Changelog
Track the latest updates and improvements to TraceMind.
v1.10.9
Coming soon- ChangedA large backup now downloads as one .zip file you save once, instead of one download per part (which could mean dozens of Save dialogs with Chrome's "Ask where to save each file" turned on). Importing is just as easy: pick the .zip and TraceMind reads every part out of it, with duplicates still skipped automatically. Older split backups (numbered .json files) still import by selecting them together.
- FixedIf an interrupted backup import cannot reach the background worker while resuming, the dialog now stays recoverable instead of leaving its controls disabled. You can try again or cancel the import.
- FixedThe export dialog no longer shows an "Include encryption metadata" checkbox that had no effect. Whether a backup is encrypted follows your passphrase setting in Settings > Privacy & Encryption; the dialog now states plainly whether the file will be encrypted or readable JSON.
- ChangedTraceMind's privacy and retention wording now clearly separates local core capture, indexing, search, screenshots, and analytics from optional provider-backed Chat; captured history from the separate Chrome-history import; no TraceMind page cap from browser and device storage limits; Free local storage from optional Pro passphrase encryption; and sandboxed reading snapshots from complete archives.
v1.10.8
Coming soon- FixedFixed a crash that could happen when searching a very large history from the full dashboard. On big histories the search could run out of memory and force a reload; searches are now bounded and no longer load every saved screenshot into memory at once, so they stay fast and stable.
- FixedThe "Try" search suggestions under the search box now actually reflect what you browse most. Because of an indexing bug they had been quietly falling back to the same generic examples for everyone.
- ChangedYou can now set how many results show per page right from the pagination bar on the full page (25 / 50 / 100 / 200), instead of digging into Settings for "Max results per search." Your choice is remembered.
- AddedLarge backups now export as several smaller files instead of one huge file, and you can select all of them together when importing. This fixes big backups (especially with screenshots) that were too large to restore in one go. Duplicates are still skipped automatically, so importing all the parts is safe.
- ImprovedScreenshots are now stored in their own place instead of alongside every page record, so browsing, filtering, and searching your history stay light and quick even with thousands of captured pages.
v1.10.7
Coming soon- FixedSaving a page for offline reading now works even on tabs that were already open when TraceMind updated. Before, an offline save could quietly do nothing on those tabs until you refreshed them; now TraceMind restores its local capture helper on the spot and completes the save.
- ChangedThe custom date and hour filter in the full dashboard is now a clear labeled control, so scoping a search to a specific day or time is easier to find, and the filter panel follows your light or dark theme instead of always appearing white.
v1.10.6
- ChangedA refreshed Memory design across the popup and full page: a serif display face, refined type and spacing, and a soft fog backdrop, with every feature right where you expect it.
- ChangedHistory retention can now be unlimited: keep your searchable history for as long as you want instead of the previous 365-day cap, and new installs default to unlimited. Free and Pro both support it; Pro can still choose a shorter window. Saved pages are never auto-removed.
- ChangedOptional multilingual semantic search is available to every user. Existing and new users start on the fast English model; choose Multilingual MiniLM L12 in Settings > Semantic Search Model and save to search across 50+ languages. The semantic index rebuilds locally in the background while keyword search stays available.
v1.10.5
- AddedNew users with little history get a one-click taste of semantic search: a single plain-language "Try" chip that finds their TraceMind page without ever typing its name. It appears only when that page is already in their history, so it never comes up empty.
v1.10.4
- AddedThe "Try" suggestions under the search box are now personalized to your own browsing, surfacing the topics that recur across the pages you visit most, instead of three fixed examples. Topics that appear across several different sites are preferred over any single site's name, and clicking one runs a normal search. New profiles keep the starter examples until there is enough history to personalize.
v1.10.3
- FixedMuch faster startup, especially with a large history. The extension used to re-compress every stored page on every launch, which could make it unresponsive for about ten seconds while opening. That cleanup now runs once and never blocks startup again; on a roughly 10,000-page library, initialization dropped from about 11 seconds to a third of a second.
- FixedThe extension now answers "are you ready yet?" from the moment it starts, so the popup and full page no longer sit on the initializing screen during a slow launch.
v1.10.2
- FixedExporting a large backup no longer fails. Big histories, especially with screenshots included, could exceed the browser's maximum text size and end with an error and no file. Backups are now streamed to disk in pieces, so they export regardless of size. Over-large encrypted backups stop with a clear suggestion instead of failing silently.
- FixedThe full page no longer opens blank and needs a refresh right after the extension updates or reloads, and the first search after opening shows results immediately instead of timing out while the on-device AI model warms up.
- FixedSearch result icons now come from the browser's own local favicon cache instead of being fetched from each website, so the results list makes zero network requests, works offline, and stops flooding the console with icon errors.
v1.10.1
- FixedThe import and reindex progress areas could show a stray "0" before any entries were counted. Nothing is shown now until there is real progress to display.
v1.10.0
- AddedThe full page now shares the popup's redesigned look: a warm "Memory" light theme and an indigo "Spotlight" dark theme, the serif wordmark, and an On-device badge, so the popup and the full dashboard feel like one calm, private space. Every existing dashboard feature (sidebar filters, statistics, storage, settings, chat, pagination, bulk select, notes, tags, pinning) stays exactly where it was.
- AddedSearch results on the full page now show whether each page matched by meaning (Semantic) or by exact keyword, with a similarity score and bar, so you can see at a glance how the AI understood your query.
- AddedA dismissible "what's new" note now appears in the popup and full view after you install or update the extension, with a small NEW badge on the toolbar icon and a link to this changelog. No account, email, or push notification required, and it never reappears once dismissed.
- ChangedThe full-page search box is now the popup's debounced pill input and shows the detected intent of your query. Clearing it returns you to your recent history instead of a "no results" screen.
- FixedThe full-page sidebar's Chat item no longer shows a Pro tag to users who already have Pro; it now matches how the other Pro features are marked.
v1.9.1
- FixedThe popup opens instantly again. The redesigned popup loaded its fonts from Google Fonts, and that network request could hold the popup on a blank loading screen for seconds. Fonts now ship inside the extension, so opening the popup makes zero network requests and no longer contacts Google.
- FixedLarge histories no longer stall the extension after browser startup. Startup checks used to load every stored AI vector into memory; they now compare only database keys, and the vector index loads the first time you actually search instead of when the popup opens.
v1.9.0
- AddedDate chips in the popup. Tap Today, Yesterday, or Last week to scope results to a date range without opening the full page, both while browsing and while searching. Tap the active chip again to go back to all results.
v1.8.0
- AddedBackup export and import are now free for everyone. Saving your history to a plain JSON file and restoring it are no longer Pro features, so you can move between machines and recover from a profile wipe. Your data is yours.
- ChangedPassword-protected (encrypted) backups are now the Pro option, using your own passphrase. Importing an encrypted backup is always free for everyone, so nobody is ever locked out of their own data, even after a subscription lapses.
v1.7.0
- AddedMultilingual semantic search. A new Semantic Search Model setting lets you switch the on-device model to one that understands 50+ languages, so an English search can find a page written in German, French, or Macedonian, and the other way around. Switching rebuilds your search index in the background; keyword search keeps working throughout.
- AddedAI chat now quotes the passage you actually asked about, centering each source snippet on the matching part of the page instead of its opening lines.
- AddedAI chat handles short follow-ups. A terse question like "cheaper?" now reuses your previous question when searching your history, so it retrieves relevant pages instead of coming up empty.
- ChangedEach stored embedding now records which model produced it, so switching models rebuilds only the vectors that need it and never mixes incompatible vectors in one search.
v1.6.3
- FixedThe popup and the full page now return the same results for the same search. A candidate-pool sizing quirk let the popup drop a strong meaning-based match that the full page kept; the pool is now a fixed size on every surface, so the popup's results are an exact prefix of the full page's.
- FixedThe full page no longer flashes your whole history over search results, and no longer shows a previous search's results after you run a new one. The latest search always wins.
- FixedBackground AI indexing now restarts on its own within a few minutes after a large import, even if the service worker dropped the rebuild, and the index-rebuild-speed note no longer says CPU when your GPU is actually in use.
- ChangedThe full page now opens as a single tab: Open full view, Chat, and Settings links focus the existing tab instead of stacking duplicates. The popup shows at most 50 results with a link to see the rest in the full view.
v1.6.2
- FixedThe Saved filter now shows in the active filters bar and can be cleared.
- FixedBlocking a domain from a result card no longer resets your other settings; settings updates now merge instead of overwriting the rest.
v1.6.1
- FixedLarge imports no longer stop partway. A shared status flag could make the background index rebuild look like a cancel to an in-progress import; import now tracks its own signal and the rebuild waits until import finishes. A 4,314-entry import now lands all 4,314.
- ImprovedBetter semantic ranking: a strong meaning-based match with no title keyword now scales its boost by how closely it matches, instead of being buried under weaker pages that merely repeat a query word.
- FixedVisited search-engine result pages no longer clutter semantic results, and existing installs now get the sharper default search-confidence threshold via a one-time migration that keeps any value you set yourself.
v1.6.0
- AddedChat is easier to find: a left-sidebar item in the full page and a one-click Chat button in the popup (Pro). External-AI Chat is now strictly opt-in, gated by a Settings toggle (off by default) and a consent screen that explains what is sent to your chosen provider.
- AddedAI index rebuild speed control. Background re-embedding is idle-aware (it pauses while you are actively using your computer) with a Full speed option, and its progress is durable across restarts, shown as a toolbar badge and in-UI chips.
- ChangedThe AI index now rebuilds on the GPU when your device has one (WebGPU) instead of always on a single CPU thread, and is much faster thanks to batched embedding.
- FixedFull-page sort now follows intent (Recent when browsing, Relevance when searching), the popup is fast and complete on a cold open again, and the footer shows the true number of indexed pages.
v1.5.1
- FixedCold start no longer leaves the popup stuck on the Initializing screen. The startup database maintenance pass now runs in the background, so your history loads quickly even after a long gap.
v1.5.0
- AddedChat with your history (Pro): ask questions in plain language and get answers drawn from the pages you have visited or saved, with clickable source citations. Bring your own OpenAI, Anthropic, or Google Gemini key; your question and the matched snippets are sent only to the provider you choose, and your key is stored locally and never exported.
- ChangedLicense revalidation now refreshes features within the hour, so newly granted features and updated expiry reach an active subscription without waiting for a re-activation.
v1.4.0
- AddedSave / Bookmark (Free): explicitly save any page and pull it up instantly from a new "Saved" filter in the full-page view. Save from the popup with "Save current page" or with the bookmark button on any history card.
- AddedSaved pages are protected from automatic retention cleanup, so your bookmarks never expire, even past the history-retention window.
- AddedSave works even on pages that weren't auto-captured (excluded domains, or when capture is off): saving the current tab captures the page on demand.
- ChangedDatabase schema upgraded to v8 with an additive savedAt index that auto-upgrades with no data migration.
v1.3.2
- SecurityNarrowed web-accessible resources to model and runtime assets only, so no website can embed the extension's own pages (removes a clickjacking surface).
- SecurityImported screenshots are now validated as data:image URLs and rendered via DOM APIs instead of HTML string injection, closing a markup-injection path from crafted import files.
- SecuritySetting an encryption passphrase now enforces a minimum PBKDF2 iteration floor, so a malformed or too-low value can't weaken key derivation.
- FixedThe offline fallback tokenizer could break on Windows checkouts from stray carriage returns; line endings are now pinned and the parser tolerates CRLF.
- ChangedBundled the static-embedding (Model2Vec) assets so fresh builds always ship the semantic search fallback.
- ChangedUnified the free excluded-domains limit to a single source of truth (10) across code and docs, and removed dead legacy source files.
v1.3.1
- FixedChrome history import no longer collapses into a single entry with an inflated visit count. Deduplication now merges only entries that share real content, so distinct URLs are always preserved.
- FixedDeduplication now also deletes the embedding rows of merged-away entries and any newly orphaned content rows, instead of leaving them behind as storage bloat.
- FixedAdded automatic, one-time recovery for profiles already corrupted by the old import behavior: on update the extension detects the collapsed-history signature and rebuilds accurate per-URL history from the browser's own records, showing a one-time banner when it does.
- ChangedRemoved the low-value maintenance sweep that deleted unpinned, older-than-7-days pages with fewer than 2 visits. Retention is now governed solely by the tier history-retention window.
v1.3.0
- ImprovedSearch-engine internals overhauled for performance: a packed int8 embeddings cache (about 4x less memory), metadata-only reads, indexed tag lookups, and a consolidated search-fallback ladder.
- ImprovedBetter degraded-mode search. When on-device ML is unavailable, the fallback now uses a locality-preserving static embedding (a bundled Model2Vec table distilled from MiniLM) instead of a near-random hash.
- ImprovedBackground re-embedding is now resumable across service-worker restarts and only processes entries that need it; imported browser history gets real embeddings automatically once the AI is ready.
- ChangedRemoved the dormant Voy vector index in favor of the already-live brute-force cosine path (real distances, CSP-safe).
- ChangedEmbedding storage now records model, version, and dimension metadata (database schema v7, additive auto-upgrade).
- SecurityEncryption at rest now uses 600,000 PBKDF2-SHA256 iterations for new passphrases (OWASP 2023 guidance); existing encrypted data continues to decrypt.
v1.2.0
- AddedOffline page viewer for Pro users: save and view complete HTML snapshots offline with one-click save, sandboxed rendering, auto-save option, and smart storage management.
- AddedChrome History Import (free feature): import your existing Chrome history for instant text search by title and URL, with full semantic search once you revisit pages.
- AddedEnhanced export/import encryption: AES-256-GCM encryption with PBKDF2 (200K iterations) for secure backups, background processing, and smart duplicate detection.
- ImprovedPBKDF2 key derivation iterations increased from 100,000 to 200,000 for stronger encryption security.
- ImprovedDatabase schema upgraded to v6 with dedicated offlinePages and settings tables for better organization.
- ImprovedScreenshot quality system now offers 4 distinct levels: Low (320x240, Free), Medium (640x480), High (1280x720), and Ultra HD (1920x1080) for Pro users.
- ImprovedStop words list optimized to 131 carefully selected words for more accurate search results.
- ImprovedMaintenance system upgraded to 5 specialized alarms: maintenance (6h), deduplication (24h), license revalidation (1h), storage optimization (24h), and indexing (5m).
- ImprovedVector similarity search runs entirely on-device with a CSP-safe cosine index for better compliance and performance.
- ImprovedOffline page viewer enhanced with improved HTML sanitization, configurable image inclusion, and smart storage management with pruning support (waiting for review).
- ChangedSearch execution optimized to sequential processing (FlexSearch first, then vector search) for better performance and lower memory usage.
- ChangedEncryption scope expanded: AES-256-GCM now applies to both at-rest storage AND export/import (previously export only).
- ChangedSPA navigation detection improved with pushState/replaceState/popstate event handling for better single-page app support.
- FixedExcluded domains limit clarified: free tier UI now allows up to 10 domains for better usability.
- SecurityEnhanced encryption strength with 200,000 PBKDF2-SHA256 iterations (up from 100,000) and expanded encryption to cover both storage and export.
v1.1.0
- AddedPro license system with offline activation, device limits, and feature gating.
- AddedStatistics dashboard with date range filters, quick filters, and domain blocking.
- AddedBackground processing improvements with a dedicated service worker and offscreen ML inference.
- AddedCustom encryption passphrase (AES-256-GCM with PBKDF2-SHA256, 200,000 iterations) for Pro users.
- AddedNotes filtering and improved tag handling in history views.
- AddedOffline page viewer & saver for Pro users: full HTML snapshots with sanitization, sandboxed iframe rendering, auto-save option.
- AddedExport/import encrypted backups for Pro users: JSON v2 format, background processing in chunks of 200, smart duplicate detection, streaming import with validation.
- ImprovedScreenshot capture with 4 quality levels: Low (320x240, Free), Medium (640x480), High (1280x720), and Ultra HD (1920x1080) for Pro.
- ImprovedSettings navigation and UI styling with design tokens.
- ImprovedStartup performance by deferring heavy storage optimization.
- ImprovedPro plan pricing details and license key formatting.
- ImprovedIcons and branding refresh.
- FixedHybrid search ID mismatch to improve semantic results.
v1.0.0
- AddedInitial release of TraceMind, an AI-powered browser history manager.
- AddedNatural language search across your entire browsing history.
- AddedPrivacy-first local processing; your data never leaves your device.
- AddedSmart categorization and tagging of visited pages.
- AddedChrome extension with seamless browser integration.
- AddedFlexible pricing plans with monthly and yearly subscriptions.
- AddedComprehensive FAQ and documentation.
- SecurityAES-256-GCM encryption at rest for all local data storage.