Version 2.4.1 Highlights & Improvements:

• Native HTTP Compatibility (Manifest V3): Explicit CSP policy declaration allowing native plain HTTP requests without forced HTTPS upgrades. Connect your private LibreTranslate, Ollama, or LM Studio instances on localhost or local area networks (LAN) directly without requiring an SSL reverse proxy.
• LibreTranslate URL Normalization: Automatic detection and trimming of trailing /translate paths entered in preferences to prevent duplicate endpoints.
• Zero-Click Mode (Optional): Automatically translates foreign emails upon opening with zero clicks required. Powered 100% offline by Thunderbird built-in language detection engine (zero bytes sent over network for detection).
• Optional Local Language Detection: Preference toggle to identify the source language locally before any network call. Ideal for LibreTranslate and local LLMs lacking built-in server-side auto-detection.
• Optional 1-Click Translation: New setting allowing instant translation directly upon clicking the toolbar button or pressing Alt+Shift+T, with no second click required in the banner.
• Dual Translation & Smart Inversion: Automatic source ↔ target swap when an explicit language matches target, or automatic fallback to an alternate target language when the email is already in your primary language.
• 60s Extended Timeout for LibreTranslate: Increased timeout to 60 seconds for LibreTranslate and local engines (Ollama, LM Studio) to comfortably accommodate CPU neural translation without premature timeouts.
• Ergonomics Section & Modern Toggle Switches: Replaced standard checkboxes with smooth, accessible toggle switches featuring smart mutual exclusion between 1-Click and Zero-Click modes.
• NLP Filtering & Quota Protection: Advanced preprocessing before CLD2 analysis (stripping CSS blocks, quotes, and URLs) combined with strict confidence thresholds to prevent false positives and protect your paid API balances (DeepL, OpenAI, Gemini).

Version 2.3.10 Highlights & Improvements:

• Hardened Google Translate Resilience: Migrated to the modern translate-pa.googleapis.com endpoint explicitly authorizing Mozilla extensions, resolving 429 rate-limiting and Captcha challenges.
• Automatic Resilient Fallback: Seamless automatic switch to the free MyMemory service if Google Translate ever encounters rate limits or network issues.
• New MyMemory Provider: Built-in free translation without any API key required with smart sentence chunking (5,000 chars/day per IP, expandable to 50,000 chars/day with an optional email address).
• Compose Mode Hardening: Clear alert when neither Subject nor Body is checked (no more false successes), and immediate error propagation if body translation fails.

Version 2.3.9 Highlights & Improvements:

• Universal Custom Server Support: Re-enabled on-demand dynamic host authorization for any custom or self-hosted OpenAI-compatible endpoint (DeepSeek, OpenRouter, Claude proxies, private VPS, corporate gateways).
• Seamless 1-Click Authorization: Host permission requests trigger strictly on direct user clicks with native Thunderbird confirmation dialogs.
• Full Control & Reversibility: Each authorized domain automatically gets its own dedicated toggle switch in Thunderbird's Add-ons Manager and can be revoked at any time in 1 click.
• Hardened Stability: Strict WebExtension host match pattern compliance and eliminated background warnings.

Version 2.2.0 — Cloud LLMs Hub (Gemini, OpenAI, Groq, Mistral) & Local LLMs (Ollama, LM Studio)

✨ Major Highlights:
- Dedicated Google Gemini API: Native REST integration with Google AI Studio featuring real-time dynamic model synchronization and ultra-fast gemini-3.5-flash-lite by default (with full support for gemini-3.7-flash, gemini-2.5-flash, etc.).
- Universal OpenAI-Compatible Hub: Connect any Cloud or Local LLM in one click via the standard /v1/chat/completions API (OpenAI gpt-4o-mini, Groq Cloud llama-3.3-70b-versatile, Mistral AI, Ollama, LM Studio, vLLM).
- 100% Sovereign Local Translation: Seamless support for local servers like Ollama (http://localhost:11434) and LM Studio (http://localhost:1234) with zero API keys required.
- Robust Translation Prompting: Strict system prompts with automated raw text sanitation ensuring clean, high-fidelity translations without preamble or markdown fences.
- Extended Options Page: Dedicated settings forms, instant live model sync button, presets and smart autofill.