Skip to content

USER_GUIDE.md

Latest commit

 

History

History
623 lines (450 loc) · 18.8 KB

USER_GUIDE.md

File metadata and controls

623 lines (450 loc) · 18.8 KB

Compose Booster User Guide

This guide provides detailed instructions for using all features of Compose Booster.

Quick Links: Getting Started | Main Interface | Quick Actions | Style Profiles | Settings | Keyboard Shortcuts | Troubleshooting

Table of Contents

  1. Getting Started
  2. Main Interface
  3. Quick Actions
  4. Custom Processing
  5. Style Profiles
  6. Response Regeneration
  7. Output Format Options
  8. Context Window Warnings
  9. Settings
  10. Keyboard Shortcuts
  11. Troubleshooting Guide

Getting Started

Installation

Windows:

  1. Download the installer from GitHub Releases or the Microsoft Store
  2. Run the installer and follow the prompts
  3. Launch Compose Booster from the Start Menu

macOS:

  1. Download from the Mac App Store (coming soon) or GitHub Releases
  2. Open the DMG file and drag Compose Booster to Applications
  3. Launch from Applications or Spotlight

Setting Up Your API Key

Compose Booster uses OpenRouter to access AI models. You'll need an API key to get started.

  1. Get an API Key:

  2. Enter Your Key in Compose Booster:

    • Open Settings (File → Settings or Ctrl/Cmd + ,)
    • Paste your API key in the API Key field
    • Click Test Key to verify it works
    • Click Save Settings

Cost Note: OpenRouter uses pay-as-you-go pricing. A typical email costs $0.001-0.01 depending on the model. Start with free credits from OpenRouter, then add funds as needed.

Your First Email

  1. Paste your email draft into the input area (left side)
  2. Click a Quick Action button (e.g., ⚡ Quick Polish)
  3. Wait a few seconds for the AI to process
  4. Review the improved email in the output area (right side)
  5. Copy to clipboard using the Copy button or Ctrl/Cmd + Shift + C

Main Interface

The main window is divided into several key areas:

Input Area

The left text area where you enter or paste your email draft.

  • Character Counter: Shows the current character count in the bottom-left corner
  • Paste Button: Click to paste from clipboard, or use Ctrl/Cmd + Shift + V
  • Clear Button: Removes all text from the input area

Tips:

  • Include the full email thread if you want context-aware responses
  • The AI can handle emails of any length, but very long threads may approach context limits

Output Area

The right text area displays the AI-improved version of your email.

  • Copy Button: Copies the output to clipboard in your preferred format
  • Clear Button: Clears the output area
  • Try Again Button: Generates an alternative version (appears after processing)

Status Bar

The bottom bar shows:

  • Model Used: Which AI model processed your request
  • Processing Time: How long the request took
  • Cost: Estimated cost for the API call (if available)
  • Status Messages: Errors, warnings, and success indicators

Quick Actions

Quick Actions are pre-configured one-click buttons for common email improvements.

Using Quick Actions

  1. Enter your email text in the input area
  2. Click one of the three Quick Action buttons:
    • ⚡ Quick Polish - Fast general improvement
    • 🎯 Professional - Formal business tone
    • ✨ Friendly - Warm, approachable tone
  3. Or use keyboard shortcuts: Ctrl/Cmd + 1, 2, or 3

Each Quick Action combines a specific model, prompt, and tone for consistent results.

Customizing Quick Actions

You can fully customize each Quick Action button:

  1. Open Settings → Quick Actions tab
  2. For each action, configure:
    • Name: Display name on the button
    • Icon: Emoji to show on the button
    • Model: Which AI model to use
    • Prompt: Which prompt template to apply
    • Tone: Which tone to use
    • Style: Optionally apply a style profile
  3. Reorder by dragging cards or using the arrow buttons
  4. Click Save Settings

Tip: The order of Quick Actions determines the keyboard shortcuts. The first action is Ctrl+1, second is Ctrl+2, etc.

Custom Processing

For more control, use the custom processing options below the Quick Actions.

Selecting a Model

Choose from various AI models with different capabilities and costs:

Model Best For Cost
Claude 3.5 Haiku Fast, everyday emails Low
GPT-4o Mini Good balance of speed/quality Low
Claude 3.5 Sonnet High-quality writing Medium
GPT-4o Complex, nuanced emails Medium
Claude 3 Opus Best quality, complex tasks High

To change the model:

  1. Click the Model dropdown
  2. Select your preferred model
  3. The selection is remembered for next time

Choosing a Prompt

Prompts tell the AI what kind of improvement to make:

  • Improve Email - General enhancement
  • Fix Grammar - Focus on grammar and spelling
  • Make Concise - Shorten while preserving meaning
  • Expand Detail - Add more detail and context
  • Formal Tone - Professional business language
  • Casual Tone - Relaxed, friendly language
  • Reply to Email - Draft a response to an email

Setting the Tone

Tones adjust the emotional quality of the output:

  • Professional - Formal, business-appropriate
  • Friendly - Warm and approachable
  • Neutral - Balanced, no strong emotion
  • Confident - Assertive and decisive
  • Empathetic - Understanding and supportive
  • Urgent - Conveys importance and time-sensitivity

Applying a Style

Optionally select a Style Profile to match your personal writing voice. See Style Profiles for details.

Style Profiles

What Are Style Profiles?

Style Profiles teach the AI to write in your personal style by providing example emails you've written. This uses a technique called "few-shot prompting" where the AI learns patterns from your examples.

Benefits:

  • Output sounds more like you, not generic AI
  • Consistent voice across all your emails
  • Preserve your unique writing quirks and preferences

Creating a Style Profile

  1. Open Settings → Styles tab

  2. Click + Add New Style

  3. Fill in the details:

    • Style Name: e.g., "My Work Style" or "Casual Me"
    • Description: Brief notes about the style (optional)
    • Sample Emails: Paste 2-3 emails you've written

  4. Click Save

Using Style Profiles

  1. In the main window, find the Style dropdown
  2. Select your style profile
  3. Process your email as usual
  4. The AI will attempt to match your writing patterns

Styles work with:

  • Quick Actions (configure in Settings → Quick Actions)
  • Custom processing (select from Style dropdown)

Best Practices for Samples

Do include:

  • 2-3 representative emails (more isn't always better)
  • Emails that showcase your typical vocabulary and phrasing
  • A mix of short and medium-length emails
  • Emails you're proud of and consider "your voice"

Avoid:

  • Very short emails (one-liners don't show enough style)
  • Extremely long emails (uses too many tokens)
  • Emails with lots of quoted text or signatures
  • Highly unusual or one-off emails

Sample length:

  • Each sample should be 100-500 words ideally
  • Total samples shouldn't exceed ~1500 words combined
  • Too many samples can hit context limits (see Context Warnings)

Response Regeneration

Try Again Button

Not happy with the output? Click Try Again to get an alternative version.

The button appears in the output area after processing completes. Each regeneration:

  • Uses the same input, model, prompt, and tone
  • Increases the "temperature" for more creative variation
  • Stores the previous version so you can go back

Version Navigation

After regenerating, use the version navigator to browse alternatives:

← [1 of 3] →
  • Previous (←): Go to the earlier version
  • Next (→): Go to the later version
  • Indicator: Shows which version you're viewing

Up to 10 versions are stored per session. Starting a new email clears previous versions.

How Temperature Works

Temperature controls how "creative" or "random" the AI's output is:

Temperature Behavior
0.7 (initial) Balanced, predictable output
0.9 More variety, some surprises
1.1 Creative, may deviate from conventions
1.3+ Highly varied, less predictable

Each regeneration increases temperature by 0.2, up to a maximum of 1.5.

Output Format Options

Control how your copied text is formatted for pasting into other applications.

Plain Text

Setting: Settings → General → Output Format → Plain Text

Copies the output as simple text with no formatting. Best for:

  • Email clients that don't support rich text
  • Plain text editors
  • Maximum compatibility

Markdown

Setting: Settings → General → Output Format → Markdown

Copies the output preserving Markdown formatting (headers, lists, bold, etc.). Best for:

  • GitHub, GitLab, or similar platforms
  • Markdown-aware applications
  • Documentation systems

HTML (Rich Text)

Setting: Settings → General → Output Format → HTML

Copies formatted HTML that pastes as rich text. Best for:

  • Microsoft Outlook
  • Gmail compose window
  • Word processors
  • Any application supporting rich text paste

Tip: If you're pasting into a rich text email client like Outlook or Gmail, HTML format preserves formatting like bold text and bullet points.

Context Window Warnings

Understanding Context Limits

AI models have a maximum "context window" - the total amount of text they can process at once. This includes:

  • Your email input
  • The prompt template
  • Style profile samples
  • System instructions

When your input approaches these limits, the request may fail or produce poor results.

Warning Levels

Compose Booster displays warnings when you're approaching limits:

Level Threshold Indicator
Normal < 80% No warning
Warning 80-95% Yellow warning message
Critical > 95% Red warning, may fail

Warnings appear before processing and give you the option to:

  • Continue anyway (risk of failure)
  • Cancel and reduce input
  • Switch to a model with larger context

Reducing Token Usage

If you hit context limits:

  1. Shorten the email thread

    • Remove older messages from the thread
    • Keep only the most recent relevant messages

  2. Reduce style samples

    • Use fewer or shorter sample emails
    • Remove samples from your style profile temporarily

  3. Use a larger model

    • Claude 3.5 Sonnet: 200K tokens
    • GPT-4o: 128K tokens
    • Claude 3 Opus: 200K tokens

  4. Simplify the prompt

    • Use a shorter prompt template
    • Remove unnecessary instructions

Settings

Access settings via File → Settings or press Ctrl/Cmd + ,

General Settings

Setting Description
API Key Your OpenRouter API key
Test Key Verify your API key is valid
Theme Light, Dark, or System (follows OS)
Font Size Interface text size (10-20px)
Output Format Clipboard format: Plain, Markdown, or HTML
Save Window Position Remember window size and location
Clear History on Exit Don't persist undo history
Developer Tools Enable DevTools shortcuts

Models Tab

Manage the AI models available in dropdowns.

Features:

  • Enable/Disable: Toggle models on/off with checkboxes
  • Add Custom: Add models from OpenRouter's full catalog
  • Reorder: Drag to change dropdown order
  • Detailed View: Toggle to show model IDs and pricing
  • Refresh: Update model list from OpenRouter (5-minute cache)
  • Reset: Restore default model list

Adding a custom model:

  1. Click + Add Custom Model
  2. Enter the model ID (from OpenRouter)
  3. Enter a display name
  4. Set the cost tier (Low/Medium/High)
  5. Click Add

Prompts Tab

Create and manage prompt templates.

Features:

  • Add New: Create custom prompts
  • Edit: Modify existing prompts
  • Duplicate: Copy a prompt as starting point
  • Delete: Remove custom prompts (defaults are protected)
  • Reorder: Drag to change dropdown order

Variables in prompts:

  • ${content} - Replaced with user's email text
  • ${tone} - Replaced with selected tone description
  • ${date} - Replaced with current date
  • ${style} - Replaced with style profile content

Tones Tab

Manage tone options for email styling.

Features:

  • Add New: Create custom tones
  • Edit: Change name and description
  • Duplicate: Copy existing tone
  • Delete: Remove custom tones
  • Reorder: Drag to arrange

Tone description: The description is included in the AI prompt to guide the output. Be specific:

  • Good: "Warm and friendly, uses casual language, includes personal touches"
  • Less effective: "Friendly"

Styles Tab

Manage Style Profiles for personalized output. See Style Profiles for detailed instructions.

Features:

  • Add New: Create a style with samples
  • Edit: Update name, description, or samples
  • Duplicate: Copy a style profile
  • Delete: Remove custom styles
  • Reorder: Drag to arrange

Quick Actions Tab

Configure the three Quick Action buttons.

For each action:

  • Name: Button label
  • Icon: Emoji to display
  • Model: AI model to use
  • Prompt: Prompt template
  • Tone: Tone setting
  • Style: Optional style profile

Reordering:

  • Drag cards to reorder, or use ↑/↓ arrows
  • Order determines keyboard shortcuts (1, 2, 3)

Advanced Tab

Feature Description
Export Settings Save all settings to a JSON file
Import Settings Load settings from a JSON file
Reset to Defaults Restore all settings to original values

Note: Export does not include your API key for security.

Keyboard Shortcuts

Processing

Shortcut Action
Ctrl/Cmd + Enter Process with current settings
Ctrl/Cmd + 1 Quick Action 1
Ctrl/Cmd + 2 Quick Action 2
Ctrl/Cmd + 3 Quick Action 3
Escape Cancel processing

Clipboard

Shortcut Action
Ctrl/Cmd + Shift + V Paste into input
Ctrl/Cmd + Shift + C Copy output

Editing

Shortcut Action
Ctrl/Cmd + Z Undo (restore previous input)
Ctrl/Cmd + K Clear input
Ctrl/Cmd + Shift + K Clear output

Application

Shortcut Action
Ctrl/Cmd + , Open Settings
Ctrl/Cmd + Q Quit (macOS)
Alt + F4 Quit (Windows)

Troubleshooting Guide

Common Issues

"API key is invalid"

  1. Check that you copied the full key (starts with sk-or-)
  2. Verify the key at openrouter.ai/keys
  3. Ensure the key hasn't been revoked
  4. Create a new key if needed

"Insufficient credits"

  1. Check your balance at openrouter.ai/credits
  2. Add credits to your account
  3. Consider using a lower-cost model

Processing takes too long

  1. Check your internet connection
  2. Try a faster model (Haiku or GPT-4o Mini)
  3. Shorten your input text
  4. Check OpenRouter status at status.openrouter.ai

Output is cut off or incomplete

  1. The model may have hit its output limit
  2. Try regenerating (Try Again button)
  3. Use a prompt that explicitly requests complete output
  4. Try a different model

Context limit errors

  1. Your input + style samples exceed the model's limit
  2. See Context Window Warnings for solutions
  3. Use a model with larger context (Claude 3.5 Sonnet: 200K)

App won't start / crashes

  1. Windows: Try running as Administrator
  2. macOS: Check Security & Privacy settings
  3. Delete config file and restart:
    • Windows: %APPDATA%/compose-booster/
    • macOS: ~/Library/Application Support/compose-booster/
  4. Reinstall the application

Getting Help

Appendix

Glossary

Term Definition
API Key Authentication credential for OpenRouter
Context Window Maximum tokens a model can process
Few-shot Prompting Teaching AI by example (Style Profiles)
Prompt Template Instructions telling the AI what to do
Temperature Controls randomness/creativity of output
Token Unit of text (~4 characters in English)
Tone Emotional quality of the writing

Model Context Limits

Model Context Window
Claude 3.5 Haiku 200K tokens
Claude 3.5 Sonnet 200K tokens
Claude 3 Opus 200K tokens
GPT-4o 128K tokens
GPT-4o Mini 128K tokens
Gemini 1.5 Pro 1M tokens
Gemini 1.5 Flash 1M tokens

Last updated: December 2025