AI Tagger for Thunderbird

A Thunderbird/Betterbird MailExtension that classifies and tags email using a local Ollama LLM. Define tags in plain English; the extension matches incoming mail in a single LLM call, applies Thunderbird tags, runs optional rules (move, archive, trash), and marks messages as processed.

Privacy-first: email content is sent only to your local Ollama instance (localhost). No cloud API required.

Features

• One LLM call per email — all active tag definitions are evaluated together for efficiency
• Natural-language tags — describe when a tag should apply; optional keywords for small models
• Automatic processing — classify new mail as it arrives (onNewMailReceived)
• Manual runs — context menu, folder batch, or all-unprocessed across accounts
• Action rules — move, archive, or trash when tag conditions match (AND/OR)
• Priority & stop processing — control which tags win when multiple match
• Processed marker — skip already-classified messages (configurable tag)
• Built-in tester — paste sample email content and inspect JSON results in Options

Requirements

• Thunderbird or Betterbird 128+
• Ollama running locally with a pulled model (default: llama3.1:8b)

Installation

From source (development)

git clone https://github.com/Mecallie/thunderbird-ai-tagger.git
cd thunderbird-ai-tagger
npm install
npm run build

Load the built extension in Thunderbird:

1. Add-ons and Themes → gear icon → Debug Add-ons
2. Load Temporary Add-on → select manifest.json (or install the zip from web-ext-artifacts/)

From release zip

Download ai_tagger_for_thunderbird-0.1.20.zip from Releases, then load it as a temporary add-on (or install via your preferred method).

Quick start

1. Open AI Tagger settings (toolbar button or Add-ons → Preferences).
2. Confirm Ollama URL and model, then click Test Ollama Connection.
3. On the Tags tab, add tags with descriptions and click Save All Tags.
4. Enable automatic processing on the LLM & Settings tab, or use Test & Manual Run to try a sample email.

Development

npm run lint      # web-ext lint
npm run build     # package to web-ext-artifacts/
npm run start     # run with web-ext (Thunderbird target)

Background script logs appear in about:debugging → This Thunderbird → AI Tagger → Inspect. The options page has its own console.

Architecture

File	Role
background.js	New-mail listener, classification orchestration, bulk runs
options.html / options.js	Settings UI (tags, rules, Ollama config, tester)
utils/ollama.js	Prompt builder, Ollama /api/chat, JSON parsing
utils/email.js	Extract plain-text body from messages
utils/tagManager.js	Sync tags with Thunderbird's tag system
utils/actionEngine.js	Post-classification rules (move, archive, trash)
utils/folders.js	Folder picker for action rules
utils/storage.js	browser.storage.local helpers

License

MIT — see LICENSE if present, or MIT as stated in package.json.

Topics

thunderbird · betterbird · thunderbird-addon · mail-extension · webextension · ollama · llm · email-classification · ai · local-ai · email-automation