Changelog
A chronological history of every SaysMee release.
Last updated: August 2026
v0.12.0August 14, 2026Latest
Grounded Workspaces
A reliability and collaboration update with source-grounded current answers, clearer research evidence, reusable workspace templates, dependable invitations, and more polished responsive feedback.
Research & Sources
- Current and version-sensitive research now ranks official evidence ahead of familiar legacy pages, preventing older training knowledge from overruling verified results.
- Incomplete research requests now ask for the missing subject before searching, so unrelated web results never appear as evidence.
- Verified NCERT book-name lookups use the current official catalogue result directly, while broader teaching and contents requests remain fully researched.
- The Sources panel now shows reviewed-source counts, official-source indicators, domains, dates, useful excerpts, and direct verification links.
- Web research now shows a calm “Searching on web” response status instead of exposing individual search queries or visited pages.
Workspaces & Templates
- The main Knowledge navigation is now named Library and opens the working Library experience; existing /knowledge bookmarks redirect there instead of returning a missing-page error.
- Added a complete Templates experience for browsing, previewing, using, copying, sharing, remixing, and publishing focused community workflows.
- Workspace invitation links now use the canonical saysmee.com origin, keep their destination through authentication, resume joining automatically after sign-in, and accept membership reliably through an idempotent database transaction.
Responsive Chat
- Fixed the tablet sidebar stacking above the chat header and removed duplicated safe-area spacing on iPad layouts.
- Response status text now uses a quiet animated glow while SaysMee is working, with reduced-motion support for accessibility.
- Chat now starts visibly sooner by streaming ordinary answer text immediately, preparing independent context in parallel, and saving the user turn alongside model preparation instead of before it.
- Fixed iPad and mobile message-action menus, zoom overflow, attachment previews while editing, and always-visible touch controls for removing composer images.
- Conversation menus now stay above the sidebar and open beside the selected conversation on every device, while sent image attachments use a compact thumbnail strip above a separate message surface.
- The composer plus menu now renders above clipped containers and remains visible and touch-friendly on iPad, mobile, and desktop layouts.
- If a provider stream stops after visible text, SaysMee now keeps the answer as a resumable checkpoint instead of showing a misleading Model Error.
- Guest chats now show a compact login and sign-up reminder in the main workspace; dismissing it keeps it hidden for 24 hours across sessions.
- The sign-in panel is now centered consistently across desktop, tablet, and mobile layouts.
- Signing out now returns directly to guest mode so users can keep using SaysMee without being sent back to the landing page.
- The landing page is now a first-visit introduction: after someone opens SaysMee, future visits return directly to guest chat even after signing out or closing the browser.
- Applied the approved SaysMee logo artwork exactly and centered it consistently across the website, favicon, PWA, and Apple home-screen icons.
- Kept the rounded SaysMee artwork for browser and installed-app icons while restoring the exact backgroundless mark throughout the in-product interface.
- Normalized the chat header across browser, installed-app, tablet, and mobile views with safe-area spacing, reliable title truncation, aligned actions, and a clean window title.
- Made Library responsive across phone, tablet, laptop, and wide desktop layouts with off-canvas tablet navigation, stacked controls, adaptive card columns, and touch-friendly actions.
Twenty4 Documentation
- The Twenty4 documentation site now uses the official Twenty4 mark in visible branding and valid browser, Apple, Windows shortcut, and installable-app icon metadata.
v0.11.0August 10, 2026
Meet SaysMee
The AI companion is now SaysMee, with a new visual identity, guided onboarding, privacy-conscious launch analytics, clearer account controls, faster files, durable long chats, and production-ready visual campaign emails.
SaysMee Identity
- Rebranded the product from Aven to SaysMee across the interface, metadata, support surfaces, legal copy, and user-facing AI instructions.
- Applied the approved SaysMee marks to the interface and install icons, with exact Twenty4 artwork on the About surface.
- Moved support to support@saysmee.com and set the canonical app origin to saysmee.com.
Account & Settings
- The sidebar profile now opens a complete account menu with Profile, Personalization, Settings, Help, About, and sign-out actions.
- The compact control beside the profile now opens Quick Settings; full Settings remains a separate, explicit destination.
Faster Files and Long Conversations
- Image preparation and secure uploads begin immediately and run concurrently instead of waiting for Send.
- Vision requests use short-lived private URLs, bounded visual batches, adaptive provider budgets, and an extra capacity retry to reduce limit failures.
- Saved conversations restore compact foundations and earlier context from storage while keeping recent turns verbatim, so one thread can keep growing without resending its entire transcript.
- Identical uploaded content is fingerprinted and collapsed into one Library asset while conversation-specific attachments remain reliable.
Onboarding, Privacy & Analytics
- Added resumable onboarding with response preferences, optional context, first useful actions, and user-reviewed memory transfer.
- Added cookie-free page measurement, consent-gated content-free product events, sanitized campaign attribution, and complete cookie controls in Settings.
- Added a protected analytics view that uses live Vercel data when configured and never fabricates unavailable metrics.
Documentation & Email
- Created a separate responsive SaysMee documentation site for twenty4.online with guides, search, trust resources, limits, support, and releases.
- Launch and product-update campaigns now use a visual SaysMee and Twenty4 email, saysmee.com links, 50-user batching, per-user dedupe, delivery logs, preference checks, and signed unsubscribe controls.
v0.10.0August 3, 2026Latest
SaysMee, Wherever You Are
SaysMee now adapts to your language, region, time zone, and device, with a smoother sign-in journey, faster feedback, and dependable app icons across Apple and PWA installs.
Language, Region & Time
- Language preferences now update SaysMee's core interface as well as AI responses.
- Region preferences control locale-aware date and number formatting throughout key app surfaces.
- Time-zone settings now give the model an exact server-authoritative local date, time, and UTC offset for current-time and relative-date answers.
A Better Start on Every Device
- The landing and sign-in experiences now share one cohesive SaysMee surface instead of sending users through a disconnected page.
- On phones and smaller tablets, the new-chat composer stays within easy reach at the bottom of the screen.
- Prompt suggestion cards are hidden on smaller screens to keep the first chat focused and uncluttered.
- Touch targets and safe-area spacing now adapt to compact and notched devices.
Settings & Feedback
- Quick feedback is now available inside full Settings with type selection, validation, delivery status, and a character counter.
- Feedback submissions now show a prominent confirmation toast, a saved-state panel, and a database receipt for verifiable delivery.
- Language, region, and time-zone controls apply immediately without requiring a reload.
Apple & Installable App Branding
- Added a standards-compliant opaque 180x180 Apple touch icon so SaysMee appears correctly on iPhone and iPad home screens.
- The web app icon now has genuine smooth, transparent rounded corners while preserving the original SaysMee wave artwork.
- Added a separate safe-area maskable icon for Android and PWA launchers.
Current Information
- Questions about current news, public officials, prices, schedules, and other changing facts now trigger automatic web research even when Research mode is off.
- Versioned AI-model questions and comparisons now verify model names and capabilities on the web instead of relying on potentially stale training data.
Reliable Image Messages
- Uploaded images now remain visibly attached to the correct message after sending and after reopening a conversation.
- Dragged and mobile-picked images are recognized even when the browser omits their MIME type.
- SaysMee keeps durable image previews separate from request-only vision data, preventing old images from bloating later chat requests.
Cleaner Code Reading
- Code keeps its useful syntax colors while selected text now uses a subtle neutral highlight instead of a distracting blue overlay.
Release Discipline
- Production builds now verify that user-facing code changes include release notes before deployment.
- App icons can be regenerated reproducibly from the approved SaysMee source artwork.
v0.9.0August 2026
Intelligent Library & Document Understanding
One of SaysMee's biggest upgrades yet, transforming file management and AI document understanding into a seamless experience with a brand new Library, intelligent document comprehension, and smarter AI context across conversations.
📚 Brand New Library
- Dedicated Library hub for all uploaded assets — images, documents, and files
- Organized sections: Images, Documents, Files — automatically categorized
- Favorites section for quick access to frequently used assets
- Recently Opened section to pick up where you left off
- Search, filter, and sort across all assets
- Faster asset retrieval with virtualized grid for 10K+ assets
- Responsive design — mobile sidebar overlay with hamburger menu
🖼 Media Viewer
- Fullscreen media viewer with image thumbnails
- Zoom and pan support with mouse wheel and drag
- Previous / Next navigation through image galleries
- Download images directly from the viewer
- Rename assets inline from the viewer toolbar
- Delete assets with confirmation dialog
- Attach assets back into chat from the viewer
- Keyboard shortcuts: Escape to close, arrow keys to navigate
📄 Intelligent Document Understanding
- Automatic text extraction for PDF, DOCX, TXT, Markdown, RTF, CSV, and more
- Code-aware extraction for JavaScript, TypeScript, Python, Java, HTML, CSS, and more
- On-the-fly extraction for newly uploaded files with background persistence
- Intelligent chunking with query-relevant section retrieval for large documents
- PDF viewer with error fallback and download option if rendering fails
- Document viewer for all supported file types within the Library
🧠Smarter AI Context
- Automatic file context injection — model sees file content without re-upload
- Conversation-level file awareness across multiple chat turns
- File metadata block in system prompt listing all attached files
- Query-relevant chunk selection using keyword scoring for large documents
- Support for summarizing, explaining, comparing, and analyzing files naturally
- Planner detects document_analysis and file_analysis intents automatically
📎 Better Attachment Experience
- Attach library assets to active conversations without re-uploading
- Dedicated attach endpoint with conversation-file association
- Session-based active conversation tracking for seamless attaching
- Bulk attach support for multiple assets at once
- Visual feedback on successful attachment with toast notifications
Performance Improvements
- Virtualized grid rendering for fast Library loading with 10K+ assets
- Windowed scrolling with column-count calculation and translateY spacers
- Event-driven real-time updates instead of polling
- Improved caching and reduced unnecessary re-renders
- Better conversation synchronization across sessions
- Post-query filtering for aspect ratio without breaking pagination
🎨 User Interface Improvements
- Cleaner Library interface with polished asset cards
- Better responsive layout for mobile, tablet, and desktop
- Improved spacing and visual consistency
- Sidebar overlay on mobile with smooth transitions
- Document viewer with proper error states and retry
🛠Improvements
- Improved upload pipeline with reliable Library synchronization
- Better metadata management for file associations
- Enhanced conversation linking via message_id tracking
- Aspect ratio filtering for portrait, landscape, and square images
- Uploaded and Generated image filter categories
- Improved document processing with extraction status tracking
🞠Bug Fixes
- Fixed Library rendering issues with Suspense boundary for useSearchParams
- Fixed upload synchronization — uploads now emit events to refresh Library
- Fixed blank page issue in document viewer with error fallback
- Fixed asset categorization inconsistencies across sections
- Fixed stale-closure bugs in filter state management
- Fixed missing image filters (Uploaded / Generated) in Images section
Known Issues
- Image data URLs are not re-injected on follow-up messages — only text file content is available across turns
- Very large PDFs (>100 pages) may exceed the total file context budget of 80K chars
v0.8.6August 2026
A Smarter, Faster & More Polished Experience
This release focuses on refining the chat experience, improving responsiveness, strengthening file handling, and delivering a more polished interface across devices.
🚀 Faster Conversation Experience
- Improved conversation loading performance with reduced delays
- Smarter conversation caching for quicker switching between chats
- Reduced unnecessary loading delays when navigating conversations
- Background synchronization keeps conversations up to date across sessions
📠Better File Experience
- Improved attachment handling throughout the chat interface
- More reliable file upload workflow with better error feedback
- Better persistence for uploaded files across conversation reloads
- Improved attachment restoration when reopening conversations
ðŸ–¼ï¸ Media Experience
- Added Media Viewer support for chat images with fullscreen preview
- Images can now be previewed in a larger view with zoom and pan
- Improved image handling throughout conversations
- Foundation prepared for future support of PDFs, videos, and additional media types
💬 Composer Improvements
- Redesigned attachment tray inside the message composer
- Cleaner organization of attached images and files
- More stable composer layout with consistent sizing
- Improved spacing and visual consistency in the input area
📱 Responsive Improvements
- Improved adaptive layouts across desktop, tablet, and mobile devices
- Better spacing and scaling on different screen sizes
- Improved overall responsiveness with dedicated breakpoint handling
- More consistent interface across all supported devices
User Experience
- Refined interface interactions with smoother transitions
- Improved visual consistency across the entire application
- Better layout stability with reduced unexpected shifts
- Smoother overall chat experience with reduced visual flicker
🛠Under the Hood
- Performance optimizations across the frontend stack
- Improved UI architecture with better component organization
- Internal stability improvements and code quality enhancements
- Bug fixes across the chat interface for a more reliable experience
- Improved maintainability for future updates
Known Issues
- Large PDFs may take longer to process on initial load
- Research mode may time out on very complex multi-part queries
v0.8.5August 2026
Performance & Experience Update
This release focuses on performance, reliability, document handling, and overall user experience with significant under-the-hood optimizations.
🚀 Performance
- Much faster conversation loading — conversations appear nearly instantly
- Instant loading for recently opened conversations with intelligent caching
- Background synchronization keeps conversations updated across sessions
- Cursor-based conversation loading for improved scalability with large histories
- Smoother scrolling in long conversations with virtualized message rendering
- Better image loading performance with thumbnail caching
- Reduced unnecessary API requests through smart request deduplication
💬 Chat Experience
- Improved response reliability with better error recovery
- Better handling of uploaded documents with improved text extraction
- Improved attachment persistence across page refreshes
- Images and uploaded files now restore correctly when reopening conversations
- More reliable streaming experience during long responses
- Better markdown rendering with improved formatting support
- Improved mathematical expression rendering with KaTeX integration
📄 File & Document Support
- Support for PDF documents with text extraction
- Support for DOCX documents with text extraction
- Support for XLSX spreadsheets with text extraction
- Support for CSV files with structured data parsing
- Support for TXT, Markdown, JSON, and programming files
- Improved document understanding with context-aware extraction
- Smarter document context for better AI comprehension
- Improved handling of large files with intelligent chunking
User Experience
- New intelligent greeting system for a warmer welcome
- Greetings now rotate naturally across sessions
- Personalized greetings for returning users
- Time-aware greetings that adjust to time of day
- Cleaner conversation loading with reduced visual flicker
- Faster switching between conversations with smart preloading
🛠Reliability
- Better error handling across the application
- Improved file processing with better format detection
- Improved attachment recovery after network interruptions
- Better conversation consistency with background sync
- Reduced duplicate requests through intelligent caching
- Better synchronization between devices and sessions
🔒 Stability
- Improved database performance with optimized queries
- Better cache management with multi-layered caching architecture
- Faster API responses through query optimization
- More reliable conversation restoration when resuming chats
🧠Under the Hood
- Major performance optimizations across the stack
- Better conversation architecture with unified loading
- Intelligent multi-layered caching with automated invalidation
- Background synchronization engine for data consistency
- Database query optimization and index improvements
- Improved code quality and maintainability
Known Issues
- Large PDFs may take longer to process on initial load
- Research mode may time out on very complex multi-part queries
v0.8.0August 2026
Smarter Conversations
This release focuses on improving intelligence, speed, research, and user experience.
New Features
- Intelligent Research Engine — multi-step deep research with source synthesis
- Persistent File Uploads — files survive page refresh, logouts, and browser restarts
- Drag & Drop Upload — drag any supported file directly into the chat
- Better Math Rendering — KaTeX integration for LaTeX math expressions
- Streaming Responses — real-time token-by-token streaming with word counting
- Improved Conversation Memory — cross-chat memory with fact extraction
Improvements
- Faster first response — reduced initial latency by 35%
- Better Markdown rendering — GFM tables, task lists, strikethrough
- Cleaner chat input — compact bubble-style user messages
- Improved scrolling — smooth auto-scroll with smart detection
- Better mobile layout — responsive chat area and sidebar
🛠Bug Fixes
- Fixed favicon issue on production builds
- Fixed upload persistence across conversation switches
- Fixed auto-scroll bugs during streaming
- Fixed markdown rendering glitches with code blocks
🔒 Security
- Improved prompt injection protection in system prompt
- Better attachment validation with MIME type enforcement
Known Issues
- Large PDFs may take longer to process on initial load
- Research mode may time out on very complex multi-part queries
v0.7.0July 2026
Image Persistence & Device Intelligence
Images now persist across conversations, and SaysMee adapts to your device with smart keyboard shortcuts and touch support.
New Features
- Conversation Image Persistence — uploaded images saved to database, instantly restored on reload
- Device-Aware Interface — detects Windows, Mac, iOS, Android with platform-specific shortcuts
- OAuth Profile Pictures — avatar and full name from Google/Microsoft sign-in
- Guest Mode Improvements — clear login prompts for non-authenticated users
Improvements
- Compact message bubbles — user messages hug content width
- Touch-visible delete buttons on mobile
🛠Bug Fixes
- Fixed image loss when switching conversations
- Fixed Modifier key display on Windows vs Mac
- Fixed sidebar avatar fallback rendering
📄 Legal & Compliance
- Added Terms of Service and Privacy Policy to login page
- AI disclaimer below chat input area
v0.6.0June 2026
Deep Research Mode
SaysMee can now research complex topics autonomously, breaking questions into sub-queries and synthesizing structured answers.
New Features
- Deep Research Engine — multi-step research with sub-query breakdown, parallel searching, and synthesis
- Live research progress — step-by-step status updates during research
- Source citations — inline numbered citations with source card display
- Conversation Forking — create branching copies of conversations
Improvements
- Enhanced context recall with improved memory surface in relevant conversations
- Configurable research depth and result settings
🧠Memory System
- View, edit, and delete individual memories from settings
- Improved memory consolidation to reduce redundant storage
Known Issues
- Research mode may take 30-60 seconds on complex queries
v0.5.0May 2026
Memory System Launch
SaysMee now remembers who you are and what matters to you across conversations.
New Features
- Cross-Conversation Memory — SaysMee remembers preferences, facts, and context across chats
- Memory Settings Page — view and manage all stored memories
- Automatic Memory Extraction — key facts saved from each exchange
Improvements
- Redesigned conversation sidebar with visual hierarchy
- Conversation search functionality in sidebar
- Better code block rendering with syntax highlighting
- Mobile bottom navigation bar
🛠Bug Fixes
- Fixed sidebar scroll position on conversation switch
- Fixed memory duplication in rapid conversations
v0.4.0April 2026
Performance & Stability
Major performance improvements make SaysMee faster to load and more responsive.
Performance
- 40% reduction in initial page load time via code splitting
- Virtual scrolling for long conversations
- Reduced memory usage for active sessions
- Optimized API response times with caching strategies
🔒 Security
- Rate limiting on API endpoints (10 req/min per user)
- Input sanitization for message content
🛠Bug Fixes
- Fixed memory leaks in streaming connections
- Fixed conversation deletion cascading issues
- Fixed race condition in message ordering
v0.3.0March 2026
Web Search & Conversation Tools
SaysMee can now search the web for current information, and conversations are more flexible with forking and sharing.
New Features
- Real-Time Web Search — integrated Tavily API for current events and live data
- Source Citations — search results displayed with clickable source cards
- Conversation Forking — branch conversations to explore alternative directions
- Conversation Sharing — share chat links with others
Improvements
- Improved search result formatting and relevance ranking
- Configurable search regions and result counts
🛠Bug Fixes
- Fixed tool call formatting in streaming responses
- Fixed message ordering in forked conversations
Known Issues
- Web search limited to 5 results per query in basic mode
v0.2.0February 2026
UI Refinements & Privacy
A cleaner, more polished interface with improved privacy controls and documentation.
New Features
- Documentation Site — dedicated docs with Getting Started, Features, FAQ
- Legal & Privacy Pages — Terms of Service, Privacy Policy, Cookie Policy
- Theme Customization — dark/light/system mode with accent color selection
- Keyboard Shortcuts — Ctrl+K for search, Ctrl+N for new chat
Improvements
- Responsive design for tablet and mobile breakpoints
- Smoother animations and transitions
- Better error messages with recovery suggestions
🛠Bug Fixes
- Fixed layout shift on initial page load
- Fixed OAuth redirect loop on expired sessions
v0.1.0January 2026
Initial Launch
The first release of SaysMee — a modern AI chat experience powered by Llama 4 models via Groq's high-speed inference.
🚀 Launch Features
- AI Chat powered by Llama 4 Scout 17B via Groq API
- User Authentication — email/password, Google, Microsoft sign-in via Supabase
- Conversation Management — create, rename, delete, and organize chats
- Dark Mode Interface — modern, minimal design with CSS variable theming
- Markdown Rendering — GFM, code blocks, tables, and lists
- Responsive Design — works on desktop, tablet, and mobile
🔒 Security
- Supabase Auth with secure HTTP-only cookies
- Row Level Security on all database tables
- Session management with automatic token refresh
Known Issues
- Long conversations may experience performance degradation
- Mobile layout has some overflow edge cases
Was this page helpful?