🤖 AI Tag Manager for Thunderbird

An intelligent Thunderbird extension that uses artificial intelligence to automatically manage, optimize, and clean up your email tags.

✨ Features

AI-Powered Tag Management

• Find Similar Tags: AI analyzes your tags and finds similar or duplicate tags (e.g., "Project", "Projects", "project")
• Improve Tag Names: Get suggestions for more consistent and clearer tag names
• Automatic Categorization: Organize your tags into meaningful categories (Projects, Priorities, Status, etc.)
• ⭐ Interactive Suggestions: All AI suggestions are displayed in a clear preview page - you decide which ones to apply!
• 🔥 Deep Analysis Mode: Analyze ALL tags in a single request for comprehensive similarity detection

Tag Management

• Tag Analysis: Overview of all tags with usage statistics
• Merge Tags: Automatically merge similar tags
• Cleanup: Remove unused tags automatically
• Consistency: Ensure uniform naming conventions
• Smart Warnings: Token limit detection with actionable recommendations

Supported AI Providers

• OpenAI (GPT-4o, GPT-4o-mini, GPT-3.5)
  • ⭐ GPT-4o recommended for 1000+ tags (16K output limit)
• Anthropic (Claude 3.5 Sonnet, Claude 3.5 Haiku)
  • ⭐ Claude 3.5 Sonnet recommended for 1000+ tags (8K output limit)
• Google AI (Gemini 1.5 Flash, Gemini 1.5 Pro)
  • ⭐ Gemini 1.5 Pro recommended for 1000+ tags (8K output limit)
• Ollama (local, free - llama3.2, mistral, gemma2, etc.)

📦 Installation

From Source

1. Clone or download the repository

   git clone https://github.com/hendkai/ai_tag_manager
   cd ai_tag_manager

2. Install in Thunderbird

   • Open Thunderbird
   • Go to Menu → Add-ons and Themes (or press Ctrl+Shift+A)
   • Click the gear icon ⚙️
   • Select Install Add-on From File...
   • Navigate to the folder and select manifest.json

Alternative: Debug Mode

For development and testing:

1. Open about:debugging in Thunderbird
2. Click This Thunderbird
3. Click Load Temporary Add-on...
4. Select the manifest.json file

🚀 Getting Started

1. Configure API Key

Open the extension settings:

• Click the AI Tag Manager icon in the toolbar
• Select ⚙️ Open Settings

Choose your preferred AI provider and enter your API key:

OpenAI

• Register at platform.openai.com
• Create an API key at API Keys
• Recommended model: gpt-4o for 1000+ tags, gpt-4o-mini for <1000 tags

Anthropic (Claude)

• Register at console.anthropic.com
• Create an API key
• Recommended model: claude-3-5-sonnet-20241022 for best quality

Google AI

• Create an API key at Google AI Studio
• Recommended model: gemini-1.5-pro for 1000+ tags

Ollama (Local & Free!)

• Install Ollama
• Run: ollama pull llama3.2
• No API costs! Runs completely locally.

2. Analyze Tags

1. Click the AI Tag Manager icon
2. Click 📊 Analyze Tags
3. You'll see an overview of your tags

3. Use AI Functions

Find Similar Tags

• Click 🔍 Find Similar Tags
• Enable 🔥 Deep Analysis Mode for comprehensive analysis (recommended for 500+ tags)
• AI analyzes your tags and suggests merges
• Preview page opens: Review each suggestion individually
• Select which suggestions to apply
• Click "Apply Selected" to execute chosen changes

Improve Tag Names

• Click ✨ Improve Tag Names
• Get suggestions for more consistent names
• Interactive selection: Choose only the improvements you like
• E.g., "ToDo" → "todo", "IMPORTANT" → "important"

Categorize Tags

• Click 📁 Categorize Tags
• AI sorts your tags into meaningful categories
• Overview view: Informative display of your organized tags
• Helps organize large tag collections

📊 Understanding Token Limits

What are tokens?

Tokens are pieces of text that AI models process. When analyzing tags, the AI needs tokens for both reading your tags (input) and writing suggestions (output).

Why does this matter?

Each AI model has a maximum output limit. With many tags (500+), the AI might find hundreds of similar groups, but if it hits the token limit, it can only return a fraction of them.

Model Recommendations

Tag Count	Best Model	Output Limit	Quality
< 500 tags	Any model	-	Good
500-1000 tags	GPT-4o-mini, Gemini 1.5 Flash	16K / 8K	Good
1000-2000 tags	⭐ GPT-4o, Claude 3.5 Sonnet, Gemini 1.5 Pro	16K / 8K	Excellent
2000+ tags	⭐ GPT-4o only	16K	Excellent

What happens with insufficient tokens?

If you have 1000 tags and use a smaller model, the AI might only return 30-50 groups instead of 100+. The extension will:

• ✅ Automatically recover partial results
• ⚠️ Display a warning with coverage percentage
• 💡 Recommend appropriate models for your tag count

⚙️ Settings

Tag Management Options

• Automatic Merging: Tags are merged after confirmation
• Similarity Threshold: How similar tags must be (0.0 - 1.0)
• Remove Unused Tags: Automatic deletion during cleanup
• Case Sensitive: Distinction during analysis

Advanced Options

• Max. Tags per Analysis: Limit for very large tag lists
• Detailed Logging: For debugging and development
• 🔥 Deep Analysis Mode: Send ALL tags in ONE request for maximum coverage

🔒 Privacy & Security

• API keys are stored locally only (in Thunderbird's Storage API)
• No data collection: This extension collects no user data
• Only tag names are sent to AI APIs (never email content!)
• Ollama option: Use AI completely locally without cloud connection

🛠️ Development

Project Structure

ai_tag_manager/
├── manifest.json           # Extension manifest
├── scripts/
│   ├── background.js       # Background logic
│   └── aiIntegration.js    # AI API integration
├── popup/
│   ├── popup.html          # Popup UI
│   └── popup.js            # Popup logic
├── results/                # ⭐ NEW
│   ├── results.html        # Preview page for AI suggestions
│   └── results.js          # Interactive selection logic
├── options/
│   ├── options.html        # Settings page
│   └── options.js          # Settings logic
├── icons/                  # App icons
└── README.md

Debugging

1. Open about:debugging in Thunderbird
2. Click on the extension
3. Open the Console for logs
4. Enable "Detailed Logging" in settings

View Logs

// In Browser Console (Ctrl+Shift+J)
browser.runtime.sendMessage({action: 'analyzeTags'})
  .then(console.log)

📝 Known Limitations

• API Costs: OpenAI, Anthropic, and Google AI incur costs (use Ollama for free alternative)
• Rate Limits: AI APIs have request limits
• Token Limits: Large tag lists (1000+) may require larger models for complete results
• Thunderbird >= 91: Requires Thunderbird version 91 or higher

🤝 Contributing

Contributions are welcome! Please create a pull request or open an issue.

Set Up Development Environment

# Clone the repository
git clone <repository-url>
cd ai_tag_manager

# Load in Thunderbird (about:debugging)

📄 License

MIT License - see LICENSE file

🆘 Support

For questions or issues:

1. Check the FAQ
2. Open an issue on GitHub
3. Enable "Detailed Logging" for debug information

❓ FAQ

Why doesn't the AI find similar tags?

• Your tags might already be very well organized
• Try reducing the similarity threshold
• Ensure the API key is correct
• For 1000+ tags, use a larger model (GPT-4o recommended)

Are my emails sent to the AI?

No! Only the tag names are analyzed, never email content.

Can I use the extension without internet?

Yes, with Ollama! Install Ollama locally and all AI functions run offline.

How much does it cost?

• Ollama: Free (local)
• OpenAI: ~$0.001 per request (gpt-4o-mini), ~$0.005 per request (gpt-4o)
• Anthropic: ~$0.003 per request (Claude 3.5 Haiku), ~$0.015 per request (Claude 3.5 Sonnet)
• Google: Free (with limits), then ~$0.001 per request

Which AI provider is best?

• For cost: Ollama (free) or OpenAI gpt-4o-mini
• For quality with large tag lists (1000+): GPT-4o or Claude 3.5 Sonnet
• For privacy: Ollama (completely local)
• For speed: Gemini 1.5 Flash or GPT-4o-mini

What is Deep Analysis Mode?

Deep Analysis Mode sends ALL your tags in ONE single AI request for comprehensive comparison. Benefits:

• ✅ Finds every possible similarity in a single pass
• ✅ Compares all tags with each other
• ✅ Best results for large tag collections
• ⚠️ Requires larger models (GPT-4o, Claude Sonnet) for 1000+ tags
• ⚠️ Higher API costs but fewer total requests

I'm getting token limit warnings. What should I do?

When you see "Token limit reached" warnings:

1. Best solution: Switch to a larger model (GPT-4o for 1000+ tags)
2. Alternative: Disable Deep Analysis Mode (uses batches instead)
3. Optional: Remove unused tags first to reduce total count

The extension automatically recovers partial results and shows you exactly how many groups were found vs. expected.

🎯 Roadmap

• Deep Analysis Mode for comprehensive tag comparison
• Token limit detection and warnings
• Automatic partial result recovery
• Model recommendations based on tag count
• Undo/Redo for tag changes
• Tag rules and automation
• More AI providers (Mistral, etc.)
• Tag suggestions based on email content
• Import/Export tag configurations
• Multi-language user interface

---

Developed with ❤️ for better email management