* fix(Markdown): eliminate hydration error from image `<div>` nested in `<p>`
Signed-off-by: Chan Lee <Leetimemp@gmail.com>
* feat: add support for reading local files in binary format
Signed-off-by: Chan Lee <Leetimemp@gmail.com>
* refactor(ImageViewer): Consolidate image rendering for unified display and context menu
Signed-off-by: Chan Lee <Leetimemp@gmail.com>
---------
Signed-off-by: Chan Lee <Leetimemp@gmail.com>
* feat: add update channel functionality for beta testing
- Introduced a new IPC channel for setting the update channel.
- Implemented logic in AppUpdater to handle update channel changes.
- Updated settings to include a beta testing toggle, allowing users to switch between stable and beta update channels.
- Enhanced the settings UI to reflect the new beta testing option.
* add i18n
* update i18n
* update i18n
* refactor: rename update channel to feed URL and update related functionality
- Changed IPC channel from App_SetUpdateChannel to App_SetFeedUrl.
- Updated AppUpdater to set feed URL instead of update channel.
- Modified preload and settings to reflect the new feed URL functionality.
- Added constants for production and early access feed URLs.
* refactor: remove setAutoUpdate method from API
- Eliminated the setAutoUpdate method from the API object in preload index, streamlining the IPC communication interface.
* refactor: update early access feed URL and improve tooltip descriptions
- Changed EARLY_ACCESS_FEED_URL to point to the latest GitHub release.
- Simplified the setEarlyAccess function to directly set the feed URL.
- Added tooltips for early access settings in multiple languages to inform users about potential instability and the need for data backup.
* feat(migrate): add early access setting to state configuration
- Introduced a new state setting 'earlyAccess' and initialized it to false in the migration configuration.
* fix(i18n): update early access tooltip translations for clarity
- Revised the tooltip descriptions for the early access feature in English, Simplified Chinese, and Traditional Chinese to enhance clarity and ensure consistency in messaging regarding potential instability and the importance of data backup.
* feat: introduce FeedUrl enum for centralized feed URL management
- Added a new enum `FeedUrl` in the constants file to define production and early access feed URLs.
- Updated relevant IPC handlers and services to utilize the `FeedUrl` enum for type safety and consistency.
- Refactored the configuration manager to include methods for getting and setting the feed URL using the new enum.
* feat(settings): initialize early access and auto-update settings in AboutSettings component
- Added initialization for early access and auto-check update settings in the AboutSettings component to enhance user configuration options.
---------
Co-authored-by: beyondkmp <beyondkmkp@gmail.com>
fix: enhance OpenAI stream handling and error management
- Updated the `openAIChunkToTextDelta` function to include error handling with a try-catch block, improving robustness during stream processing.
- Refined the `readableStreamAsyncIterable` function to ensure proper handling of stream completion and errors, including a return method for cleanup.
- Adjusted type definitions for better clarity and consistency in the handling of async iterables.
This commit introduces support for displaying and downloading
inline base64-encoded images (specifically PNG and JPEG formats)
within Markdown content.
Key changes:
- Modified 'urlTransform' in the Markdown component to allow 'data:image/png'
and 'data:image/jpeg' URLs, enabling their rendering.
- Updated the 'download' utility to handle 'data:' URLs,
allowing users to save these inline images.
Signed-off-by: Chan Lee <Leetimemp@gmail.com>
This update modifies the OpenAIProvider to yield tool_calls only if they exist and have a length greater than zero, improving the handling of delta content. Additionally, a minor cleanup was performed by removing an unnecessary blank line in the code.
This commit updates the AssistantsTab component to only display group titles for tagged assistants, excluding the 'untagged' category. This change enhances the UI by reducing clutter and improving clarity in the display of assistant groups.
* chore: update OpenAI package to version 5.1.0 and adjust related patches
- Updated OpenAI dependency from version 4.96.0 to 5.1.0 in package.json and yarn.lock.
- Removed obsolete patch for OpenAI 4.96.0 and added new patch for OpenAI 5.1.0.
- Adjusted types for image handling in OpenAIResponseProvider to use Uploadable instead of FileLike.
- Minor code refactoring for better clarity and maintainability.
* refactor(OpenAIResponseProvider): remove logging for image generation process
This commit restructures the localization strings for Notion settings in the Japanese language file, moving them from a nested structure to a more accessible format. This change improves clarity and maintainability of the localization data.
* Add support for TokenFlux image generation service
This commit integrates TokenFlux as a new painting provider with dynamic
form generation based on model schemas, real-time generation polling,
and full painting history management.
Key features:
- Dynamic form rendering from JSON schema input parameters
- Model selection with pricing display
- Real-time generation status polling
- Integration with existing painting workflow and file management
- Provider-specific painting state management
The implementation follows existing patterns from other painting pages
while adding TokenFlux-specific functionality like schema-based form
generation and asynchronous polling for generation results.
* Add image upload support and comparison view to TokenFlux
Implements file upload handling for image parameters with base64
conversion, random seed generation, and side-by-side comparison
layout when input images are present.
* Refactor TokenFlux to use service class and components
Extract form rendering logic to DynamicFormRender component and API
logic to TokenFluxService class. Simplifies the main component by
removing duplicate code for model fetching, image generation polling,
and form field rendering.
* Refactor TokenFlux to fix state management and polling
- Change painting field from modelId to model for consistency
- Fix updatePaintingState to use functional state updates
- Add automatic polling for in-progress generations on mount
- Group models by provider in the selection dropdown
- Separate prompt from other input params in form data handling
- Improve error handling in the paintings store
* Auto-select first model when models are loaded
* Add image generation UI localization strings
Add translation keys for model selection, input parameters, image
labels, pricing display, and form validation across all supported
locales (en-us, ja-jp, ru-ru, zh-cn, zh-tw). Update TokenFluxPage
component to use localized strings instead of hardcoded English text.
* fix: Add a right border to the first child of the ImageComparisonSection
* style: Remove padding from UploadedImageContainer in TokenFluxPage
* feat: Implement caching for TokenFlux model fetching and update image upload handling
* feat: Enhance localization support by adding language context handling in TokenFluxPage
* refactor: Simplify layout structure in TokenFluxPage by removing unnecessary SectionGroup components and improving section title styling
---------
Co-authored-by: kangfenmao <kangfenmao@qq.com>
* feat(markdown-export): add option to show model name in export
* refactor(export): Refactor the Obsidian export modal to Ant Design style
* refactor(obsidian-export): export to obsidian using markdown interface & support COT
* feat(markdown-export): optimize COT export style, support export model & provider name
Add a new setting to toggle displaying the model provider alongside the
model name in markdown exports. Update the export logic to include the
provider name when enabled, improving context and clarity of exported
messages. Also fix invalid filename character removal regex for Mac.
* feat(export): add option to export reasoning in Joplin notes
Introduce a new setting to toggle exporting reasoning details when
exporting topics or messages to Joplin. Update the export function to
handle raw messages and convert them to markdown with or without
reasoning based on the setting. This improves the export feature by
allowing users to include more detailed context in their Joplin notes.
* feat(export): update i18n for new export options
* fix(settings): remove duplicate showModelNameInMarkdown state
* feat(export): add CoT export for notion & optmize notion export
* feat(export): update Notion settings i18n
* fix(utils): correct citation markdown formatting
Swap citation title and URL positions in markdown links to ensure
the link text displays the title (or URL if title is missing) and
the link points to the correct URL. This improves citation clarity.
purefkh