* refactor(GeminiService): migrate to new Google GenAI SDK and update file handling methods
- Updated import statements to use the new Google GenAI SDK.
- Refactored file upload, retrieval, and deletion methods to align with the new SDK's API.
- Adjusted type definitions and response handling for improved type safety and clarity.
- Enhanced file listing and processing logic to utilize async iteration for better performance.
* refactor(GeminiProvider): update message handling and integrate abort signal support
- Refactored message content handling to align with updated type definitions, ensuring consistent use of Content type.
- Enhanced abort signal management for chat requests, allowing for better control over ongoing operations.
- Improved message processing logic to streamline user message history handling and response generation.
- Adjusted type definitions for message contents to enhance type safety and clarity.
* refactor(electron.vite.config): replace direct import of Vite React plugin with dynamic import
* fix(Gemini): clean up unused methods and improve property access
* fix(typecheck): update color properties to use CSS variables
* feat: 修改画图逻辑
* fix: import viteReact
---------
Co-authored-by: eeee0717 <chentao020717Work@outlook.com>
* feat: implement OAuth client provider and lockfile management
* feat: implement OAuth callback server and refactor authentication flow
* fix(McpService): restrict command handling to 'npx' for improved clarity
* refactor: make callbackPort optional in OAuthProviderOptions and clean up MCPService
* refactor: restructure OAuth handling by creating separate callback and provider classes, and remove unused utility functions
* feat(MessageContent): 添加引用内容折叠功能，优化用户界面交互
* feat(Citations): add hideTitle prop to control title visibility in CitationsList
* feat(Messages): add message update functionality and manage UI state for citations and web search
* fix: web search citation
---------
Co-authored-by: 自由的世界人 <3196812536@qq.com>
* feat(model): add ModelCard schema and related types for input/output capabilities
* refactor(model): make limits and price properties optional in ModelSchema
* feat(model): add textGeneration capability to ModelSchema
* feat: integrate AoxisProxy for HTTP requests in rerankers and CopilotService
- Replaced direct axios calls with aoxisProxy in JinaReranker, SiliconFlowReranker, and VoyageReranker to utilize proxy settings.
- Introduced AoxisProxy service to manage axios instances with proxy configurations.
- Updated CopilotService to use aoxisProxy for API requests, ensuring consistent proxy handling across services.
* refactor(AxiosProxy): improve proxy handling and initialization logic
* fix tyop
* fix tyop
* Add support for custom HTTP headers in MCP servers
Allow users to configure custom HTTP headers for SSE and streamable HTTP
MCP server connections. This enables authentication and other API
requirements.
* Add custom headers i18n strings
* Add typed MCPCallToolResponse interface and format converters
The commit introduces a structured response type for MCP tool calls with
proper handling for different content types (text, image, audio). It adds
provider-specific converter functions to transform tool responses into
the appropriate message format for OpenAI, Anthropic, and Gemini.
* Support vision models in tool call responses
Add isVisionModel parameter to tool response formatters to conditionally
handle content based on model capabilities. Non-vision models now receive
JSON stringified content, while vision models get proper multimedia parts.
Add support for webpage summarization, direct URL references, and
better query processing using a structured XML format. Move web content
fetching to dedicated utility functions with improved error handling
and format options.
- Added 'icon' class to various icons in Sidebar for consistent styling.
- Removed unused loading state from Messages component.
- Cleaned up iconStyle variable in ModelSettings as it was no longer needed.
feat(Inputbar, Settings): add backspace delete model functionality and localization updates
- Implemented a new setting to enable backspace key functionality for deleting models/attachments in the Inputbar.
- Added corresponding localization strings for English, Japanese, Russian, Chinese (Simplified and Traditional) in the i18n files.
- Updated the QuickPanelBody styling to inherit border-radius.
- Migrated the new setting to the state management for persistence.
Co-authored-by: 亢奋猫 <kangfenmao@qq.com>
* fix(provider config): update dashscoop new links
* feat(EditModelsPopup): add grouping function for bailian
* fix(isWebSearchModel): Correctly handle the priority of manually setting model support for web search
- Reduced avatar size in ModelList for better alignment.
- Simplified rendering logic for model descriptions in EditModelsPopup to enhance readability.
- Updated MainContainer to use flex display for better layout.
- Increased margin in NpxSearch component for improved spacing.
- Adjusted ResultList to use two columns instead of three for better content presentation.
* refactor(AgentPage): Refactor AgentPage UI
* style(AgentCard): update HeaderInfoEmoji styling for improved layout and visual consistency
* fix(AgentCard): conditionally render HeaderInfoEmoji to prevent rendering of undefined
* feat(AgentsPage): add handleAddAgent function to streamline agent addition process
* style(AgentsPage): remove unnecessary whitespace in title rendering
- Added a new SVG icon for npm in the MCP settings.
- Introduced a custom hook `useMCPServer` for retrieving a specific MCP server by ID.
- Updated localization files to include new error messages for tool and prompt loading in English, Japanese, Russian, and Chinese.
- Refactored MCP settings components for improved navigation and state management, including the use of React Router for routing.
- Enhanced the Npx search functionality and UI for better user experience.
- Introduced a styled container to the NewContextButton component to hide it on smaller screens (max-width: 800px).
- Ensured the tooltip and button functionality remain intact while enhancing the component's layout.
- Removed unused QuestionCircleOutlined icon and Popconfirm from Inputbar, replacing it with a direct button click for clearing topics.
- Refactored message display logic in Messages component to enhance clarity and maintainability, while preserving existing functionality.
* feat(MCPSettings): implement server selection and navigation with back button
* chore(ui)
* chore(UI): npx search padding
* feat(NpxSearch): add server selection and navigation; update styles
---------
Co-authored-by: eeee0717 <chentao020717Work@outlook.com>
* feat(MessageTools): add error handling and status indicator for tool responses
* feat(i18n): add error message for tool invocation in English, Japanese, and Russian locales
* Add MCP prompt listing and retrieval functionality
* Add generic caching mechanism for MCP service methods
Refactor caching strategy by implementing a higher-order withCache function
to centralize cache logic and reduce code duplication. Separate implementation
details from caching concerns in listTools, listPrompts and getPrompt methods.
# Conflicts:
# src/main/services/MCPService.ts
* Add MCP prompts listing feature
- Add IPC handlers for listing and getting prompts
- Create UI component to display available prompts in settings tab
- Improve error handling in MCP service methods
* fix(McpService): add error handling for tool and prompt listing methods
* feat(MCPSettings): enhance prompts and tools sections with improved UI and reset functionality
* feat(i18n): add tabs and prompts sections to localization files
* feat(MCPToolsButton): add MCP prompt list functionality to Inputbar
* feat(McpSettings, NpxSearch): improve user feedback with success messages on server addition
* feat(MCPService, MCPToolsButton): enhance prompt handling with caching and improved selection logic
* feat(MCPToolsButton): enhance prompt handling with argument support and error management
---------
Co-authored-by: Teo <cheesen.xu@gmail.com>
* feat(models): add grok-3-mini support and update reasoning effort logic in SettingsTab and OpenAIProvider
* feat(settings): update reasoning effort logic for Grok models and enhance localization in multiple languages
* fix(models): correct spelling of reasoning in model support functions
* fix(settings): correct spelling of reasoning_effort in OpenAIProvider
- Introduced new models for the Qiniu provider in SYSTEM_MODELS.
- Updated migration logic to populate Qiniu provider models if they are empty during state initialization.
- Updated web search provider settings to include API key and free status indicators.
- Improved localization for English, Japanese, Russian, Chinese, and Taiwanese languages to reflect new API key and free status fields.
- Refactored web search provider management to prevent duplicates and streamline provider addition during state migration.
- Adjusted UI components to conditionally render based on provider type, enhancing user experience.
* feat(websearch): implement search window functionality and enhance search service
* feat(DefaultProvider): integrate @mozilla/readability for improved content parsing
* Add LocalSearchProvider for web page scraping
AI: Change `provider` from private to protected in BaseWebSearchProvider and implement LocalSearchProvider for web searching with browser-based content extraction.
* Add web search provider management features
Implement addWebSearchProvider function to prevent duplicates,
automatically load default providers on initialization, fix
LocalSearchProvider implementation, and update local provider
identification logic.
* Improve web search with specialized search engine parsers
Add dedicated parsers for Google, Bing, and Baidu search results,
replacing the generic URL extraction approach. Enhance page loading
with proper wait mechanisms and window cleanup. Remove DuckDuckGo
provider as it's no longer supported.
* Simplify DefaultProvider to unimplemented placeholder
* Remove default search engine from initial state
* Improve web search providers config and display
Add configuration for local search providers, remove empty apiKey fields,
and enhance the UI by sorting providers alphabetically and showing
whether they require an API key.
* Add stderr logging for MCP servers
* Make search window initially hidden
* feat(QuickPanel): enhance close action options and improve input handling
- Added 'enter_empty' as a new close action option for QuickPanel.
- Refactored input handling to include a delay before clearing search text after panel closure.
- Updated keyboard event handling to prevent default actions for specific keys.
- Improved styling for selected and focused states in QuickPanel components.
- Enhanced AttachmentPreview to utilize a separate FileNameRender component for better readability and functionality.
* feat(AttachmentPreview): enhance file icon rendering and styling
* feat(CustomTag): add closable functionality and improve styling
- Enhanced CustomTag component to support closable tags with an onClose callback.
- Updated styling for better visual integration and added hover effects for the close icon.
- Refactored usage of CustomTag in AttachmentPreview, KnowledgeBaseInput, and MentionModelsInput components for consistency.
* feat(SelectModelPopup, QuickPanel): update tag component and enhance search functionality
* feat(Inputbar, SettingsTab): add enable quick panel triggers setting and update translations
* feat(QuickPanel): integrate color library for dynamic styling and update package dependencies
* fix: 解决聊天页面图片复制失败的问题和点击编辑回复的时候，不显示图片url的问题
* fix: 解决chat模式，gemini-2.0-flash-exp-image-generation返回base64图片，无法复制的问题
* fix(MessageImage): Update the image copying feature to process base64 and URL formatted images based on their type
---------
Co-authored-by: magicdmer <magicdmer@163.com>
Co-authored-by: eeee0717 <chentao020717Work@outlook.com>
* feat: implement tool usage handling and system prompt building for AI providers
* refactor: streamline tool usage handling and remove unused code in OpenAIProvider and formats
* refactor: simplify tool usage handling in Anthropic and Gemini providers, and update prompt instructions
* refactor: remove unused function calling model checks and simplify MCP tools handling in Inputbar
* hidden tool use in message
* revert import
* Add idx parameter to parseAndCallTools for unique tool IDs
