Torcharizer is a Thunderbird add-on that uses a local Ollama model to analyze the full HTML of emails and displays an AI-generated summary at the top of the message view.
- Summarizes emails using a local Ollama model
- Adds a Markdown panel labeled "AI Analysis" at the top of each email
- Works with HTML and plain text emails
- Session-based summary caching
- Easy enable/disable via settings
- Support for custom online APIs (OpenAI, Gemini, etc)
- Queued summary processing for switching emails
- Persistent summary cache across restarts
- Support for remote/cloud Ollama models
- Improved error handling for unreachable Ollama
- Enhanced compatibility with more Thunderbird versions
- Scraping content from HTML to help AI may not work properly.
- System Prompt may not be well-defined or could lead to unexpected results.
- Switching emails while a summary is generating may show the summary for the wrong email or overwrite the new one. (Queued processing is planned.)
- Summaries may be delayed for very large emails or if Ollama is busy.
- If Ollama is unreachable or blocked by CORS, the extension may silently fail to show a summary.
- Some Thunderbird versions may not support all required APIs (e.g., messageDisplayScripts).
- Only local Ollama models are supported; remote or cloud models are not for now.
- Summaries are cached per session, but cache is not persistent across restarts.
- Install Ollama and ensure it is running locally (
ollama serve).- For development, you may need to set the environment variable to allow cross-origin requests:
$env:OLLAMA_ORIGINS="*"
- For development, you may need to set the environment variable to allow cross-origin requests:
- Install the Add-on in Thunderbird:
- Go to
Tools > Add-ons and Themes > Extensions > Settings gear > Install Add-on From File. - Select
torcharizer.zipfrom this folder (rename the zip if needed).
- Open an Email in Thunderbird. The "AI Analysis" panel will appear at the top of the message within a few seconds.
- Configure Settings using the Torcharizer toolbar button:
- Toggle "Enable AI summaries" to turn summaries on or off globally.
- Unzip and load as a temporary add-on from the manifest file for quicker iteration.
- Use Add-on Debugging (
Tools > Developer Tools > Debug Add-ons) to view logs and errors in the background page console. - Check for streaming network activity to Ollama when emails are opened.
messagesModify,messagesRead: Read and annotate displayed messages.storage: Persist user preferences.tabs: Coordinate messaging and fallback content script injections.- Host:
http://127.0.0.1:11434/*to communicate with local Ollama.
- Fetches the full message and extracts the HTML body (or falls back to text rendered as HTML)
- Sends the HTML and subject to Ollama
/api/generatewith a structured system prompt - Streams the AI response and renders a Markdown panel at the top of the email view
- Uses programmatic registration for message display scripts and loads content scripts into open tabs as needed
We need your help to resolve the issues above! Please open issues or pull requests for bug reports, feature requests, or improvements. All contributions are welcome.
See PRIVACY.md. All processing is local; no telemetry or remote data exfiltration.
This project is licensed under the MIT License. See LICENSE for details.