribes on-device
Vapor showing live voice dictation with microphone indicator

Screenshot shelf - auto-detects screenshots, add to context with one keypress
Screenshot shelf with detected screenshots

Context tray - captured pages, articles, and research in a searchable sidebar
Context tray sidebar with captured pages

Browser injection - send prompts directly into any AI chat tab
Compressed prompt injected into AI chat

Prompt compression - 40–60% token reduction, meaning preserved
Prompt compression with original and compressed output

Keyboard-First Interface

Vapor is designed to be driven entirely from the keyboard. Every core action is a single keypress or key combo: summon the window, dictate, capture, compress, and send without ever touching the mouse.

  • ⌃⌥Space - summon Vapor from anywhere on your Mac
  • Fn (hold) - start voice dictation, release to stop
  • ⌘↩ - compress the current prompt and copy to clipboard
  • ⌘⇧P - send the prompt directly into your active AI chat tab
  • ⌘⇧S - jump to the Screenshot Shelf
  • ⌘⌥C - open the Context Tray

The goal: stay in your flow. Dictate a thought, capture a screenshot, grab context from a browser tab, compress it all together, and inject it into your AI, without switching windows.

Features

Voice Dictation

Hold Fn to speak. Vapor uses Apple's on-device speech recognition. No cloud, no latency, no API key, no privacy tradeoff. Your words appear in the editor as you speak. Release Fn when done.

Screenshot Shelf

Vapor auto-detects new screenshots on your Desktop. Open the shelf with ⌘⇧S, browse thumbnails, and add any screenshot to your prompt's context with one keypress. Vapor sees what you see.

Context Tray

Captured pages, articles, browser research in one searchable sidebar. Open with ⌘⌥C, search by keyword, and insert context directly into your prompts. Everything you capture is processed through a pipeline that extracts entities, generates summaries, and builds citations.

Agent Memory API

Vapor can index local agent sessions and expose them through an authenticated localhost API so agents can search prior conversation history and tool context. See docs/agent-memory-api.md.

Research Interrogation

Scan live browser tabs for structured data: tables, JSON, XHR feeds, articles. Vapor discovers data sources on the pages you have open and captures them into context automatically.

Browser Injection

Send prompts directly into ChatGPT, Claude, Gemini, Grok, Perplexity, and any site via the DOM picker. The Chrome extension (included in the DMG) connects Vapor to your AI chat tabs. No copy-paste needed.

Prompt Compression

Reduce tokens by 40–60% while preserving meaning. Vapor strips filler words and fuses related concepts into dense, efficient prompts.

Example:

Original (36 words) write a web component that renders a canvas that changes color from blue to golden as the time of day changes mimicking the light as it would be where you are located if there were no clouds out site in the sky
Compressed (15 words) webcomponent renders canvas changes color time day mimics sky location clouds out site sky

Choose between two backends:

Local LLM OpenRouter
Cost Free ~$0.01/1M tokens
Privacy On-device Cloud
Latency <1s ~1–2s
Setup Download model (2.3–4.7 GB) API key required

Browser Integration

Chrome extension connects Vapor directly to AI chat tabs:

  1. Load the extension (included in DMG)
  2. Copy the auth token from Vapor Settings → Browser
  3. Paste into the extension's Settings → Connected

Supported: ChatGPT, Claude, Gemini, Grok, Perplexity, and any site via the DOM picker.

Keyboard Shortcuts

Action Shortcut
Start dictation Fn (hold)
Compress & Copy ⌘↩
Copy Original ⌘⇧C
Post to Browser Tab ⌘⇧P
Copy & Clear ⌘K
Focus Vapor (global) ⌃⌥Space
Focus Screenshots ⌘⇧S
Focus Context Tray ⌘⌥C
Prompt History ⌘Y
Context Tray ⌘⇧E
Toggle Compact/Full ⌘\
Keyboard Shortcuts ⌘/

Requirements

  • macOS 14.0+ (Sonoma)
  • Apple Silicon or Intel
  • ~2–5 GB for local LLM model (optional - OpenRouter works without it)
  • Chrome (for browser integration)

Building

git clone https://github.com/memetic-research-labs/vapor.git
cd vapor/Vapor
open Vapor.xcodeproj
# Build & Run from Xcode

See CONTRIBUTING.md for detailed build instructions.

Code of Conduct

This project follows the Contributor Covenant Code of Conduct. By participating, you agree to uphold it.

Trademarks

The Vapor name, logo, and app icon are trademarks of Memetic Research Labs LLC and are not licensed under the MIT License. See TRADEMARKS.md for details.

Feedback & Contributing

Vapor is an early proof of concept - we're just getting started and every bit of feedback helps shape where it goes next. If something breaks, feels slow, or doesn't work the way you'd expect, please let us know.

  • Bug reports & feature requests - open an issue on GitHub
  • Ideas & discussion - we'd love to hear how you're using Vapor and what you'd want to see next
  • Pull requests - see CONTRIBUTING.md for build instructions

License

MIT - see LICENSE