| v0.8.5 | - Add MQTT TLS encrypted connection support for secure data transmission
- Add automatic Broker restart on abnormal exit, recovering service without manual intervention
- Add MQTT credential caching to avoid frequent re-authentication latency
- Add window.neomind.callExtension() frontend API for external page-to-extension orchestration
- Add click-to-view for image metrics in Map/CustomLayer popups
- Add ConfigPanel support for community components, enabling custom configuration UI
- Add "Advanced" tab to component config dialog, organizing low-frequency settings separately
- Improve extension timeout control and crash recovery — extensions auto-reload on failure instead of breaking the main flow
- Improve extension Transform to support Complex MetricValue types for nested metric calculations
- Fix Dashboard drag position jump; replace Web dialogs with native dialogs for more stable interaction
- Improve Dashboard responsive grid layout for better adaptation across screen sizes
- Improve component library UX with streamlined browsing and selection flow
| 2026-06-03 | ⬇️ |
| v0.8.4 | - Add Vision tool enabling Agent to perform multimodal image analysis via VLM
- Add layered multimodal capability detection via LiteLLM registry, auto-identifying model vision/text support
- Add settings sidebar navigation, replacing horizontal tabs for quicker access to settings
- Improve system prompt slimmed by 75% (7,500→1,800 tokens), reducing inference overhead while preserving instruction quality
- Improve Dashboard performance: lazy-load hls.js, eliminate cascading re-renders, reduce unnecessary state updates
- Improve extension ecosystem showcase visual design, highlighting core and community extensions
- Improve Dashboard toolbar with icon-only buttons, saving toolbar space
- Improve community component categorization, splitting into "My Components" and "Marketplace" views
| 2026-06-02 | ⬇️ |
| v0.8.3 | - Add Agent memory system with dual-file persistent layout, retaining context and preferences across sessions
- Add Memory tool with 5 actions × 5 targets for comprehensive memory management (store/retrieve/update/delete/search)
- Add web_fetch, file_write, and file_edit tools for Agent to fetch web content and read/write local files
- Add Docker deployment support with official Dockerfile for one-command containerized setup
- Add Agent experience learning, auto-extracting insights from task execution and reflecting on improvement strategies
- Add NEOMIND_MAX_CONTEXT environment variable for low-memory devices to limit context window and prevent OOM
- Improve GlobalChatFab from full-screen overlay to floating window, allowing Dashboard visibility during chat
- Fix community widget data flow issues, ensuring correct data passing and state synchronization between components
| 2026-06-01 | ⬇️ |
| v0.8.2 | - Add unified Dashboard DataSource architecture (Source+Mode separation), enabling flexible and reusable data source configs
- Improve image upload with automatic compression, preventing HTTP 413 errors from large images
- Improve LED data instant rendering with faster first-load speed
- Fix image history limit (4→unlimited), restoring full image browsing capability
- Fix mobile Dashboard drawer panel overlap for better mobile interaction
- Fix Dashboard component lifecycle issues, adding error boundaries to prevent single-component crashes from affecting the entire panel
| 2026-05-29 | ⬇️ |
| v0.8.1 | - Add MQTT Broker authentication and TLS certificate management, supporting self-signed and CA-issued certificates
- Upgrade embedded Broker engine from rumqttd to rmqtt for improved connection stability and protocol compatibility
- Add Broker connection guide in the Add Device dialog, streamlining new device onboarding
- Improve floating chat panel visual design for better conversation interaction
- Fix Base64 image display issue, ensuring embedded images render correctly
- Fix macOS resource limits issue, preventing file descriptor exhaustion under high concurrency
| 2026-05-27 | ⬇️ |
| v0.8.0 | - Add global AI chat floating action button (FAB) with full-screen overlay, enabling Agent interaction from any page
- Add 5 notification channels with channel editor, supporting email, Webhook, MQTT, in-app, and custom channels
- Add data push module with batch aggregation and delivery log tracking
- Improve messaging system architecture with retry mechanism, deduplication, and reliable delivery guarantees
- Fix installer service configuration and version upgrade compatibility issues
- Fix auto-updater not restarting app after update and version placeholder not being replaced
| 2026-05-26 | ⬇️ |
| v0.7.9 | - Add complete Dashboard rewrite with unified type system, API layer, data layer, and component architecture
- Add CLI resource management commands for devices, dashboards, rules, agents, messages, and widgets
- Add AI Build CLI integration, injecting CLI reference into Agent prompts for CLI-driven operations
- Add Webhook device adapter with auto-discovery and token management, enabling device integration via HTTP callbacks
- Fix Dashboard scroll white screen, eliminating render flicker during fast scrolling
| 2026-05-25 | ⬇️ |
| v0.7.8 | - Add AI Build Mode foundation framework, providing the underlying layer for visual AI orchestration
- Fix update dialog reappearing after restart, showing only once after upgrade confirmation
- Fix chat message flicker after streaming completion, stabilizing the final rendered result
- Fix extension marketplace loading timeout and telemetry data sorting issues
| 2026-05-16 | ⬇️ |
| v0.7.7 | - Improve telemetry storage identifier unified to "device:" prefix format, eliminating identifier ambiguity in queries
- Improve Transform Builder toolbar visual design for more intuitive operation entry
- Fix Dashboard scroll white flash for smoother long-page scrolling experience
- Improve chart component deduplication and unified caching, reducing redundant data requests
| 2026-05-15 | ⬇️ |
| v0.7.6 | - Add HTTP read/body timeout mechanism to prevent slow-client DoS attacks
- Improve storage sync debounce to 500ms, reducing write amplification on disk I/O
- Improve telemetry query condition pushdown to storage layer, reducing unnecessary data transfer
- Add extension component preloading, eliminating first-use loading delay
- Fix WKWebView Dashboard white screen, ensuring proper web view initialization
| 2026-05-14 | ⬇️ |
| v0.7.5 | - Add unified execution engine, consolidating into Focused / Focused+ / Free execution modes
- Add Agent image analysis persistence, retaining visual insights across conversation turns
- Improve tool loop exit guidance, helping smaller models properly stop after task completion
- Fix panic caused by UTF-8 truncation, ensuring safe handling of multi-byte characters
- Fix Dashboard grid layout and WKWebView rendering issues
| 2026-05-13 | ⬇️ |
| v0.7.4 | - Add NO_WEB backend-only deployment mode for running pure API services without frontend
- Add Dashboard sharing system with proxy authentication and ShareManager for managing share links
- Add community component marketplace with browse, install, and ZIP upload support for custom components
- Add Dashboard component device binding, linking components directly to specified device data
- Add extension device management and log viewer for debugging extensions without switching to terminal
- Upgrade security: migrate sensitive APIs to protected routes with unified authentication checks
| 2026-05-11 | ⬇️ |
| v0.7.3 | - Improve ReAct reasoning to unified single-loop architecture, eliminating redundant nested loops for higher efficiency
- Improve multimodal vision-first prompt strategy, auto-adjusting prompts for better analysis when images are present
- Improve Agent tool error messages and guidance copy, lowering the debugging barrier
- Add chat virtual scrolling, fixing lag in long message lists
- Improve extension push stream performance, reducing data delivery latency
| 2026-05-08 | ⬇️ |
| v0.7.2 | - Add data export with Excel/CSV format (date range filtering) and image ZIP batch export
- Add multi-instance management with API Key authentication and quick instance switching
- Add API Key service account pattern for third-party system integration
- Fix embedded Broker startup failure and device last_seen timestamp not updating
| 2026-05-06 | ⬇️ |
| v0.7.1 | - Add BLE provisioning for zero-touch device setup (WiFi / Cat.1 / HaLow network types)
- Add Broker auto-discovery guidance, prompting available Brokers when onboarding new devices
- Add Tauri native BLE transport layer with re-provisioning support
- Improve mobile navigation: hamburger menu replaced with scrollable tab navigation for better efficiency
- Fix crash caused by missing BLE permissions on macOS
| 2026-05-06 | ⬇️ |
| v0.7.0 | - Add redesigned Agent execution modes, consolidating into Focused (formerly Chat) and Free (formerly React), simplifying execution strategy selection
- Add VLM Vision Dashboard component with real-time video analysis and WebSocket-streamed AI inference results
- Add AI Metrics Registry and AI Metric Tool, enabling Agent-defined data collection with dynamic Data Explorer tabs
- Add AI Analyst Dashboard with data-driven auto-trigger analysis, invoking Agent to generate insights on metric changes
- Add event-driven architecture where extension outputs and device metric changes auto-trigger Agent execution
- Add Aurora/Glass UI design system using OKLCH color space, improving visual consistency and accessibility
- Improve quality hardening: replace all hot-path unsafe unwrap() with safe error handling, eliminating potential panics
- Improve input validation across all API endpoints with unified parameter verification and error response format
- Improve performance: adopt DashMap lock-free concurrency replacing mutex locks, eliminate N+1 query issues
- Improve UX: introduce skeleton loading, empty states, error boundaries, and confirmation dialogs throughout
| 2026-04-28 | ⬇️ |
| v0.6.9 | - Add transform aggregated tool for LLM-driven data transform
- Fix Agent cache invalidation, context compaction, and tool result formatting
- Fix anti-hallucination handling for tool results
| 2026-04-16 | ⬇️ |
| v0.6.8 | - Simplify LLM thinking control, unify thinking display in multi-round conversations
- Expose virtual metrics to LLM and add write_metric action
- Harden OTA update flow (fix done state, dual dialog, version sync)
- Fix rule builder extension support
- Reuse session-level tool result cache
| 2026-04-15 | ⬇️ |
| v0.6.7 | - Performance optimizations: paginated storage queries, concurrent query limits, batched frontend state updates
- Fix updater download progress stuck and Windows URL error
- Improve LLM vision detection and Agent thinking panel UX
| 2026-04-14 | ⬇️ |