Share content-addressed parsing and embedding caches across isolated test runs, fingerprint relevant dependencies and model assets, and remove implicit UI warm-up. Add realistic two-round Skill delivery evaluation and cache regressions.
- Implement unit tests for workspace index job store contract to validate job creation, active job retrieval, progress updates, and failure handling.
- Create unit tests for workspace index worker contract to ensure embedding batch limits and file processing logic are correctly enforced.
- Add tests for interrupted job recovery to verify that running jobs and file indexing statuses are correctly marked as failed upon recovery.
- Introduce tests for workspace preparation contract to confirm that only indexable and failed files are processed during updates.
- Develop tests for workspace scanning to ensure file index records are generated correctly and hidden files are filtered out.
- Add end-to-end tests for sidebar app update functionality to validate update visibility, download progress, and error handling.
- Updated multiple e2e test files to replace 'launchMode: real' with 'serviceMode: real' for consistency in service mode configuration.
- Refactored the electronApp.ts to introduce E2EServiceMode and related types for better clarity and maintainability.
- Enhanced run-e2e-case.ts to accommodate the new service mode structure, ensuring proper handling of test cases.
- Created workspaceFixture.ts to manage temporary workspaces more efficiently during tests.
- Deleted the `missing-llm-notification-startup.e2e.test.ts` file as it is no longer needed.
- Enhanced `send-guard-mode-switch-and-preferences.e2e.test.ts` to include checks for missing model notifications.
- Removed `browser-use-deepseek-financing-search.e2e.test.ts` due to redundancy.
- Cleaned up `chat-run-service-contract.test.ts` by removing performance tests that are no longer relevant.
- Updated `aiLiveTestData.ts` to remove unused constants and types related to risk reports.
- Removed `aiToolHarness.ts` function for writing sample documents as it is no longer utilized.
- Updated `llmConfig.ts` to infer LLM provider kind from environment variables.
- Cleaned up `cache-summary.ts` by removing unused JSON conversion functions.
- Deleted `browserLlmConfig.ts` as it was redundant.
- Enhanced `settings.ts` to support new application settings and improved HTTP server response handling.
- Removed unused functions from `workbench.ts` related to waiting for assistant answers.
- Renamed constants and types related to the workspace index worker to reflect its new role as a prepare worker.
- Removed the old workspace index worker implementation and created a new prepare worker implementation.
- Updated the client for the prepare worker to handle requests and responses appropriately.
- Adjusted tests to accommodate the changes in worker functionality and ensure proper handling of document preparation.
- Introduced new tests for embedding runtime to ensure serial execution of inference requests.
- Updated user-message-actions.e2e.test.ts to replace local OpenAI server setup with live model configuration through UI.
- Refactored reference-token-lifecycle.e2e.test.ts to streamline workspace preparation using new utility functions.
- Simplified multi-file-apply-tabs.e2e.test.ts by removing unnecessary workspace indexing logic.
- Enhanced file-preview-actions.e2e.test.ts to utilize new workspace preparation methods.
- Modified run-navigation-snapshot-replay.e2e.test.ts to adopt temporary workspace setup and removed redundant workspace creation logic.
- Updated settings.ts to consolidate workspace creation and binding logic, improving maintainability.
- Removed deprecated methods for creating real workspaces and replaced them with UI-driven approaches.
huangyuan