- Add useCallback for togglePin and handleModelSelect to prevent unnecessary re-renders
- Refactor dropdown menu styling with more specific CSS scoping
- Simplify dropdown open/close logic
- Improve performance by memoizing function dependencies
- Adjust dropdown overlay styling and animation
* ✨ feat: add MCP servers in chat input
- Introduce MCPToolsButton component for managing MCP servers
- Add new icon for MCP server tools in iconfont.css
- Update Inputbar to include MCP tools functionality
- Add toggle functionality for enabling/disabling MCP servers
- Implement styled dropdown menu for server selection
- Add necessary type imports and useState for MCP server management
* ✨ feat: add support for enabling/disabling MCPServers per message (main)
- Added `enabledMCPs` property to the `Message` type to track enabled MCPServers.
- Modified `MCPToolsButton` to enable all active MCPServers by default using a new `enableAll` state.
- Introduced `filterMCPTools` utility to filter tools based on enabled MCPServers.
- Updated `AnthropicProvider`, `GeminiProvider`, and `OpenAIProvider` to filter tools using `filterMCPTools`.
- Enhanced `Inputbar` to include `enabledMCPs` in the message payload when set.
- Add dynamic group management button that changes based on group's current state
- Simplify group add/remove logic with a single button
- Enhance visual feedback for group-level model management
* feat: a button a add a whole group of models
* feat: search as typing in EditModelsPopup
* feat: add a button to remove a whole group of models
* feat: add remove button for model group in the model list
- Add database version 5 with schema updates
- Create `upgradeToV5` function to migrate Tavily web search metadata to new format
- Update types to support new web search metadata structure
- Minor code cleanup and formatting improvements
- Remove `enabled` flag from WebSearchProvider type
- Add `hasObjectKey` utility function to check optional properties
- Update WebSearchService to check web search availability based on API key/host
- Modify WebSearchSettings and WebSearchProviderSetting components to support API key/host validation
- Add Searxng provider in migration script
- Simplify web search provider configuration and validation logic
* ✨ feat: add Model Context Protocol (MCP) server configuration (main)
- Added `@modelcontextprotocol/sdk` dependency for MCP integration.
- Introduced MCP server configuration UI in settings with add, edit, delete, and activation functionalities.
- Created `useMCPServers` hook to manage MCP server state and actions.
- Added i18n support for MCP settings with translation keys.
- Integrated MCP settings into the application's settings navigation and routing.
- Implemented Redux state management for MCP servers.
- Updated `yarn.lock` with new dependencies and their resolutions.
* 🌟 feat: implement mcp service and integrate with ipc handlers
- Added `MCPService` class to manage Model Context Protocol servers.
- Implemented various handlers in `ipc.ts` for managing MCP servers including listing, adding, updating, deleting, and activating/deactivating servers.
- Integrated MCP related types into existing type declarations for consistency across the application.
- Updated `preload` to expose new MCP related APIs to the renderer process.
- Enhanced `MCPSettings` component to interact directly with the new MCP service for adding, updating, deleting servers and setting their active states.
- Introduced selectors in the MCP Redux slice for fetching active and all servers from the store.
- Moved MCP types to a centralized location in `@renderer/types` for reuse across different parts of the application.
* feat: enhance MCPService initialization to prevent recursive calls and improve error handling
* feat: enhance MCP integration by adding MCPTool type and updating related methods
* feat: implement streaming support for tool calls in OpenAIProvider and enhance message processing
* feat: Enhance MCPServer and MCPTool interfaces with optional properties and unique IDs
* fix(mcp): Refactor SSE transport initialization to use URL object
* fix(OpenAIProvider): correct inputSchema properties reference in tool parameters
* feat(MCPSettings): enhance server settings UI with new fields and improved layout
* feat(MCPSettings): add multilingual support for MCP server settings
* fix: remove unnecessary console log statements
* ✨ feat: add Model Context Protocol (MCP) server configuration (main)
- Added `@modelcontextprotocol/sdk` dependency for MCP integration.
- Introduced MCP server configuration UI in settings with add, edit, delete, and activation functionalities.
- Created `useMCPServers` hook to manage MCP server state and actions.
- Added i18n support for MCP settings with translation keys.
- Integrated MCP settings into the application's settings navigation and routing.
- Implemented Redux state management for MCP servers.
- Updated `yarn.lock` with new dependencies and their resolutions.
* 🌟 feat: implement mcp service and integrate with ipc handlers
- Added `MCPService` class to manage Model Context Protocol servers.
- Implemented various handlers in `ipc.ts` for managing MCP servers including listing, adding, updating, deleting, and activating/deactivating servers.
- Integrated MCP related types into existing type declarations for consistency across the application.
- Updated `preload` to expose new MCP related APIs to the renderer process.
- Enhanced `MCPSettings` component to interact directly with the new MCP service for adding, updating, deleting servers and setting their active states.
- Introduced selectors in the MCP Redux slice for fetching active and all servers from the store.
- Moved MCP types to a centralized location in `@renderer/types` for reuse across different parts of the application.
* feat: enhance MCPService initialization to prevent recursive calls and improve error handling
* feat: enhance MCP integration by adding MCPTool type and updating related methods
* feat: implement streaming support for tool calls in OpenAIProvider and enhance message processing
- Modify AppUpdater to separate update dialog logic
- Add new IPC handler for manually showing update dialog
- Update renderer hooks and store to track downloaded update state
- Switch import for UpdateInfo from electron-updater to builder-util-runtime
* feat(export): Enhance markdown-to-docx export with table support
Add comprehensive table rendering capabilities to the ExportService, including:
- Support for table headers and body rows
- Configurable cell styling and alignment
- Handling of inline text formatting within table cells
* feat(export): Add hyperlink support in markdown-to-docx export
Enhance ExportService to handle hyperlinks during document export:
- Implement link detection in inline tokens
- Create ExternalHyperlink with proper styling
- Preserve link text and URL in exported document
- Add new `filterUserRoleStartMessages` function in MessagesService
- Update Anthropic, Gemini, and OpenAI providers to use new message filtering
- Refactor message handling to ensure user messages start the conversation
- Remove redundant message filtering logic from individual providers
- Add vision model check before image upload in Inputbar
- Implement flexible message start forcing for specific models
- Improve provider-level message routing logic
* feat: add resize handle to input textarea with drag interaction
* handle auto size inputbar
* optimize auto resize function and add i18n
* fix: expand button bug in inputbar and rebase to latest main
* rebase to main
- Fixed typo in Tencent Cloud TI provider key from 'tentent-cloud-ti' to 'tencent-cloud-ti'
- Enhanced ProviderSettings search input with search icon and custom styling
- Adjusted button border radius for consistent UI design
- Implemented `useKnowledgeFiles` hook for managing knowledge base files
- Added localization support for knowledge base file management in multiple languages
- Created UI option to remove all knowledge base files in DataSettings
- Updated file size formatting utility function
- Modified ContentView and FilesPage to use file size correctly
- Implemented detailed backup progress tracking in BackupManager
- Added new BackupPopup component for backup process visualization
- Enhanced backup process with file copy progress and stage tracking
- Updated localization files with backup progress translations
- Integrated backup progress reporting to renderer process
- Refactored reasoning effort handling for Claude models
- Added type definition for ReasoningEffort
- Simplified budget token calculation
- Improved type safety and readability of the method
- Refactored getReasoningEffort method to handle reasoning configuration more robustly
- Added type definitions for reasoning effort and configuration
- Simplified logic for calculating budget tokens
- Improved type safety and readability of the method
* refactor: Improve Ellipsis component and usage
- Modify Ellipsis component to use children instead of text prop
- Add support for multi-line and single-line ellipsis with styled-components
- Update KnowledgeContent to use new Ellipsis component structure
- Enhance ClickableSpan styling for better text truncation
* fix: Improve text wrapping in Ellipsis component
Add overflow-wrap: break-word to ensure long words are properly truncated in multi-line ellipsis
* refactor: Improve link and tooltip rendering in KnowledgeContent
- Wrap links with ClickableSpan for better interaction and styling
- Adjust Tooltip and Ellipsis placement for improved readability
- Remove unnecessary inline styling for links in ItemInfo
---------
Co-authored-by: lizhixuan <zhixuan.li@banosuperapp.com>
- Created detailed technical documentation explaining the KnowledgeService's task queue and workload management
- Added JSDoc comment to KnowledgeService.ts to provide high-level overview
- Documented key aspects of concurrent task processing, including workload evaluation, task state management, and resource optimization strategies
* feat: Improve file upload for the knowledge base
* feat: Improve file upload for the knowledge base
* feat: Improve file upload for the knowledge base
---------
Co-authored-by: 亢奋猫 <kangfenmao@qq.com>
- Added Zhihu (知乎直答) mini app logo
- Updated minapps configuration to include Zhihu mini app
- Modified store migration to enable Zhihu mini app by default
- Added Cici mini app logo in WebP format
- Updated mini apps grid to display 8 columns instead of 6
- Updated store migration to include Cici in enabled mini apps
- Updated Cici app logo import path
- Move StyleSheetManager logic from App.tsx to a new dedicated context component
- Simplify App.tsx by importing the new StyleSheetManager
- Preserve existing prop validation and filtering logic
- Add Babel plugin for styled-components with performance and naming optimizations
- Integrate @emotion/is-prop-valid for improved prop filtering in StyleSheetManager
- Update Antd Input components to use variant="borderless" instead of deprecated bordered prop
- Add new dependencies for styled-components configuration
- Add hunyuan-turbos-latest
- Add hunyuan-embedding
- Fix the issue of the embedding model displaying the internet icon (when the provider fully supports web search)
- Add min-width and min-height to ModelAvatar to ensure consistent sizing
- Separate AssistantName into a new styled component for better text overflow handling
- Refactor AssistantItem to improve text display and icon alignment
- Implement alphabetical sorting for model groups
- Use lodash's sortBy and toPairs to order model groups
- Modify rendering to use sorted model groups
- Create new ModelEditContent component in ProviderSettings
- Separate model editing logic from ProviderSetting
- Adjust styling for AssistantItem and TopicsTab list items
- Refine Segmented component styling with transparent background
- Improve modal layout and interaction for model type configuration
- Implement new setting to show/hide model icons in assistant list
- Add localization support for new assistant settings
- Update UI to conditionally render model avatar in AssistantItem
- Modify settings store to include showAssistantIcon state
- Enhance display settings with new toggle switch for assistant icon
- Remove custom Segmented styling from AntdProvider
- Update HomeTabs Segmented component with refined styles
- Consolidate Segmented styling in a single location
- Improve visual consistency with rounded corners and transparent background
- Update window vibrancy to 'sidebar' for better visual effect
- Refine input bar styling with softer background and rounded corners
- Adjust toolbar button sizes and styling
- Modify topic position selection to use Segmented component
- Tweak light theme background opacity
* feat: add "More Settings" option in multiple languages and enhance model type selection UI
* feat: add "More Settings" option with warnings and confirmation prompts in multiple languages
* fix: improve modal close handling and reset model type visibility
Restrictions are introduced in earlier beta versions of MacOS Sequoia.
However, in the public release, Apple revoke this restrictions.
Alt+[Char] shortcut has no restrictions now.
- Add location check to prevent navigating to settings when already on settings page
- Enable hotkey on content editable and form elements
- Use useLocation to track current route
* feat: Implement robust favicon loading with fallback mechanisms
* refactor: Improve favicon loading state and use Promise Method
* refactor: Extract FallbackFavicon into a separate component
* feat: Add Splitbee favicon service to fallback favicon URLs
* feat: Add search button to the knowledge base, and adjusted the display for drag-and-drop uploads
* feat: Add search button to the knowledge base, and adjusted the display for drag-and-drop uploads
- Standardize spacing and capitalization in Chinese (Simplified and Traditional) locales
- Enhance readability of error and configuration messages
- Consistent formatting for API keys, database IDs, and URLs
- Update Tavily description across locales to emphasize AI agent capabilities
- Add blacklist functionality for web search results
- Introduce configurable maximum search results
- Improve localization for web search settings in multiple languages
This commit involves minor reorganization and cleanup of translation files across multiple languages (en-us, ja-jp, ru-ru, zh-cn, zh-tw). Changes include:
- Reordering of some translation keys
- Removing duplicate entries
- Slight restructuring of nested translation objects
- Minor formatting adjustments
This commit removes unnecessary console.log statements and makes minor code cleanup in several files:
- PlantUML component
- WebSearchSettings
- WebSearchService
- Store settings
Additionally, it updates the store migration to handle LM Studio provider addition and reverts the store version
- Added `gemini-2.0` to `visionAllowedModels`.
- Introduced new Gemini models in `SYSTEM_MODELS`:
- `gemini-1.5-flash-8b`
- `gemini-2.0-flash`
- Updated Yi model configurations:
- Renamed `yi-lightning` to "Yi Lightning".
- Deprecated `yi-medium`, `yi-large`, and `yi-vision` (replaced by `yi-lightning` per documentation).
- Added `yi-vision-v2` as a new version of Yi Vision.
- Added new image asset `codestral.png` for Codestral model.
- Updated `embedings.ts` to include the `mistral-embed` embedding model with `max_context: 8000`.
- Updated `models.ts` to:
- Add support for Codestral model with `codestral-latest` ID.
- Introduce new Mistral models, including:
- `ministral-3b-latest`
- `ministral-8b-latest`
- `mistral-large-latest`
- `mistral-small-latest`
- `mistral-embed`
- Organize models into appropriate groups (e.g., "Mistral Chat", "Mistral Code").
- Adjusted `getModelLogo` function to include Codestral and Ministral logos.
This commit involves several improvements to localization files across different languages:
- Reordered and cleaned up translation keys
- Removed redundant entries
- Ensured consistent ordering of keys
- Added missing translations for various features
- Normalized whitespace and formatting
- i18n Updates:
- Refactored WebDAV-related translations into nested JSON structures for better organization.
- Added support for pluralization in time intervals (minutes and hours) across all locales (en-us, ja-jp, ru-ru, zh-cn, zh-tw).
- UI Enhancements:
- Updated `DataSettings` and `WebDavSettings` components to use the new i18n keys for time intervals.
- Improved the `Select` dropdown for sync intervals with dynamic pluralization based on locale.
- Adjusted input field widths for better alignment and consistency.
- Code Cleanup:
- Removed redundant comments and unused code in `WebDavSettings.tsx`.
- Simplified button and input styling for a cleaner layout.
* feat: agent can select multiple knowledge bases
* feat: basic search multiple knowledge base
* fix bug: knowledge base is delete, assistants and agents sync delete
* fix bug: assistant and knowledge base button sync
* feat: allow to search multiple knowledge base
* chore: finish rebase to upstream/main
- Update llm.ts to include LM Studio and ModelScope in initial system providers
- Modify migrate.ts to add migration logic for adding these new providers
- Ensure providers are added only if they don't already exist in the configuration
PR #1253 fixed reasoning time calculation for APIs that return reasoning
content in `delta.content`, but introduced a regression for those
returning it in `delta.reasoning_content`. This commit fixes the
regression.
Fixes#1593
- Remove image loader from dependencies
- Update file loading to skip image, video, and audio files
- Add logging for knowledge base file processing
- Modify common file extensions list
- Add type ignore for LocalPathLoader to resolve TypeScript issues
- Update OpenAI package to version 4.77.3
- Add confirmation modal for max tokens setting
- Modify max tokens input to use InputNumber instead of Slider
- Update localization strings for max tokens and URL input
- Enhance knowledge base URL input to support multiple URLs
- Improve error handling and input validation
* feat: knowledge base support more file types
* chore: add common document types
* feat: finish basic odloader
* feat: finish open document loader
* feat: support more type from dictionary
* fix: 删除目录时仍能检索到信息
* chore
* feat: 🎸 使用@呼出模型选择列表
输入第一个字符为@符号的时候可以呼出选择模型的列表
* feat: 🎸 Only one can be chosen at a time
一次只能选择一个模型，选择后自动关闭。选择过的模型不在出现在列表，避免删除模型的时候显示异常。
* fix: 🐛 When choosing the model, Enter will send a message
* feat: 🎸 选中的模型显示供应商
* feat: 🎸 pinned module show privoder
* feat: 🎸 only selected modle show provider
* feat: 🎸 删除@符号以后自动关闭
* feat: 🎸 增加模糊搜索
---------
Co-authored-by: duanyongcheng77 <duanyongcheng77@gmail.com>
* fix: retain selected model when regenerating message instead of reverting to default model
* fix: use props model
---------
Co-authored-by: Shern <shenkeji@corp.netease.com>
- Incremented version from 60 to 61 in the persisted reducer configuration.
- Updated migration logic to change the provider type for 'qwenlm' from 'openai' to 'qwenlm', ensuring correct identification in the state management.
- Updated Message component to fallback to message.model if model retrieval fails, improving robustness.
- Refactored MessageHeader to utilize getModelName for better user name display based on message role, enhancing clarity.
- Introduced getModelName function in ModelService to streamline model name retrieval, improving code modularity and readability.
- Updated AddAssistantPopup to improve layout and styling, ensuring better overflow handling and text display.
- Refactored AgentsPage to utilize a new utility function for grouping agents, enhancing data management and organization.
- Exported getAgentsFromSystemAgents function for better modularity and reusability across components.
- Updated translateText function to accept an optional callback for handling translated text directly within the function.
- Refactored MessageMenubar to utilize the new callback mechanism, improving the flow of translated content handling.
- Enhanced error handling during translation to ensure better user feedback in case of failures.
- Introduced FileService to handle file reading operations via IPC.
- Implemented a new IPC handler for reading files, enhancing the application's ability to access and manage data.
- Integrated system agents from a JSON file, allowing dynamic loading of agent data into the application.
- Updated the AgentsPage and AddAssistantPopup components to utilize the new system agents, improving user experience and functionality.
- Enhanced application state management by adding resourcesPath to the runtime state, ensuring proper resource handling across components.
- Replaced direct usage of modelId with model object in Message, MessageHeader, MessageMenubar, and TranslatePage components for consistency.
- Introduced getMessageModelId utility function to streamline model retrieval from messages.
- Updated event handling in Messages component to align with new model structure.
- Enhanced code readability and maintainability by reducing redundancy in model handling.
- Updated MessageMenubar to display the resend button only for user messages, enhancing user experience and preventing unnecessary actions for other roles.
- Refactored the children prop of TextEditPopup to include conditional rendering logic based on message role.
- Added new styles for message thought containers and group message wrappers to improve UI layout.
- Updated MessageGroup component to dynamically set the selected message index based on message length.
- Introduced a new event for appending messages, enhancing message handling capabilities.
- Refactored MessageMenubar to support the new append message functionality.
- Adjusted multi-model message style setting to 'fold' for better user experience.
- Improved responsiveness of message grid layout for smaller screens.
- Replaced hardcoded language translation options with a dynamic mapping from TranslateLanguageOptions.
- Improved maintainability and scalability of the translation feature by utilizing a centralized configuration for language options.
- Introduced a new image asset for the FlagOpen model in the assets directory.
- Updated the models configuration to include the FlagOpen logo, allowing for its use in the model logo mapping.
- Refactored `getMessageBackground` function for better readability.
- Updated `MessageContent` component to use a new `withMessageThought` utility for extracting reasoning content from messages.
- Changed fragment usage to `Fragment` for consistency in JSX.
- Enhanced message handling by separating reasoning content from the main message content.
* Fix: ESC key to exit the expanded editor
* Add the shortcuts for 'clear' and 'new context' to the input bar
Clear Messages: Ctrl+L
New Context: Ctrl+R
https://github.com/CherryHQ/cherry-studio/issues/740https://github.com/CherryHQ/cherry-studio/issues/766
* Fix: the paste issue when copying from an email (content was pasted as an image; ensure it is pasted as text). Prioritize the text in the clipboard during pasting.
If there is no subsequent message or if the next message is from the user, this message should be resent. delete the old message after processing is complete.
