Release Notes

​

🧠 Agent, Tools & Chat Experience

• Assistant-First Tool Surface: Re-engineered the main agent interface to natively handle inline tool discovery, tag-based tool searches, and clear tool-invocation logging.
• Managed Skill Cleanups: Streamlined agent capabilities by retiring obsolete creative skill bundles and introducing inline managed-skill activation.
• Smarter Modeling & Caching: Added native Claude prompt caching for faster response times, alongside explicit max_tokens protection to eliminate mid-stream chat cuts.
• Enhanced Failure Resilience: Upgraded the system to gracefully handle long-turn tool loops with soft-stop alerts and automated retries for transient model or provider drops.

​

🔌 Connectors, OAuth & Cloud Sync

• Hardened OAuth Integration: Migrated security client configurations directly to the backend architecture and enabled proxy-driven PKCE security for robust connector stability.
• Platform-Wide App Links: Finalized mobile connector handoffs for Google Drive and GitHub via a single universal App Link callback mechanism.
• Deep Platform Read/Write: Expanded integration capabilities, adding secure writing scopes to post directly to Slack and label-scoped reads for precise Gmail access.
• Baidu Netdisk Support: Deployed Baidu Netdisk as a completely standalone cloud synchronization destination target.
• Browser Extension Upgrades: Extended the companion browser extension to support automated history sync across Qwen and Kimi engines.

​

🏗️ Memory, Documents & Ingest Pipeline

• Advanced Directory Browsing: Upgraded the system Documents view to support an intuitive filesystem folder-hierarchy view, explicit typed file-path imports, and clean AI-generated file indicators.
• Granular Folder Watcher Rules: Enhanced the background folder monitor with multi-select rules, a real-time preview tool, and selective file exclusion pathways.
• Digestion Performance Pass: Drastically lowered system CPU usage during heavy background monitoring while unlocking full search capability inside complex Microsoft Excel (.xlsx) multi-sheet archives.
• Prioritized Queues: Taught the ingest engine to automatically process chat-appended documents immediately, letting you interact with fresh data without waiting on background backlogs.

​

📱 Mobile Parity (iOS & Android)

• No More Mobile-Lite: Deployed a full-featured native Documents screen matching the desktop layout, complete with support for generated file artifact previews.
• True Chat Attachments: Mobile users can now natively upload and attach physical files directly into active chat threads.
• Sync Keep-Alive: Implemented an Android foreground notification layer and specialized iOS keep-alive hooks to maintain background knowledge sync without the operating system putting the app to sleep.
• UI Modernization: Overhauled mobile exploration with a slick new home drawer layout, native action selector sheets, and a dedicated Beta application badge.

​

🛠️ Desktop Polish & Experimental Features

• Experimental Mode App Builder: Power users can now go to Settings → Advanced to access the brand-new App Builder. This allows early explorers to create editable apps and custom tools supported by a hardened widget runtime.
• Invite-a-Friend Referrals: Integrated a native beta sharing portal featuring simple referral-code management right in the app.
• Interface Smoothing: Added multi-axis webview zooming (keyboard + trackpad gestures), instant sidebar visual updates after background tasks complete, and fresh macOS menu bar and Windows packaging installers.

​

🏗️ Memory, Documents & Ingest Pipeline

• Resource Optimization: Drastically reduced idle CPU and hardware overhead during active background document monitoring and file ingestion.
• Deep Spreadsheet Searching: Upgraded the data digestion pipeline to parse and index every sheet within complex Excel (.xlsx) files, making them completely searchable in chat.
• Smart File-Watcher Filters: Enhanced the automated folder watcher to detect and skip Microsoft Office temporary/owner files (with a ~$ prefix) to prevent processing loops.
• Storage Pinning: Fixed a newline expansion bug in memory persistence and pinned storage providers to a stable local base_dir path configuration.

​

🧠 Agent & Chat Experience

• Thread-Level Connectors: Implemented interactive data engine connectors accessible and manageable natively directly within an active chat thread window.
• Dynamic Session Artifact Paths: Added system support for per-session artifact output locations. Relocated artifact paths are written back cleanly to the core data layer to preserve conversation histories.
• Input Isolation: Isolated active chat draft inputs to prevent accidental text bleeding across multiple open sessions.

​

🛠️ Desktop Polish & System Maintenance

• Background Life Cycle Sliders: Introduced granular utility sliders inside the General Settings pane, including Keep running in menu bar and Launch at Startup for persistent background execution.
• UI Action Wording: Cleaned up desktop text variables by unifying monitored-folder action buttons to read “Stop monitoring” for clearer user guidance.
• Windows UI Concealment: Modified underlying runtime mechanics to hide console windows for spawned child processes, keeping the environment clean for Windows users.

​

📱 Cross-Platform Mobile Enhancements (iOS & Android)

• Model Picker Alignment: Aligned the mobile platform’s model selector interface to honor desktop regional configurations and gateway streaming logic.
• Persistent Android Syncing: Deployed a persistent Android notification service layer to keep background knowledge synchronization active.
• Gated Startup Sequences: Protected early data initialization sync routines against explicit local user identity states to completely avoid blank onboarding loops.
• Native Issue Reporting: Integrated an in-app error and feedback utility interface directly into the mobile layout.

​

🧠 Agent & Chat Experience

• ZeroClaw Runtime Transition: Replaced the legacy skill-framework path with the unified ZeroClaw runtime, incorporating layered tool resolution and isolated multi-model tool replays.
• Local Artifact Generation & Workflow: Added native engines to generate structured documents (PDF, DOCX, PPTX, XLSX) and region-aware raster images via AIGW, featuring side-drawer previews and inline document chips.
• Active Workspace Mechanics: Implemented session pinning/unpinning, inline conversational search, empty-state quick actions, and automated ingestion of chat-generated files back into active memory.
• App Builder & Extensibility: Introduced an editable App Builder with a portable sidecar runtime, a visual Skill Wizard, slash-command pickers, and tool-token debug metadata panels.

​

🔌 Connectors & External Knowledge

• OAuth-Backed Umbrella: Standardized the external knowledge layout to support system-wide, per-turn tool visibility across desktop distributions.
• Deep Platform Integrations: Added direct OAuth pipelines with granular per-tool permission drawers for Slack (including channel-scoping), Gmail, GitHub, Google Calendar, OneDrive/SharePoint, Dropbox, Box, Outlook, and Confluence.
• Background Browser Extension Sync: Enhanced the web extension to support background chat-history synchronization across Claude, DeepSeek, and Doubao engines.

​

🏗️ Memory, Documents & Ingest Pipeline

• Unified Core Library Navigation: Consolidated the sidebar layout to map to the new Documents view (unifying text assets and chat artifacts) and renamed the user engagement logs from Ideas to Active Chats.
• Automated Folder Watcher: Implemented a localized background file-system watcher equipped with an Ingest Queue UI, next-step call-to-actions, and specialized path routing for OneDrive and Dropbox renames or moves.
• Advanced Processing Pipeline: Upgraded data digestion to parse large XLSX sheets, execute vision-based text extraction on PDFs/images, and partition content into optimized segments using the core extractor skills.
• MemLight Infrastructure Hardening: Secured gateway routing defaults, added cross-user cache isolation during OAuth account switches, and deployed a local filesystem fallback framework for real-time diagnostics.

​

🛠️ Desktop Polish & System Maintenance

• Onboarding Agent Workflow: Deployed a dedicated “Getting Started” onboarding agent to guide users through initial vector space configuration and memory structure maps.
• Telemetry & Privacy Control: Integrated a monthly LLM token consumption panel alongside a secure, end-user data privacy reset button.
• Distribution Safeguards: Polished Windows MSI passive installation execution loops and implemented hardened application update handoffs.

​

🏗️ Advanced Organization

• Dedicated Memory Sync Tab: Transitioned Memory Sync settings to a standalone interface to streamline setup and improve cloud storage transparency.
• Improved Office Extraction: Enhanced the system’s ability to read and understand Microsoft Office documents, ensuring higher quality for professional research.
• Memory Space UI Polish: Refined the Memory Space creation panel with a more compact design and corrected template scrolling behavior.
• The “Memories” Rebrand: Updated internal names from “Memory” to Memories to better reflect the interconnected nature of your Personal Knowledge Base.

​

🧠 Smarter Chat

• Triple Model Mobile Support: Mobile users can now select and toggle between up to three active AI Experts simultaneously.
• Default Reasoning Upgrade: Updated the core model logic to provide faster and more configurable reasoning out of the box.
• Enriched System Prompts: Improved multi-turn conversation accuracy by better organizing how context is sent to the AI.
• Response Utility: Added a copy toolbar to code blocks and a dedicated “Copy Response” button for seamless content transfer.

​

🛠️ Core Infrastructure

• Mobile Config Editor: Introduced a dedicated settings editor for mobile power users to manage app behavior on the go.
• Sync Resilience: Resolved connection failures for Google Drive on mobile and stabilized sync settings during cloud storage transitions.
• Performance Optimization: Improved the display of long conversations to prevent UI lag and added “Scroll to Bottom” controls for easier navigation.
• Engine v0.5.5.1: Updated the core desktop engine for better stability and refreshed cloud storage configurations.

• Module Segmentation: Organized backend components for better maintainability and faster diagnostic cycles.
• Expert Alignment: Standardized how different AI Experts behave to ensure consistent results across different models.
• Sync Guard Fix: Improved document reading logic to prevent synchronization errors.

• Data Repair: Stabilized draft synchronization and implemented automated repair for corrupted session logs.
• Compact Workspaces: Redesigned the “Create Memory Space” panel to be more compact, optimizing space for smaller displays.
• Guidance UI: Resolved issues with the onboarding pages to ensure a smoother start for new users.

• Interactive Refinements: Polished the sidebar, user menu, and Product Tour popovers for better visual feedback.
• Feature Usage: Integrated tools to report which features are used most, helping our team understand which tools are most valuable to you.
• Markdown Utility: Added table actions and improved code styling for more readable research outputs.

• Interactive Product Tour: A new guided experience introducing Personas, Memory Spaces, and AI Expert selection.
• Smart Adding: Documents can now be assigned to specific Memory Spaces instantly via direct drag-and-drop onto the sidebar.
• Refined Document UI: Replaced “Upload” with Add and redesigned empty states for better clarity.

​

🏗️ Advanced Organization

• Meet Memory Spaces Onboarding: Use the new step-by-step visual guide to establish your first Knowledge Spaces, helping you separate work, life, and learning.
• Google Drive Sync: Connect your Google Drive to index cloud documents directly into your Personal Knowledge Base.
• Enhanced Document Management: You can now edit document titles, categories, and Memory Space assignments directly for a faster workflow.
• Guided Space Setup: Use ready-made templates to set up pre-configured spaces in seconds for new projects.

​

🧠 Smarter Chat

• Upgraded Reasoning: Experience more precise and stable AI performance with the transition to updated models.
• UI Customization: Manage complex conversations better with the new expand/collapse functionality for AI responses.
• Mobile Experience: Enjoy improved clarity on the go with the new rendering engine for mobile chat messages.

​

🛠️ Core Infrastructure

• PDF Recovery: The document engine now allows you to recover text from partially corrupted or non-responsive PDF pages.
• Memory Optimization: Experience a more stable desktop engine with new limits designed to prevent system errors.
• Session Stability: Enjoy a smoother login experience with optimized handling that reduces refresh errors.

​

🧠 Smarter Chat

• Last Used Settings: MemoryBox now automatically restores your last used Persona and Memory Space when switching between chats.
• Response Rendering: Improved the reliability of how text and formatting appear during live AI responses.
• Model Stability: Enhanced multi-turn stability for Gemini models across complex tasks.

​

🏗️ Advanced Organization

• Deep Document Search: Introduced new tools allowing AI Experts to perform deeper, more detailed analysis of your local documents.

​

🛠️ Core Infrastructure

• Login Stability: Resolved an error message where the desktop app would occasionally report a failed login status upon reopening.
• Mobile Reliability: Fixed a critical build failure to ensure continued stability for the mobile app.

​

🧠 Smarter Chat

• Unified Entry Point: The chat start area has been relocated for a more streamlined, contextual start to your work.
• Better Memory: Conversations are now better organized for more accurate multi-turn reasoning.
• Performance Tuning: Improved chat responsiveness for a smoother experience.

​

📱 Mobile Evolution

• Memory Lists: Added dedicated list and detail pages for saved memories on mobile.
• Resilient Sync: Enhanced sync settings and introduced a graceful way to disconnect Google Drive.
• Mobile Runtime: Improved the mobile engine with a more resilient connection for on-the-go queries.

​

🛠️ Core Infrastructure

• Real-time Sync Fix: Resolved a critical issue with ChatGPT real-time synchronization in the extension.
• Tool Connection: Improved how tools authorize and connect to prevent timeouts.

​

🏗️ Advanced Organization

• Document Editing: Users can now edit documents and reassign them to different Memory Spaces directly within the app.
• Search Transparency: Added a status banner to notify users when the search index is being updated.
• Automated Indexing: The system now automatically updates your search index when the AI model is changed.

​

🧠 Smarter Chat

• Chat Toolbar: Added quick selectors directly into the chat toolbar for faster switching between tools and personas.
• Default Personas: Refined the logic to display your designated main persona across the platform.

​

🛠️ Quality of Life

• Mobile Session Management: Introduced a new flow for managing active mobile sessions.
• Manual Updates: Added a manual “Check for Update” button in the About section of the Desktop app.

​

🧠 Smarter Chat

• Doc Retrieval: Added tools allowing agents to fetch and analyze the full content of your documents.
• Refined Filtering: Introduced document search and filters for faster navigation.

​

🛠️ Core Infrastructure

• Cloud-Local Sync: Added sync status and a dedicated settings UI for better cloud visibility.
• Improved Adding: Expanded ChatGPT export support to handle divided files and missing attachments gracefully.
• Desktop UI Polish: Refined the sidebar layout and resizing behavior.

​

📱 Mobile Expansion (iOS & Android)

• Cross-Platform Knowledge: Access your Personal Knowledge Base on the go with mobile apps tuned for chat, personas, and memory spaces.
• Cloud-Local Sync: Integration with iCloud and other cloud services ensures your memories travel with you securely.

​

🏗️ Advanced Organization

• Memory Spaces: A completely rebuilt model for project management. Organize your world with topics, filters, and drag-and-drop links.
• Multi-Persona Workflows: Create distinct personas for work, creative, or personal mindsets.

​

Feature Enhancements

• Memory Assistant: Refined reasoning to improve the accuracy of how the AI recalls your information.
• Desktop Reliability: Optimized the build process for better performance across different computers.