- Quota-aware Gemini retries: On HTTP 429, AutoSort+ honors Google's retry delay and retries the request up to five times. Failed responses do not increment the add-on's local usage counter; successful API responses do, even if the generated label cannot be used.
- More reliable Gemini classification: Increased the output-token budget and use any returned label text when Gemini reports
MAX_TOKENS. - Automatic usage refresh: Gemini counters are reconciled every 10 seconds and when usage is refreshed. Each key's countdown starts only when that key reaches the add-on's 20-request daily cap; keys below the cap show that the countdown starts at the limit. One key's timer does not affect another. Optional desktop notifications announce resets for keys that had usage.
- Accurate parallel usage counters: Concurrent requests reserve per-key daily slots before sending. Successful requests commit usage, failed requests release reservations, and counters stay within 20/20.
- Gemini key management: Test all configured keys sequentially, reveal or hide each key, and remove a key and its usage card immediately.
- Completion sound: Optional sound notification after a batch finishes, with a longer, more audible chime.
- Download the attached
autosortplus.xpifile. - Open Thunderbird → Menu (☰) → Add-ons and Themes.
- Gear icon → Install Add-on From File...
- Select the downloaded
.xpifile. - Click Add and restart Thunderbird if prompted.
- Batch progress panel: Sorting multiple emails now opens a dedicated panel with a live progress bar, per-email results (sorted / skipped / failed), running counts, and live per-key API usage cards. Includes Cancel and Retry Failed only controls, and re-syncs automatically if you reopen the tab.
- Parallel batch processing: Emails are analyzed concurrently in chunks (5 at a time by default, configurable from 1 to 20 in Settings → General Settings → Parallel batch size), so large batches finish much faster while per-key usage counting stays correct.
- Bulk import Gemini API keys: Paste many keys at once (one per line) into Settings → Multiple Gemini API Keys → Bulk Import. Duplicate keys are detected before import so you never add the same key twice.
- Toolbar popup: Click the AutoSort+ icon in the Thunderbird toolbar for an at-a-glance view of your API usage. It shows total usage across all Gemini keys (
N / total used), a color-coded progress bar, how many keys you have and how much quota is left, plus quick buttons to open Settings and the Batch Panel. - Dark mode: Full dark-theme support across the Settings page, the popup, and the Batch Panel. It follows your OS/Thunderbird color scheme automatically — no more bright-white panels in a dark UI.
- One email = one AI request: Retries no longer re-analyze the same email, so batches consume quota exactly as expected (previously a batch of emails could make far more or fewer calls than intended).
- More reliable email content extraction: HTML bodies are now properly converted to text (the conversion is awaited) and messages without MIME parts are handled, so fewer emails are skipped (previously some would move based on empty content).
rateLimit is undefinedfixed: Per-key Gemini usage is normalized with safe defaults, so the settings and popup no longer crash when a key has no usage record yet.- Usage race condition fixed: Concurrent requests could clobber the per-key daily counter; a lock now serializes reads/writes so every email is counted correctly.
- Download the attached
autosortplus.xpifile - Open Thunderbird → Menu (☰) → Add-ons and Themes
- Gear icon → Install Add-on From File...
- Select the downloaded
.xpifile - Click Add and restart Thunderbird when prompted
- Gemini 404 fixed: Google now limits the legacy
gemini-2.5-*models to accounts that have used them in the past. New API keys therefore returned404 ... no longer available to new users, even thoughmodels?key=still listed the model. The default Gemini model is nowgemini-3.5-flash-lite, which is available to new projects. - Choose your model: A new Gemini Model dropdown lets you pick any Gemini model, with a Load Available Models button that queries your API key and lists the models it can actually use (plus a "Custom model…" option). Other providers (OpenAI, Anthropic, Groq, Mistral) now accept an optional Model override as well.
- Gemini 3 support: Requests to Gemini 3.x now use
thinkingConfig.thinkingLevel; the legacy numericthinkingBudgetis only sent to 2.5.x models. Sending the wrong one returns a400. - Clearer errors: Model-not-found responses now include a hint to pick a different model in Settings.
- Refreshed test tooltips: The hover text on the API key "Test" result and the "Test API Connection" result now updates on every run instead of keeping a stale previous message.
- Reset counter fixed: The Reset Counter button now actually clears today's usage stats (it previously wrote to a legacy field that the app no longer reads).
- Download the attached
autosortplus-v1.2.5.xpifile - Open Thunderbird → Menu (☰) → Add-ons and Themes
- Gear icon → Install Add-on From File...
- Select the downloaded
.xpifile - Click Add and restart Thunderbird when prompted
- Addresses #11 (Gemini
404on new API keys; configurable model)
- Groq model updated: The previous model (
llama-3.3-70b-versatile) was shut down in August 2026 and returned amodel_not_founderror. Groq now uses the supportedopenai/gpt-oss-120bmodel. - Reliable email classification:
openai/gpt-oss-120bis a reasoning model, so its reasoning tokens counted against the old 50-token limit and produced empty responses. We now setreasoning_effort: 'low'and increased the token budget so every email returns a valid label. - Gemini note: The Gemini API works correctly when a proper API key is configured. If you see a 404 on "Test", double-check the key was copied completely and that the Generative Language API is enabled — see the setup guide.
- Fixed: Manual label application via right-click (AutoSort+ > AutoSort Label > pick any) now works reliably in all Thunderbird message list views.
- Root Cause: The extension previously relied on a content script (content.js) to fetch selected messages, but Thunderbird does not inject content scripts into privileged mail/message tabs. This caused the error: "Could not establish connection. Receiving end does not exist."
- Solution: The logic for fetching selected messages and applying labels is now handled entirely in the background script using the mailTabs API. This removes the dependency on content scripts for this feature and ensures robust operation.
- No other features are affected by this change; all AI, Ollama, and label management features continue to work as before.
- Ollama Model Download UI: Buttons are now always below the download input and never overlap.
- Ollama response parsing and label matching are more robust and tolerant (closest match if not exact).
- Improved error handling for model/tag mismatches and 404s.
- Clear error messages in tooltip and status
- Cannot save or test duplicate keys
- Better Key Management - Users notified if key is already added
Ollama Model Download UI: Buttons are now always below the download input and never overlap.
Ollama response parsing and label matching are more robust and tolerant (closest match if not exact).
Improved error handling for model/tag mismatches and 404s.
- Visual indicator with red border and background
- Test button blocked for duplicates
- Save prevented with helpful error message
- Toolbar Icon - Quick access button in top-right of Thunderbird
- Click the AutoSort+ icon to instantly open settings
- Hover for tooltip "AutoSort+ Settings"
- No need to navigate through menus anymore
- Faster settings access - one click from anywhere in Thunderbird
- Better addon visibility - always accessible in toolbar
- Tooltip provides clear button function
- Fixed single-key usage tracking - Now properly displays "Today's Usage:" when using only 1 Gemini API key
- Fixed rate limit notification - Persistent notification now shows when limit is reached on single or multiple keys
- Improved API test feedback - Shows specific error messages:
- 429 error: "⛔ Limit reached" (key has exhausted daily quota)
- 401/403 error: "✗ Invalid key" (key is invalid or expired)
- Other errors: "✗ Failed (status code)"
- Added tooltips - Hover over test results to see detailed error explanations
- Better error messaging for API key testing
- More intuitive status indicators with help text
- Cursor changes to "help" icon when hovering over test results
- Single-key configurations now display in new multi-key format
- Multiple Gemini API Key Support - Add multiple API keys from different Google Cloud projects
- Automatic Key Rotation - Seamlessly switches between keys when rate limits are reached
- Per-Key Usage Tracking - Monitor usage statistics for each API key individually
- Smart Key Management - Visual indicators show which key is active and available
Free Gemini tier provides 20 requests/day per project. With multiple keys:
- Add keys from different Google Cloud projects
- Extension automatically rotates to next available key
- Example: 5 keys = 100 requests/day total
- Each key tracks its own rate limit independently
- Enhanced UI for multi-key management
- Individual test buttons for each API key
- Real-time status indicators (Active, Ready, Near Limit, Limit Reached)
- Better error messages when all keys are exhausted
- Backward compatible with single-key configurations
- Fixed API key test function for Gemini keys
- Fixed test result display to show inline status per key
- Improved key validation feedback
Free API tiers are severely limited when processing emails! Email content is large text, which counts heavily against rate limits:
- Gemini: ~15-20 emails before hitting limits
- OpenAI: ~5-10 emails (very strict on free tier)
- Anthropic: ~10-15 emails
- Groq: ~20-30 emails (best free option)
- Mistral: ~10-15 emails
For daily email processing, paid API plans are strongly recommended.
Free tiers are suitable for:
- Occasional use (a few emails per day)
- Testing the addon
- Light personal email management
For regular use, consider:
- Upgrading to paid API tiers ($5-20/month)
- Processing emails in small batches with delays
- Using Groq for the highest free tier limits
- Multi-provider AI support (Gemini, OpenAI, Anthropic, Groq, Mistral)
- IMAP folder discovery - Automatically load folders from mail accounts
- Batch email processing - Select and sort multiple emails at once
- Move history tracking - Last 100 email moves recorded
- Smart label matching - Skips null categories, auto-creates custom folders
- Professional UI - Provider info cards, real-time validation
- Added support for 5 AI providers with easy switching
- Updated Groq to llama-3.3-70b model
- Improved error handling and validation
- Fixed batch email processing bugs
- Enhanced folder management
- Added rate limit guidance
- Download the autosortplus.xpi file
- Open Thunderbird
- Click the Menu button (☰) and select "Add-ons and Themes"
- Click the gear icon and select "Install Add-on From File..."
- Select the downloaded autosortplus.xpi file
- Click "Add" when prompted to install the add-on
- Restart Thunderbird when prompted
- AI-powered email sorting
- Custom categories/folders management
- Gemini API integration
- Automatic email organization
- Removed bulk actions
- Made AI sorting always enabled
- Improved UI with detailed AI description
- Updated terminology from labels to categories/folders
- Download the autosortplus.xpi file
- Open Thunderbird
- Click the Menu button (☰) and select "Add-ons and Themes"
- Click the gear icon and select "Install Add-on From File..."
- Select the downloaded autosortplus.xpi file
- Click "Add" when prompted to install the add-on
- Restart Thunderbird when prompted