Commit Graph
200 Commits
Author SHA1 Message Date
hy b52b603256 refactor: rename HumanDecision to HumanThoughtDecision and update related logic
- Updated the decision structure to use 'thought' instead of 'decision' in various functions and tests.
- Refactored decision-building functions to align with the new structure.
- Adjusted tests to accommodate the changes in decision structure, ensuring consistency across the codebase.
- Removed the delivery summary test as it was no longer relevant.
- Added new tests for handling references in the brain's next action.
- Introduced markdown reference parsing tests to validate reference handling.
2026-05-12 14:45:22 +08:00
hy f5b34986c7 refactor: rename workspace_run_command to terminal_run across the codebase
- Updated tool specifications to reflect the new terminal_run naming.
- Adjusted artifact extraction logic to accommodate the new tool name.
- Modified test cases to replace workspace_run_command with terminal_run.
- Refactored utility functions and constants to use terminal as the skill ID.
- Added new tests for terminal command policy and ensured inline interpreter commands are handled correctly.
- Ensured all references in the codebase are consistent with the new naming convention.
2026-05-11 18:10:01 +08:00
hy 05dfc8540b fix: update documentation and tests for docx batch write limits and constraints 2026-05-11 15:34:29 +08:00
hy 8362f20cdd fix: refine mouth prompt principles and simplify risk report test conditions 2026-05-09 19:03:46 +08:00
hy 2df77de3a2 fix: update user input text for risk report draft to remove specific mention of txt files 2026-05-09 18:44:22 +08:00
hy 02157d085c fix: update user input text for risk report draft to specify focus on txt files 2026-05-09 18:39:09 +08:00
hy 260e073678 fix: update user input text for risk report draft to improve clarity 2026-05-09 18:15:20 +08:00
hy cc46e4e2dc fix: update user input text for risk report draft and remove unused template references 2026-05-09 17:23:02 +08:00
hy 9a7bb6a341 Restore docx batch limit precheck and isolate watch hints 2026-05-09 16:57:10 +08:00
hy ef613616de feat: add docx batch halt reason and enhance handling of batch write limits
- Introduced `docxBatchHaltReasonSchema` to define various reasons for halting docx batch writes.
- Updated `humanWatchHintSchema` to include a new watch hint for halted batch writes.
- Enhanced replay tree logic to manage successor and predecessor nodes for improved document editing flow.
- Implemented stable locator resolution for table cells and paragraphs to ensure accurate updates during mixed operations.
- Added unit tests to validate behavior when exceeding batch write limits and handling halted operations.
- Improved error handling and signaling for follow-up actions in case of batch write interruptions.
2026-05-09 15:57:03 +08:00
hy 125343bc15 Refactor LLM config and model selection flow 2026-05-08 18:11:22 +08:00
hy 87e2b3d70d fix: correct user input text for risk report draft to improve clarity 2026-05-07 18:54:24 +08:00
hy 207b225c21 fix: update user input text for risk report draft to include all work document information 2026-05-07 18:47:26 +08:00
hy c1cc63f84d Fix draft artifact loading in risk report live test 2026-05-07 18:40:32 +08:00
hy f04285239a Normalize docx readback path matching 2026-05-07 17:29:24 +08:00
hy e6e2ba8e03 Update LangChain deps for provider cache observability 2026-05-06 15:58:45 +08:00
hy dbfb783320 feat(docx): enhance readback validation and add new tests for docx handling 2026-04-30 18:51:19 +08:00
hy b050ffb188 feat(docx): implement readback requirement for docx write operations
- Added a new watch hint 'docx_write_requires_readback' to enforce readback after docx write actions.
- Introduced utility functions to check if a docx write operation requires a readback.
- Updated the state transition logic to return the new watch hint when applicable.
- Enhanced the docx tool capabilities to support the new readback requirement.
- Created tests to validate the behavior of the new watch hint and ensure proper handling of docx write operations.
- Refactored docx review logic to hide deleted generated content and ensure accurate review outputs.
2026-04-30 18:45:55 +08:00
hy e73defd940 Revise live human risk report fixtures and docx tests 2026-04-30 16:59:50 +08:00
hy 4b448e168a Simplify live LLM config and remove search strategy prompt 2026-04-30 15:09:13 +08:00
hy ec0ff46b1e Align settings tests with PRD topology 2026-04-30 13:48:28 +08:00
hy fedaa129f3 Refactor test domains to remove 'workbench'; update related tests and error handling in model download page 2026-04-30 11:16:35 +08:00
hy 9e5d319f6d Add tests for workspace removal and scanning; remove obsolete test files
- Implement `workspace-remove-source-contract` test to verify that removing a workspace clears associated nodes and web sources.
- Add `workspace-scan-contract` test to ensure the workspace scanning API returns the correct progress payload.
- Delete unused `.gitkeep` files and obsolete test cases related to panel search and session persistence.
- Remove outdated tests for reference selection model and review navigation.
2026-04-30 10:30:20 +08:00
hy fbd08af1c1 Refactor workspace KB indexing for live human tests 2026-04-30 09:33:04 +08:00
hy 18c970154e Update AI test plan and mock-llm tests to reflect latest run results and improve structured fact exposure descriptions 2026-04-29 17:03:05 +08:00
hy 6f51d22c95 Refactor docx file paths and update test fixtures for workspace integration; add caching for KB fixture vectors 2026-04-29 16:37:52 +08:00
hy 47f93e03e3 Refine workspace fixtures and KB caching 2026-04-29 16:27:34 +08:00
hy 0d47c1282d Refactor learning tool tests to focus on docx skills and update expected outcomes
- Updated test cases in `learning-tool-budget-tracking.test.ts` to reflect changes in skill learning, focusing on `docx-read` and `docx-write`.
- Removed unnecessary learning decisions for `workspace-cli` in multiple test files, including `multi-tool-collaboration.test.ts`, `tool-failure-recovery.test.ts`, and `tool-limit-converge.test.ts`.
- Adjusted assertions in `watchhint-retry.test.ts` to validate learning outcomes related to `docx-read`.
- Introduced new workspace fixture files for enhanced testing scenarios.
- Updated `aiLiveTestData.ts` to streamline workspace paths and improve fixture management.
- Added new unit tests in `runtime-skill-contract.test.ts` to validate skill definitions and effective skill lists.
2026-04-29 15:51:13 +08:00
hy e53ced95b8 Refactor risk report live case for workspace evidence 2026-04-29 15:14:12 +08:00
hy 41348f7bc5 Refactor knowledge base skill names and descriptions for clarity; replace 'kb-document-discovery' and 'kb-evidence-query' with 'kb-retrieval' across documentation and codebase. 2026-04-28 18:08:36 +08:00
hy 9beb579ce2 Refine user input for comparison draft to improve clarity and focus 2026-04-28 16:02:22 +08:00
hy 4bcccccf1d Raise default human tool limit and relax mouth gate 2026-04-28 15:51:05 +08:00
hy 6042c513d7 Remove brittle prompt snapshot tests 2026-04-28 10:54:46 +08:00
hy 6f46f30ced Decouple artifact names and tighten mouth output rules 2026-04-28 10:37:57 +08:00
hy 22548f751c Refactor DOCX read and write tools to support read sessions and improve caching
- Introduced DocxReadSession and DocxReadSessionCache types for better session management.
- Updated fetchOutline and fetchBlocksByRange functions to utilize caching and session parameters.
- Modified existing functions to ensure they work with the new session structure.
- Enhanced error handling and response structures in worker requests.
- Added tests to verify that read operations do not create replay sessions and that editing after reading upgrades to replay sessions.
- Ensured that long documents maintain their tail blocks during session upgrades.
2026-04-27 18:38:04 +08:00
hy c3925ad12f Enhance event logging and artifact collection with improved runtime event handling and file reading capabilities 2026-04-27 17:27:04 +08:00
hy d72ddc7656 Refine delivery guidance handoff for mouth 2026-04-27 15:13:35 +08:00
hy 22ae0a783a Refactor constants and test data for improved clarity and consistency 2026-04-27 13:47:56 +08:00
hy 9752ba0c43 Support create-target docx reviews 2026-04-24 18:51:32 +08:00
hy 370579b3ab Add docx_create and update cross-source summary live case 2026-04-24 16:43:59 +08:00
hy 2971f00749 Refine structured output fallback classification 2026-04-24 14:33:29 +08:00
hy 433278e9be 需要对齐的只有 doc/test-plan/ai-test.md,我已经补了这 4 个正式 case 的登记;其余 docx 设计文档和 2026-04-23 18:55:40 +08:00
hy 1e2009387d feat: enhance docx handling with style text support and review highlights 2026-04-23 17:41:08 +08:00
hy 7815a9fb21 refactor: update HumanStableMemory checklist structure and related logic
- Changed checklist from a string to a structured array of objects with item and status fields.
- Updated related interfaces, schemas, and types to reflect the new checklist structure.
- Removed readyToDeliver field from HumanDecision and related logic, replacing it with checklist validation.
- Adjusted tests to accommodate the new checklist format and removed references to readyToDeliver.
- Enhanced prompt and documentation to clarify the new checklist requirements and usage.
2026-04-23 16:34:16 +08:00
hy e1715700b1 Document brain/mouth history split and positioned fallback 2026-04-23 14:20:16 +08:00
hy cd5624f364 Refine mouth prompt and simplify risk report live test 2026-04-22 19:05:48 +08:00
hy e0e924dfff Refactor file reviews and mouth runtime inputs 2026-04-22 18:31:16 +08:00
hy c2ef5e32dd Refactor AI document workflow 2026-04-22 18:21:44 +08:00
hy fc04dbdadf Add cross-source summary live case and retune KB prompt 2026-04-22 16:14:04 +08:00
hy 1cb4911acc Rename docx read tool and skill catalog prompt 2026-04-22 15:21:50 +08:00