- Added AppErrorLogService to handle error logging with file management and redaction of sensitive information.
- Integrated error logging into various services including EmbeddingRuntimeClient, StorageService, and ChatRunService.
- Enhanced IPC handlers to support error log directory access and renderer error logging.
- Updated settings panel to allow users to open the error log directory.
- Added unit and end-to-end tests for error logging functionality and migration repairs.
- Updated llm-executor-text tests to simplify transport handling and improve model invocation logic.
- Revised model-catalog tests to enhance clarity in filtering models and preserving modalities.
- Adjusted structured-chat-activity tests to remove unnecessary method assertions and streamline output validation.
- Refined structured-output-auto-fallback tests to focus on single path execution and error handling.
- Enhanced structured-output-fallback tests to improve error classification and handling of various API responses.
- Updated sqlite-storage tests to include a new migration for dropping LLM strategy preferences.
- Removed llmStrategyPreferenceStore from realAiDepsFactory to streamline dependency injection.
- Replaced `resolveWorkspacePath` with `resolveToolPath` in various services to streamline path resolution.
- Updated error messages for invalid paths and workspace boundary checks.
- Enhanced attachment service to create source attachments directly from file paths without copying to runtime bucket.
- Removed unnecessary checks and legacy code related to workspace boundary enforcement.
- Improved handling of dropped and clipboard attachments in the workbench composer.
- Adjusted tests to reflect changes in attachment handling and workspace boundary logic.
- Introduced AiExecutionPermission type to manage execution permissions.
- Updated AiRunStore to include execution_permission in the database schema.
- Implemented execution permission handling in the WorkbenchComposer component.
- Enhanced ChatPage to manage execution permission state and persistence.
- Added UI elements for selecting execution permissions in the workbench.
- Updated tests to cover new execution permission functionality and its effects on AI run behavior.
- Introduced new types for ChatSession UI state management including ChatSessionUiStateV1 and related structures.
- Implemented functionality to save and retrieve UI state for chat sessions.
- Added tests for structured output fallback scenarios in the AI human live backend.
- Enhanced unit tests for LLM executor to validate strategy replacement and persistence across runs.
- Created end-to-end tests to ensure session UI state isolation and persistence after application restart.
- Updated llmConfig to load structured fallback configurations from environment variables.
- Updated PPT tool specifications to replace output schemas with more specific ones.
- Changed warning properties to issues in PPT-related types.
- Removed warnings from PDF text normalization and related types.
- Adjusted tests to reflect the removal of warnings and ensure proper handling of tool outputs.
- Enhanced schema rendering to exclude runtime constraints while preserving TypeScript shapes.
hy