- Redesign API access format cards into full-width horizontal rows, one per line
- Enforce strictly equal heights between control and API access panels
- Keep control action buttons in a single row without wrapping at 1247x703
- Tighten spacing between brand heading and URL input while keeping `http://` prefixes vertically aligned
- Unify card surfaces with a clean, bright, high-contrast theme across light and dark modes
- Enable click-to-copy directly on URL code blocks with visual feedback indicators
- Add separate configuration entries for IDE and CLI
- Integrate Gemini-compatible routing and scoped credentials
- Support detection, launch, backup, restore, and IDE restart
- Add localized guidance and configuration, UI, and native integration tests
- Support installation detection, configuration, backup, restore, launch, and restart
- Handle both models.json formats while preserving custom settings
- Simplify model names and retain existing CPA models when switching
- Fix model visibility restoration when disabling integration
- Add regression tests, translations, and documentation
- Protect selected models' aliases when generating exclusions
- Reconcile exclusions when editing models and saving mappings
- Preserve manual exclusions and provider disabled state
- Add unit and UI regression tests for alias conflicts
- Update font stacks with locale-aware fallbacks
- Standardize font sizes, weights, and line heights
- Improve text hierarchy in credential cards, dialogs, and usage tables
- Display credentials in responsive cards with quota progress bars
- Add request statistics and detailed health and cooldown information
- Consolidate nine credential settings into a dialog with partial updates
- Reuse Codex quota reset logic and confirmation across pages
- Move notes above card actions and hide the healthy status indicator
- Add translations and regression coverage
- Prevent dropped scroll updates and position jumps
- Keep scroll ranges aligned when resizing or toggling columns
- Update logging guidance across all supported languages
- Add browser regression coverage for table scrolling
- Support Simplified Chinese, Traditional Chinese, English, and Japanese
- Follow the UI language and show a notice when notes are unavailable
- Avoid language and content fallbacks
- Generate Release bodies and update manifests from per-version Markdown
- Add collapsible notes, publication dates, and links to full releases
- Test publishing, legacy manifest compatibility, and localized rendering
Keep same-source alias entries and their metadata when rebuilding provider models, preventing API connection edits from deleting aliases created in Advanced Settings.
Fixes#271
- Add editable model lists, alias suggestions, and usage guidance
- Ignore unselected rows and validate alias conflicts and naming rules
- Fix alias routing through disabled or excluded providers
- Preserve model routes when restoring legacy backups
- Repair legacy metadata and remove unsupported context fields
- Add regression tests for validation and configuration recovery
- Add Clear integration to configuration management
- Guide users without OAuth credentials through clearing integration and signing in
- Separate apply and close actions, showing Close only when configuration is applied
- Preserve credentials and original settings across switches
- Restore the CPA model catalog when reconnecting
- Add configuration switching and UI regression tests
- Add Devin login to standard and simplified modes
- Support credentials, model aliases, and exclusions
- Display daily and weekly quotas with reset times
- Validate callbacks and prevent stale login updates
- Add translations, theme support, and integration tests
- Remove empty feedback placeholders and fixed-height rows
- Place configuration status beside the save button
- Compact save controls in the simplified view
- Add responsive layout and notification stability regression tests
- Move operation feedback into a shared floating stack to prevent layout shifts
- Allow identical errors to reappear after dismissal
- Support keyboard navigation, scrolling, and focus restoration in dialogs
- Add notification interaction regression tests
- Remove code comments throughout the project
- Align TPS and token normalization with Keeper
- Add Redis usage queue batch pulling with HTTP fallback
- Prevent cache and invalid token double counting
- Refresh usage details immediately after new records arrive
- Separate selected and unselected models with independent search and bulk actions
- Fix selection initialization and ignore stale discovery responses
- Improve keyboard navigation and responsive layouts
- Update translations and add model selection UI regression tests
- require explicit model discovery before saving DeepSeek connections
- preserve selected models when refreshing the catalog
- mark model lists stale when connection details change
- add model refresh support to easy mode
- apply built-in DeepSeek reasoning levels consistently