Skip to content

Changelog ​

2026-10-12 (0.17.4) ​

New ​

  • Reworked the desktop AI extension update experience: the settings page can check for updates and show compatibility with your Server and platform, with manual download and offline import; installed extensions at or above the minimum supported version keep working without re-downloading
  • The shared photo picker is rebuilt: search to locate photos, reused everywhere from people covers to album covers and avatars; on mobile an alphabet scrollbar enables fast jumping
  • People editing now supports picking multiple library photos as cover candidates

Improvements ​

  • Big face recognition performance gains: faster desktop face-matching queries and batch concurrency, plus an automatic CPU retry when an inference provider returns invalid results
  • Smarter background task scheduling: AI tasks yield to higher-priority work such as scans and thumbnails, switching once in-flight batches finish, so organizing photos is no longer blocked by AI analysis
  • The task queue gains scheduling indexes and photo links; deleting a photo cleans up its related tasks, and upgrading an older library backfills links and reclaims SQLite space
  • Metadata extraction reuses cached results and color analysis reads thumbnails instead of originals, making scans faster
  • Disk I/O is throttled per operation, so media processing like video metadata and motion photo extraction no longer holds the disk for long stretches

Bug Fixes ​

  • Fixed desktop annual report photos and avatars failing to load
  • Fixed video metadata bloating EXIF with cover images and other binary data, with a SQLite cleanup tool
  • Fixed the vector type import failing on fresh PostgreSQL deployments
  • Fixed photo picker entry points and interactions when changing people covers, album covers, and avatars

2026-10-11 (0.17.3) ​

New ​

  • Desktop "Data Directory" setting: move your entire photo library to another drive; migration waits for running tasks and uploads to finish, rolls back on failure, and keeps the original folder as a backup
  • Desktop "Connect Mobile App": shows a LAN address and QR code so your phone can back up photos to the same library; the port stays stable across restarts
  • Photo viewer gains "Reveal original file" in the file manager

Improvements ​

  • Desktop cold start is dramatically faster: about 18 seconds from launch to usable down to around 4 seconds
  • Much faster mobile backup and uploads: original-file save confirmation is about 88% quicker; thumbnails and metadata move to background processing; chunked uploads resume after interruption and failed files no longer block the rest
  • More reliable background task scheduling: AI tasks run in type-based batches with fewer model switches, and disk I/O is throttled per volume for a smoother UI during scans
  • HEIC/HEIF photos now display as previews directly in the browser
  • Upload queue survives minimizing the window or switching pages

Bug Fixes ​

  • Fixed the photo page flickering with a full refresh after background tasks finished
  • Fixed desktop AI chat streaming responses breaking (502) and a black console window flashing on Windows
  • Fixed desktop settings failing to save when AppData is encrypted
  • Fixed mobile photo viewer thumbnails being covered by the action bar
  • The Windows installer now closes a running TrailSnap automatically, so you no longer need to quit it before upgrading

2026-10-09 (0.17.2) ​

New ​

  • Ticket wallet redesign: train and flight tickets in one place, link tickets to albums and photos, and open a ticket to see its related memories

Improvements ​

  • Unified mobile date and time pickers covering date, time, month, and range modes
  • Smoother mobile albums and recycle bin with long-press multi-select; overlays now close automatically when navigating
  • Refined moments-style layout heights for highlighted photos
  • On This Day now supports swipe navigation on mobile with accidental taps suppressed
  • Unified dialog layering so new dialogs always stay above open panels

Bug Fixes ​

  • Fixed timeline sorting lag and incorrect photo dimensions caused by EXIF orientation
  • Fixed first-install registration being blocked by railway data initialization
  • Transient read requests now retry automatically, and duplicate connection errors are no longer shown repeatedly

2026-10-04 (0.17.1) ​

New ​

  • Mobile photo diaries now support single-page reading, automatic text pagination, date selection, and swipe navigation

Improvements ​

  • Unified liquid-glass toolbars and responsive action menus across photos, albums, and memories, with bottom sheets on small screens
  • Refined headings, cards, and spacing on home, album, and memory pages for mobile browsing
  • Reduced Server Docker image size: local measurements show about 61% / 59% smaller compressed layers on AMD64 / ARM64, while retaining common video reading, thumbnails, and daily-frame film generation

Bug Fixes ​

  • Fixed Android system-bar safe areas and keyboard layout spacing for photo covers and toolbars
  • Improved multi-architecture Server builds and synchronization to the China container registry

2026-10-03 (0.17.0) ​

New ​

  • Life chapters: organize photos and memories from a period into chapters, with discovery, manual editing, diary reading, and story browsing
  • Memory relationship graph: explore connections between people, places, and events to rediscover related experiences
  • Person timeline: review one person or shared photos of two people by year, with segment hiding, restoration, and AI interpretation
  • Daily frame: choose a photo or video clip for each day in a calendar, fill gaps, edit captions, and preview, export, and manage silent films with one second per day
  • Photo diary layout: browse photos by day and expand a day to see more images
  • Map route lines: connect photo locations to revisit travel routes
  • Background color reanalysis: reanalyze photo colors in batches to refresh color search data

Improvements ​

  • Configure local language model thread counts to suit your device
  • Improved person detail layouts and album titles that remain visible while scrolling
  • More consistent system settings and common action interfaces
  • Updated dependencies for more reliable installation and builds

Bug Fixes ​

  • Fixed virtual scrolling calculations in person and album galleries
  • Fixed coordinates not updating after map dragging and administrative regions not refreshing after saving locations
  • Fixed mixed-content requests behind HTTPS reverse proxies
  • Fixed several styling and AI concurrency issues

Upgrade Notes ​

  • Update Server, AI service, and clients together; Server applies database migrations on startup. Back up your database before upgrading
  • Daily frame film generation requires FFmpeg, FFprobe, H.264 encoding support, and a working Chinese font. Daily selections remain available when these components are missing, with an explanation shown in the interface

2026-09-24 (0.16.0) ​

New ​

  • Memory fragments: TrailSnap now discovers candidate memories from the people, places, and ticket clues in your photos and surfaces them on the home page. Confirm them to turn scattered photos into real events with a cover, a story timeline, and linked photos, people, places, and tickets. You can also create memories manually from albums, photo selections, or the AI assistant — and AI only writes a memory after you explicitly confirm it.
  • Location time compare: drag a split line to compare photos of the same place across years. Enter from a location's detail page or the "This place back then" action in the photo viewer. Includes a year track, photo swapping, and one-tap then-and-now artwork generation that hides exact coordinates by default when sharing.
  • Group albums in the people page: photos are aggregated by co-occurring people — "me and dad" or "the three of us" each become their own album with the latest group photo as the cover.
  • Custom people covers: pick a person's cover from their own photos; the edit dialog is now full-screen on mobile.
  • The batch time-fix toolbox tool now accepts an album as the processing scope, touching only that album's photos in place.

Improvements ​

  • Major visual refresh: desktop sidebar with rounded buttons and primary-gradient active states, a floating liquid-glass bottom bar with a directional bubble indicator on mobile, a unified design-token system, and navigation regrouped into Photos / Explore / More with collapsible, remembered groups.
  • Smart album overview now shows real content covers (face crops for people, representative photos for places and categories) and loads faster with new database indexes.
  • Mobile album actions: long-press a custom album for an edit/delete menu, and a new "more" menu on album detail pages (edit / change cover / delete).
  • Mobile location album filters moved to a bottom sheet, the map info drawer now snaps to three positions, and the 3D footprint globe renders at higher resolution on mobile.
  • AI assistant chat improvements: reworked model picker, loading animation during generation, and repeated regeneration disabled while busy.
  • Photo page data management hardening: state split between global and home views to stop cross-page pollution, scroll position preserved across refreshes, and timeline stats protected against request races.
  • Task management: SSE push on status changes, per-category status filters, and an explicit all-clear prompt; mobile backup notifications now deep-link to backup settings.
  • Rebuilt login and registration pages with a unified visual style.

Bug Fixes ​

  • Fixed the photo gallery showing an "empty album" placeholder on load errors.
  • Fixed duplicate/similar photo cleanup pages failing to scroll in some browsers.
  • Restored auto-focus on the name input when opening the person edit dialog.
  • Fixed overlapping elements interfering with video preview interactions.

Upgrade Notes ​

  • You can upgrade directly from 0.15.x; database migrations run automatically. Updating the Server and clients together is recommended.

2026-09-18 (0.15.0) ​

New ​

  • Added an immersive 3D footprint map to the location album: a 3D globe showing the cities you've visited and travel routes, with 2D/3D/globe views, year filters, city search, journey replay, and year tour. Layer controls include photo heatmap, province boundaries, and 3D terrain.
  • Requirement platform admins can now change requirement statuses from the web UI: suggested transitions are ranked by common flow, and a reason is required and recorded in the timeline.
  • Added an agent-native delivery workflow to the requirement platform: approved specs become the single source of truth for agents, with delivery tasks, run records, PR links, and clarification Q&A all visible. AI triage now streams its feedback.
  • The requirement platform can import cc-switch backups to track token usage: real consumption and cost in a unified accounting model, with trend charts, model and device breakdowns, and request details.

Improvements ​

  • AI requirement triage supports per-model reasoning effort settings.
  • Fixed footprint year boundaries so photos taken at the very end of a year are not rolled into the next one by millisecond rounding.

Bug Fixes ​

  • Fixed Tianditu map tiles breaking on the mobile app: SDK default layers no longer bypass the server proxy, photo markers render correctly again, and stale caches are invalidated automatically after upgrading.
  • Fixed connecting to a new server on the mobile app: scanning a QR code or typing a new address no longer fails with "Failed to fetch", expired credentials are cleared silently, and the "scan or auto-discover" dead loop is gone.

2026-09-11 (0.14.2) ​

New ​

  • The TrailSnap Requirements Platform is now live: submit feature requests and issues, track review outcomes and delivery plans, and explore trend and distribution charts on the dashboard. The in-app feedback page now points to the platform.
  • Added a developer blog to the official site, starting with an article on AI-assisted development reliability practices.

Improvements ​

  • Tianditu maps now use server-side keys: web, desktop, and mobile all load maps through your TrailSnap Server, so the real key never appears in map resource URLs. Keys are validated by the Server before saving.
  • The map settings page now explains how to migrate from old browser-type keys, with documentation updated accordingly.

Bug Fixes ​

  • Fixed maps going blank on the mobile app: the map proxy forwards the device's real User-Agent so Tianditu no longer returns 403, and failed responses are no longer cached by the WebView for 24 hours.
  • Fixed base map loading failures in the location trajectory view and the add-scene dialog in some environments.

Upgrade Notes ​

  • Users with browser-type Tianditu keys from earlier versions need to create a server-side key in the Tianditu console and replace it in Settings; if you use an IP allowlist, enter your TrailSnap Server's public egress IP.

2026-09-08 (0.14.1) ​

New ​

  • LLM settings now let you configure a display name, context window, and reasoning level for each model. Photo analysis and chat can use different models.
  • The Server now prunes old app installer caches and incomplete download files, and in-app updates retry automatically after weak-network interruptions.

Improvements ​

  • Mobile backup now computes checksums in parallel and adapts media and chunk concurrency to LAN conditions, reducing small-file requests and substantially improving throughput.
  • The mobile album list has a refreshed layout with clearer, better-organized information.
  • Refined navigation visuals and photo browsing; deleting a photo now keeps your current browsing position.

Bug Fixes ​

  • Fixed mobile backup issues involving repeated incremental scans, JPG+MP4 Live Photo pairing and upload, disk verification, type correction, and preservation of capture timestamps from the phone.
  • Fixed LLM connection editing and saving. Candidate models are fetched in the connection editor, while chats now show only saved models.
  • Fixed mobile QR-code server switching being blocked by the network policy.
  • Improved Live Photo loading to reduce waiting and flicker.

2026-09-07 (0.14.0) ​

New ​

  • The Album Agent is now available: search photos with natural language, discover travel clues, create person timelines and memory-detective stories, and organize albums from real photo evidence.
  • Added an Agent task center for proposed, running, failed, and completed plans, with confirmation, retry, cancel, continue, and undo controls.
  • Added the Album Doctor repair loop for album counts, covers, descriptions, file hashes, photo times, and locations. Every repair can be previewed, confirmed, and undone.
  • Added safe cleanup proposals: exact duplicate photos move to the recycle bin, empty-album deletion requires typed confirmation, and all operations remain auditable and reversible.
  • AI travel journals can now generate personalized HTML pages for preview, export, sharing, long-image capture, and browser-based PDF printing.
  • Added a read-only MCP service, a Pi Agent extension, and scoped access tokens so external AI tools can query TrailSnap safely.
  • The location album now includes an immersive footprint map and layered GPS travel trajectories.

Improvements ​

  • Android networking now routes through your self-hosted TrailSnap Server, with server-proxied map tiles for a more consistent experience.
  • Thumbnail reads no longer require a database session, and startup storage-cache prewarming improves browsing on large libraries.
  • Proactive memories now offer recommendation cards that open directly in an Agent conversation.
  • The default service port is now 3180, with installation scripts and Docker documentation updated.

Bug Fixes ​

  • Fixed Android map-tile direct-connection failures and refined icon centering and monochrome details.
  • Fixed a race where a cancelled task prefetched into the worker queue could still execute; cancellation now takes effect immediately.
  • Fixed several Agent streaming, artifact, and background-task edge cases to improve long-running task stability.

2026-09-04 (0.13.1) ​

New ​

  • The recycle bin now supports select-all across pages, emptying the entire bin, and restoring everything without loading every page or sending every photo ID.
  • Large purge operations can run in the background with live progress, avoiding long-running request timeouts.

Improvements ​

  • Reworked bulk deletion and added database indexes. In testing, clearing 10,000 photos dropped from about 22 seconds to about 0.9 seconds while blocking other operations less.
  • Face clustering now runs as a separate background task instead of blocking recognition; large face libraries are streamed to substantially reduce peak memory use.
  • Automatic face clustering skips redundant work when little has changed recently, while manually requested clustering still runs immediately.
  • Unified scrollbar styling across the app and improved scrolling containers on desktop and mobile layouts.
  • Improved AI-facing documentation hints on the official site so assistants can find and verify TrailSnap documentation more reliably.

Bug Fixes ​

  • Fixed the recycle-bin list failing when an item has no deletion timestamp.
  • Fixed incorrect cluster counts when multiple photos from the same cluster are permanently deleted together.
  • Fixed face similarity matching potentially scanning across users and ensured each user's configured clustering threshold is respected.

2026-09-03 (0.13.0) ​

New ​

  • The Android App can now check for updates automatically, download the APK in-app, and launch the installer, with progress reporting, retry support, and the option to skip a version.
  • Added video metadata parsing for common formats such as MP4 and MOV, so capture time, location, and device information can place videos correctly on the timeline and map.
  • The AI assistant now supports semantic memory retrieval, prioritizing photo memories most relevant to the current question in long conversations.
  • Added automatic context compression for the AI assistant, preserving a concise summary plus recent messages to reduce context overflow.
  • The AI assistant input now supports multiline text, with Enter to send, Shift+Enter for a new line, and correct Chinese IME behavior.

Improvements ​

  • The mobile photo viewer now uses a three-track swipe layout with adjacent-image preloading, making rapid swiping smoother and extending left/right navigation to search results, the recycle bin, folders, and more.
  • The mobile viewer layout now follows familiar system-gallery patterns, with clearer capture information and more focused actions.
  • Redesigned mobile backup and settings navigation and improved multi-select and viewing experiences.
  • Vectorized face clustering to speed up organizing large photo libraries.
  • Added AI-friendly documentation and QR-code entry points to the official site and refreshed demo images.

Bug Fixes ​

  • Fixed mobile backup issues involving Live Photo pairing, repeated failed uploads, adaptive-gallery backup speed, and worker database connection isolation.
  • Fixed mobile photo viewer track offset and the Android adaptive launcher icon safe-area ratio.
  • Fixed local API-prefix handshake failures on desktop.

2026-09-01 (0.12.1) ​

Bug Fixes ​

  • Fixed a Docker Server restart loop when upgrading from 0.11.0 because the database migration queried the not-yet-created backup_key column.
  • Instances where the upgrade already failed can update directly to 0.12.1 and safely retry the migration; instances that have not upgraded can also update directly without manual database changes.

2026-08-31 (0.12.0) ​

New ​

  • Added incremental system-gallery backup to the Android App, with destination selection, progress reporting, pause/resume controls, original media preservation, and paired Live Photo support.
  • Added desktop management for external libraries and configurable upload destinations for importing photos into a chosen external folder.
  • Added a side-by-side comparison view to the similar-photo tool for easier review and cleanup.
  • Swipe-to-declutter now saves decisions and progress, and can synchronize offline actions after reconnecting.
  • The desktop installer can now switch AI runtimes to better match the device.

Improvements ​

  • Unified connection entry points across desktop, mobile, and browser clients, with clearer mobile server setup and improved Docker LAN service discovery.
  • Improved desktop startup, session handling, and automatic updates to reduce repeated configuration and interrupted upgrades.
  • AI task concurrency now adapts to runtime pressure and retries failures; waiting tasks resume after model downloads complete.
  • Simplified AI performance choices and improved responsive layouts, dark mode, and keyboard accessibility throughout the Web interface.
  • Restructured multi-user storage for stronger file isolation and added support for longer photo paths.

Bug Fixes ​

  • Fixed mobile backup converting original media, losing paired Live Photo files, or mishandling optional upload fields.
  • Fixed malformed vision-model output, tasks remaining paused after model downloads, and visual-analysis request timeouts.
  • Fixed cleanup failures for locked user directories on Windows, plus several mobile navigation and timeline interaction issues.

2026-08-22 (0.11.0) ​

New ​

  • Added photo-anchored memory to the AI assistant, allowing meaningful moments from conversations about specific photos to inform more coherent future responses.
  • Added proactive memories: when On This Day photos are available, the assistant can prepare a personalized greeting and show an unread indicator in the sidebar.
  • Added folder filters to conditional albums, with multi-select support and automatic inclusion of all descendant folders.
  • Added multiple PP-OCRv6 tiers and the Buffalo M face model so users can choose between speed and accuracy for their hardware.

Improvements ​

  • Unified AI model browsing, downloading, switching, and removal; switching is faster and failed removals can fall back to an available configuration.
  • Improved AI assistant photo search with concise samples and aggregate summaries for exploratory queries, plus automatic grid layouts for multi-photo results.
  • Refreshed Web, PWA, Android, and iOS icons for a more consistent product identity.

Bug Fixes ​

  • Fixed the assistant sidebar indicator affecting accessible names and unauthenticated proactive requests disrupting login state.
  • Fixed model preparation and URL configuration issues in desktop builds, improving desktop AI extension installation and startup reliability.

2026-08-18 (0.10.2) ​

New ​

  • Added Reprocess to the photo viewer, allowing face recognition, OCR, scene classification, visual description, and image embedding to be rerun for a single photo with live progress and results.

Improvements ​

  • Interactive photo-processing tasks now receive priority and can continue while background classification is paused.
  • Improved download-button fallback so users can still reach the matching release page when the GitHub API is temporarily unavailable.
  • Synchronized Android and iOS version metadata for clearer upgrades and repeat installations.

Bug Fixes ​

  • Fixed Android WebView security rules blocking photos, album covers, avatars, and On This Day images when the App connects to an HTTP self-hosted server.
  • Fixed database connection exhaustion during large first-time scans, which could break thumbnail generation and cascade into downstream task failures.

2026-08-16 (0.10.1) ​

New ​

  • Completed the desktop AI extension's full service capabilities and added an AI model management page for on-demand model viewing and management; stabilized AI runtime startup and bundled face-recognition dependencies.
  • External links on desktop now open in the system browser; mailto and tel links are handed to system apps.

Improvements ​

  • Switched the Windows MSI installer language to Simplified Chinese.
  • Optimized app icon asset sizes for a smaller installer.
  • Improved the official-site download and deployment docs, added a one-click Docker install guide, and fixed download-page platform-asset matching.

Bug Fixes ​

  • Fixed Tianditu tile loading on desktop so the album location map and trajectory view render correctly.
  • Fixed map boundary and location data 404 after desktop packaging; map endpoints no longer fail when the working directory differs.

2026-08-15 (0.10.0) ​

New ​

  • Added a Tauri 2 cross-platform desktop app with an embedded SQLite database for standalone use on Windows, macOS, and Linux.
  • Added optional desktop AI extensions with on-demand runtime installation and model downloads.
  • Added server switching and recent-server history to the mobile App login screen.
  • Added desktop and server-side folder pickers for selecting external photo-library directories.
  • Added pinch-to-zoom density controls for mobile photo grids while keeping the current photo anchored.
  • Added multi-key Tianditu management and validation, plus direct AI-service connectivity checks in Settings.

Improvements ​

  • Improved incremental photo loading, cross-page state synchronization, album details, and the mobile timeline navigator.
  • Improved photo viewer, dialog, and Android back-button behavior so the active overlay closes first.
  • Simplified desktop and mobile navigation and regrouped travel-related entries.
  • AI assistant messages now use the user's custom avatar, and the mobile model selector uses a more compact layout.
  • The one-click installer can now back up existing data and either preserve configuration during upgrades or perform a clean reinstall.
  • Improved China-registry compatibility, multi-architecture frontend image builds, and release-asset naming.

Bug Fixes ​

  • Fixed desktop SQLite date statistics, vector search, startup, and AI-extension runtime issues.
  • Fixed removing photos from regular albums, permanently deleting recycle-bin items, and flight-ticket update feedback.
  • Fixed single-photo-directory mounts and Android debug-signature conflicts during repeat installations.

2026-08-08 (0.9.2) ​

New ​

  • Added rescan support for reviewed identities, making it easier to refresh person recognition after face-data updates.
  • Added download progress for offline map data.
  • Persisted offline reverse-geocoding data to avoid repeated downloads and lookups.
  • Demo mode can now prefill the login credentials automatically.

Improvements ​

  • Added an Alibaba Cloud registry option for deployments in China.
  • Improved initial frontend loading and Nginx delivery.
  • Added pre-deployment and data-safety guidance, and improved the home page and FAQ.
  • Expanded automated coverage for photo, settings, media, face, and album modules.

Bug Fixes ​

  • Fixed llama-server startup failures caused by a missing libgomp1 runtime dependency in the AI service image.

2026-08-02 (0.9.1) ​

New ​

  • Added Capacitor-based Android/iOS App projects. On first launch, the App lets users enter and save their self-hosted TrailSnap Server address.
  • Added automated Android APK builds. Pushing a v*.*.* tag uploads the APK and CLI artifacts to the same GitHub Release.
  • Added PWA support to the Web frontend, including installability and improved offline / refresh behavior.
  • AI Moments copy generation can now show its reasoning progress in real time.

Improvements ​

  • Improved Android back-button and gesture behavior: close overlays first, navigate back next, then move to the background from the home screen.
  • Unified App and CLI publishing rules: manual builds keep Actions artifacts, while version tags create or update Releases.
  • Added mobile App installation, server configuration, update, and privacy guidance to the official site.

Bug Fixes ​

  • Fixed executable permissions for the Android Gradle Wrapper on Linux GitHub runners.
  • Fixed PWA refresh behavior and branch image-tag handling.

2026-08-02 (0.9.0) ​

New ​

  • Added Moments Diary: generate, edit, and clear AI social copy by day, with main locations and photo highlights.
  • Added location album collage view with nationwide / province browsing, automatic or manual photo selection, tile replacement, and similar-photo removal.
  • Added map double-click drill-down and collage selection strategies including memory value, quality, latest, and random.
  • Added mobile bottom-tab navigation, map/collage gestures, and an Agent drawer.

Improvements ​

  • Reduced Server and AI image size and startup overhead; added graceful stopping and health monitoring to the AI service.
  • Improved social-image aspect ratios, mobile layout, and collage rendering performance.

Bug Fixes ​

  • Fixed timezone aggregation, streaming generation, and single-image aspect ratio issues for Moments copy.
  • Fixed mobile collage blank screens, reverse-geocoding parsing, and several external-library and task-navigation issues.

2026-07-28 (0.8.1) ​

New ​

  • Added an external photo library setup wizard for directory selection, path mapping, and scan configuration.
  • Added a global notification center for background tasks and system messages.
  • Added scheduled version-update checks.

Improvements ​

  • Improved AI assistant context trimming and photo retrieval for long conversations and complex libraries.
  • Expanded automated coverage for the frontend, server, and AI service.

Bug Fixes ​

  • Fixed reinstall, health-check, metadata dispatch, and AI-timeout issues affecting some deployments.

2026-07-11 (0.7.0) ​

New ​

  • Added "Letting Go" photo filter: swipe left/right to quickly decide each photo's fate — an easy way to declutter a bloated library.
  • Added Storage Center: a unified view of storage usage across sources; the home storage card jumps straight to the corresponding photo filter.
  • Photo editing now preserves original EXIF metadata (capture time, location, etc.), so timelines and map pins stay accurate.
  • The AI assistant floating button now supports free dragging and auto half-hiding at screen edges.
  • Clicking the search box now pre-warms the AI service to avoid first-search cold-start delay.
  • First-time users are redirected straight to the registration page.
  • The install script now auto-configures a China Docker mirror and enables it by region.

Improvements ​

  • The recycle bin now supports a selection mode for batch restore/delete.
  • Improved mobile experience for the Settings and Toolbox pages.

Bug Fixes ​

  • Fixed the photo location-edit popup failing to reopen after being closed.
  • Fixed the worker watchdog restarting in a loop when only paused tasks remained.

2026-06-07 (0.4.1) ​

New ​

  • Added Photo editing: crop, rotate, doodle, text, and shape tools.
  • Added Sidebar navigation: redesigned frontend layout with collapsible sidebar.
  • Sidebar shows hover tooltips when collapsed.
  • Added Sidebar global search: quickly find photos, albums, people, etc.
  • Added Theme customization: choose from 5 brand colors (Sky, Emerald, Violet, Rose, Amber) that apply globally with one click.
  • Added Custom sidebar navigation items: customize the sidebar navigation menu.
  • Unified theme handling with reusable components.

Improvements ​

  • Optimized location map page and view component; refactored backend photo query logic.
  • Improved multi-module query performance.
  • Unified icons and refined frontend conventions.
  • Improved page UI layout and fixed import issues.

Bug Fixes ​

  • Fixed album page theme color not responding and dark mode contrast issues.
  • Fixed missing primary theme utility class mappings (hover/shadow not following theme).

2026-05-28 (0.4.0) ​

New ​

  • Photo organize toolbox now supports "Time directory structure" option — flat or recursive nested structure (e.g. 2026/01/01).
  • Photo organize supports automatic grouping by person.
  • Added Built-in AI connection: works out of the box, using MiniCPM-V-4_6-Q4_K_M multimodal model by default.
  • Added Local LLM hosting: host custom models via llama.cpp.

Improvements ​

  • Migrated all AI services to ONNX Runtime, removing redundant dependencies.
  • Refactored AI service LLM deployment logic — replaced pyllama-cpp-python with natively compiled llama-server binary.
  • Implemented priority-based task filtering when queue is full.
  • Improved image processing with WebP format auto-conversion.
  • Improved LLM settings page with built-in connection badge.
  • Improved failed task cleanup mechanism.
  • Fixed organize task progress calculation error.

2026-05-14 (0.3.9) ​

New ​

  • Added "Skiing" category to smart classification.
  • Smart classification now supports setting cover photos.
  • Added ability to delete recognized people.

Improvements ​

  • Thumbnails now use WebP format to reduce file size.
  • Fixed task priority anomalies.
  • Improved smart classification accuracy, reducing misclassification.
  • Improved some UI details.

2026-04-28 (0.3.8) ​

New ​

  • Added Recycle Bin: restore deleted photos.
  • Added TrailSnap CLI: interact with TrailSnap via command line.
  • Added Skills integration: connect to OpenClaw, Claude Code, etc.

Improvements ​

  • Improved some UI details.
  • Improved AI chat experience.

2026-04-22 (0.3.6) ​

Improvements ​

  • Fixed tasks occasionally getting stuck.
  • Optimized task process management to reduce memory usage.
  • Speeded up LLM task processing to reduce waiting time.
  • Improved AI chat thinking/tool visualization so users can see intermediate steps.
  • Improved smart classification accuracy.

2026-04-18 (0.3.3) vibe coding prompt ​

New ​

  • Added manual photo classification so users can correct misclassified photos.
  • Added multiple LLM connections so users can choose which model to use.

Improvements ​

  • Improved photo classification and ticket recognition accuracy.
  • Improved task processing efficiency to reduce waiting time.
  • Fixed photo delete failing in some scenarios.
  • Supported legacy Apple Live Photo formats (.jpg + .mov).
  • Improved some UI details.

2026-03-29 (0.3.2) vibe coding prompt ​

New ​

  • Added AI Assistant: chat with LLMs to search and understand your album content.
  • Added Footprint Timeline in location albums: browse photos by time across locations.
  • Added Trajectory View: connect photo locations across a time range to show travel routes.
  • Added Token Management: create tokens for third-party apps to access album data without logging in.

Improvements ​

  • Location album and ticket stats pages support custom time ranges.
  • Fixed metadata parsing failures for Apple HEIC photos.
  • Improved search: fixed scenic spot search issues and improved suggestion loading speed.
  • Improved some UI details.

2026-03-21 (0.3.1) vibe coding prompt ​

New ​

  • Added Similar Photos Cleanup.
  • Added Photo Calendar on the home page.
  • Video playback supports speed control.

Improvements ​

  • Fixed leftover Live Photo/thumbnail files after deletion.
  • Fixed black screen when switching videos.
  • Improved people album loading speed.
  • Fixed permission issues in ticket recognition.
  • Improved some UI details.

2026-03-12 (0.3.0) vibe coding prompt ​

New ​

  • Added multi-user support with data isolation.
  • Added LLM photo analysis: generate descriptions and scores.
  • Added On This Day: browse photos from past years, sorted by score.
  • Added Album Cleanup: clean analyzed photos based on scores.

Improvements ​

  • Support hiding a people album.
  • Support adding photos to a people album.
  • Improved face recognition accuracy.
  • Improved some UI details.

2026-02-25 (0.2.3) vibe coding prompt ​

New ​

  • Added image file filtering.
  • Location albums support filtering by year.

Improvements ​

  • Fixed photo ordering issues in people albums.
  • Improved page layout and UI.

2026-02-05 (0.2.2) vibe coding prompt ​

New ​

  • Support exporting paper-style train tickets (blue/red).

Improvements ​

  • Added Tianditu tile caching to reduce network requests.
  • Improved location map rendering.
  • Fixed resource release issues in metadata rebuild tasks.
  • Improved UI display in some pages.

2026-02-01 (0.2.0) vibe coding prompt ​

New ​

  • Supported iPhone Live Photos.
  • Supported downloading offline maps for multiple countries, or uploading custom map data (requires rerunning Metadata Extraction task).
  • Added 358 5A scenic spots in China to the map album (requires rerunning Metadata Extraction task).
  • Supported editing custom scenic spot locations.
  • Supported searching by text, location, people, album, folder, filename, etc.
  • Supported recognizing flight tickets (order screenshots), requires rerunning Ticket Recognition task.
  • Supported ticket import/export (CSV/JSON).

Improvements ​

  • Fixed duplicate photo display.
  • Fixed missing default covers after rerunning face recognition tasks.
  • Improved ticket recognition accuracy and speed.
  • Fixed upload failure for images larger than 1MB.
  • Improved some UI details.

2026-01-17 (0.1.0) ​

FeatureStatusDescription
Upload & view photos/videos✓Upload local photos and videos, view and play them.
Add external folders✓Add external folders as data sources; TrailSnap will scan and index photos/videos automatically.
Live Photo✓Support vivo/oppo/xiaomi and more.
Timeline✓Smooth timeline scrolling experience.
View photos on map✓View all uploaded photos on a map; filter by province/city/district; list & map views supported.
Lit-up cities✓See cities appearing in photos; click a city to browse its photos.
Visited scenic spotsIn DevCount visited 5A scenic spots; support custom scenic spot areas; auto filter photos within the area.
Face recognition✓Recognize people in photos and add people tags.
Smart scene classification✓Auto classify scenes, e.g. night view, pets, food, selfie.
Smart search✓Search by people, content, time, etc.
Tags✓Add/remove tags manually, or auto-tag based on AI results.
Smart albums✓Auto generate albums from content, e.g. “Selfies by the sea with my girlfriend”.
Ticket recognitionIn DevRecognize train tickets and itineraries, auto extract travel info.
Annual reportIn DevAuto generate yearly travel report: photo wall, cities, scenic spots, travel timeline, mileage, etc.
Travel logPlannedGenerate travel logs from manual input or AI-recognized itinerary.

Released under the AGPL-3.0 License