- Changed text color for decision layer (Thought) and response content to normal body color (#37352F) for better readability.
- Updated various components in the AssistantTurnCard to reflect the new color scheme, ensuring consistency across the UI.
- Adjusted styles in markdown rendering to maintain uniformity in text presentation.
- Enhanced visual feedback for process statuses, ensuring failed states are clearly indicated while maintaining a cohesive design.
- Revised the `queue-message-during-run` test case to clarify button behavior during message queuing.
- Enhanced the `AssistantTurnCard` component to display AI and human review results distinctly.
- Modified the `WorkbenchComposer` to streamline button states and tooltips based on current actions.
- Updated message log projection to include terminal review decisions.
- Adjusted types to accommodate new terminal review decision states.
- Improved utility functions for handling terminal review results.
- Refined end-to-end tests for AI question flow and approval batch flow to ensure accurate UI behavior and state management.
- Added support for terminal review events including `terminal_review_started`, `terminal_review_resolved`, and `terminal_review_blocked`.
- Enhanced `projectProcessGroups` to handle terminal review states and update process items accordingly.
- Introduced `removePendingTerminalReviewItems` utility to filter out items based on their terminal review state.
- Updated types to include `terminalReviewState` in `AssistantProcessGroupItem`.
- Modified event handling in `reduceAssistantProcessGroups` to accommodate new terminal review events.
- Enhanced tests to cover terminal review scenarios and ensure correct state transitions.
- Updated the terminal review prompt to clarify decision-making criteria for commands.
- Improved the formatting of run duration display to handle cases with zero minutes.
- Refactored E2E tests for AI terminal review to use more descriptive test names and ensure clarity in approval scenarios.
- Enhanced E2E tests for AI question flow to validate partner switching and state restoration after refresh.
- Added observation timeout for approval batch flow tests to ensure proper handling of approval states.
- Adjusted the logic for reading approval waiting snapshots to include new UI elements for guidance.
- Improved the validation of duration formats in process tests to ensure consistency in displayed time.
- Refined the locator logic for waiting on the latest assistant process to avoid conflicts with item and toggle prefixes.
- Renamed TerminalReviewPromptOperation to TerminalReviewPromptCommand and updated its properties.
- Simplified the terminal review system prompt for clarity and conciseness.
- Modified rendering functions to accommodate new command structure.
- Updated types to reflect changes in terminal review decisions from operations to commands.
- Enhanced AssistantTurnCard to display run duration and manage process expansion state.
- Added handling for run start and finish timestamps in ChatPage.
- Adjusted tests to align with the new command-based terminal review structure.
- Improved UI tests to verify assistance badge visibility and process duration display.
- Updated terminal command execution policies to allow auto-approval under specific conditions.
- Revised AI test cases to validate the new terminal review and approval mechanisms.
- Introduced a new prompt structure for terminal command reviews, ensuring clarity in decision-making.
- Enhanced the handling of terminal command operations, including better tracking of operation IDs.
- Improved test utilities to facilitate the collection of terminal review decisions.
- Adjusted end-to-end tests to verify the correct behavior of the approval batch flow in the workbench.
- Ensured that the terminal's working directory is correctly set to the user's home directory.
- Implemented `stripHereDocumentBodies` function to handle here-documents in terminal commands.
- Updated `analyzeTerminalRiskPolicy` to auto-allow empty commands and improve command parsing.
- Added new annotation type for terminal review resolution in shared types.
- Created comprehensive tests for terminal command review processes, including auto-approval and human review escalation scenarios.
- Refined approval batch handling in tests to ensure accurate decision-making and event tracking.
- Improved handling of dangerous commands and refined risk analysis logic in terminal command policies.
- Enhanced e2e tests for workbench run scenarios to validate approval workflows and execution permissions.
framenavigated reinstalled IPC listeners on the same document, so each
workspaceDirectoryList was counted twice and workspace-prepare-refresh
failed AC7 despite a single product refresh.
Graceful app.exit already waits briefly; calling electronApp.close() afterward
can hang for 15s+. Force-kill immediately when the process is still alive, and
cover the SIGKILL allowlist in unit tests.
prefix-cache-retain-live assertions could pass then hang ~45m in
electronApp.close during teardown. Race close with a 15s timeout and
SIGKILL the process so release:start live gate can finish.
Thought markdown must not use delivery linkAccess sidecar; that marked
in-thought file refs unavailable (no icon, not clickable). Pass only
onOpenReference so refs stay openable like streaming.
- Changed the structure of AskUserQuestionAnswer to use questionId, selectedOptionIds, and userText instead of id, selected, and custom.
- Updated the HumanInteractionCard component to reflect the new answer structure, including handling user text input.
- Modified the AssistantTurnCard and message log projection to accommodate the new answer format.
- Adjusted tests to ensure compatibility with the updated schema and answer structure.
- Cleaned up code for better readability and consistency across components.
- Enhanced HumanInteractionCard to support partner name and cancellation state.
- Introduced WorkbenchInteractionCard and WorkbenchInteractionFooter for better code organization and reusability.
- Updated placeholder functions for partner guides.
- Modified types to include 'cancelled' state for AssistantTurnViewModel.
- Adjusted utility functions to handle run cancellation events.
- Updated ChatPage to utilize new components and handle approval interactions more effectively.
- Improved e2e tests to validate cancellation behavior and UI state.
- Implemented resolveQuestionInteraction utility to extract interaction details from assistant process items.
- Updated AssistantActivityItemView to utilize the new interaction resolution logic.
- Refactored messageLogProjection to ensure thoughts are only created when thinking is open.
- Enhanced reduceAssistantProcessGroups to handle human interaction events and apply question interaction results.
- Added applyQuestionInteractionResult to manage answers from resolved interactions.
- Updated tests to cover new interaction handling and ensure correct behavior across various scenarios.
- Improved e2e tests for scheduled tasks to validate interaction flows and partner management.
hy