Skip to content

Latest commit

 

History

History
94 lines (71 loc) · 2.77 KB

File metadata and controls

94 lines (71 loc) · 2.77 KB

AutoSort+ v1.2.3.1-ollama-test - Ollama Support (Test Release)

🧪 Test Release - Ollama Local AI Integration

This is a test release with experimental Ollama support for local AI email classification. Please report any issues!

✨ What's New

Local Ollama Support:

  • 🏠 Run AI email classification completely locally with Ollama
  • 🔒 No data sent to external APIs
  • 🆓 No API keys or rate limits
  • 🎯 Support for any Ollama model (tinyllama, llama3.2, phi, gemma, etc.)
  • ⚙️ CPU-only mode option for systems without GPU

🚀 Quick Start with Ollama

Prerequisites

  1. Install Ollama: https://ollama.com/download
  2. Pull a model: ollama pull tinyllama (or llama3.2, phi, gemma, etc.)
  3. Verify it's running: ollama list

Setup in Extension

  1. Open AutoSort+ settings
  2. Select Ollama as AI provider
  3. Leave URL as http://localhost:11434 (default)
  4. Select your model from dropdown
  5. Click Test Connection - should show your installed models
  6. Click Save Settings

Test It

  1. Select an email
  2. Right-click → Analyze with AI
  3. Watch as Ollama classifies it locally!

🐛 Known Issues & Debugging

If you get errors:

  1. Check Ollama is running:

    curl http://localhost:11434/api/tags

    Should return list of models

  2. Test chat directly:

    curl -X POST http://localhost:11434/api/chat \
      -H "Content-Type: application/json" \
      -d '{"model":"tinyllama","messages":[{"role":"user","content":"test"}],"stream":false}'

    Should return a response

  3. Enable debug logging:

    • Open Browser Console (Ctrl+Shift+J)
    • Watch for [Ollama] messages during analysis
    • Look for any error messages
  4. Common issues:

    • 403 errors: Fixed in this release with tab injection approach
    • Timeout: Increase wait time or use faster model
    • Model not found: Run ollama pull <model-name>

Report Issues

Please include:

  • Thunderbird version
  • Ollama version (ollama --version)
  • Model used
  • Console logs showing the error

📝 Technical Details

This release uses a tab injection approach to bypass Thunderbird's fetch restrictions:

  1. Opens hidden tab at Ollama origin
  2. Injects script to make POST request
  3. Retrieves result and closes tab
  4. Works like curl - no special permissions needed

🔄 Upgrading from Previous Version

If you tested earlier Ollama builds:

  1. Uninstall old version
  2. Install this XPI
  3. Reconfigure Ollama settings
  4. Test connection again

⚠️ Disclaimer

This is a test release. The Ollama integration is experimental and may have bugs. Please backup your settings before installing.


Installation: Download autosortplus.xpi and drag into Thunderbird Add-ons page

Feedback: Open an issue on GitHub with your results!