- Replace confusing "Current ask promise was ignored" errors with more descriptive messages
- These errors occur during normal operation when handling partial messages
- The errors are caught by callers but the messages were confusing AI models
- Fixes issue #7312 where gpt-oss-120b model was confused by these error messages
When a subtask completes and the parent task resumes, set skipPrevResponseIdOnce = true
to ensure the next API call sends the full conversation including the subtask result.
This fixes the issue where GPT-5 would only receive the new message without the subtask
result context when using previous_response_id.
Fixes#7251
- Remove assistantMessageParser from experiment definitions in packages/types
- Remove ASSISTANT_MESSAGE_PARSER from shared experiments configuration
- Update Task.ts to always initialize and use AssistantMessageParser
- Remove conditional logic that checked for experiment flag
- Remove unused parseAssistantMessage import
- Update test files to remove assistantMessageParser references
- All tests passing successfully
Co-authored-by: Roo Code <roomote@roocode.com>
* feat: add optional todos parameter to new_task tool with experimental setting (#6329)
- Add optional todos parameter to new_task tool for hierarchical task planning
- Implement experimental setting to optionally require todos parameter
- Add clean state-based UI rendering to avoid spurious messages
- Export and reuse parseMarkdownChecklist function
- Add comprehensive test coverage for both optional and required modes
- Maintain full backward compatibility (todos optional by default)
* fix: update new_task tool example to include todos parameter
- Updated the example in tool-use.ts to show the todos parameter
- This prevents AI confusion about whether todos is a valid parameter
- The example now demonstrates the complete tool usage pattern
* fix: make new_task tool definition dynamic based on experimental setting
- Tool description now changes based on newTaskRequireTodos setting
- When disabled: shows todos as (optional)
- When enabled: shows todos as (required) with no mention of configuration
- Added tests to verify dynamic behavior
- Ensures AI models get unambiguous instructions based on current settings
* fix: add translations for newTaskRequireTodos experimental setting
- Added translations for all 17 supported languages
- Ensures consistent UI experience across all locales
* test: update snapshots for new_task tool example with todos parameter
- Updated 13 snapshot files to reflect the new tool-use example
- All tests now pass with the updated example format
* Update src/core/tools/newTaskTool.ts
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
* fix: address PR review comments
- Replace any[] with TodoItem[] type in ExtensionStateContext.tsx for better type safety
- Remove redundant initialTodos parameter from startTask call in Task.ts (todos already set in constructor)
- Improve code clarity in newTaskTool.ts by checking provider reference early and reusing state
# Conflicts:
# src/core/task/Task.ts
# webview-ui/src/context/ExtensionStateContext.tsx
* fix: revert order of operations in startTask to fix integration test timeout
The change in order of operations (calling say() before postStateToWebview()) was causing the XML content file test to timeout. Reverting to the original order fixes the issue.
* fix: hide todos parameter from new_task tool prompt when experiment is disabled
- Modified getNewTaskDescription to completely omit todos parameter when experiment is off
- Updated tests to verify todos parameter is not shown at all when disabled
- Ensures tool prompt remains unchanged when experimental setting is disabled
- Maintains backward compatibility while providing cleaner prompt interface
* fix: update snapshots for new_task tool todos parameter
- Updated snapshots in add-custom-instructions.spec.ts
- Updated snapshots in system-prompt.spec.ts
- All tests now passing with the new todos parameter documentation
* feat: move newTaskRequireTodos from experimental to VSCode settings
- Added newTaskRequireTodos as a VSCode configuration property in src/package.json
- Added description in src/package.nls.json
- Updated newTaskTool.ts to read from VSCode configuration instead of experiments
- Removed NEW_TASK_REQUIRE_TODOS from experimental settings in src/shared/experiments.ts
- Removed newTaskRequireTodos from packages/types/src/experiment.ts
- Updated tests to use VSCode configuration mocking instead of experiments
- Removed references from experiments test file
- Maintains backward compatibility (defaults to false)
* fix: make new_task tool description dynamically reflect VSCode setting
- Updated new-task.ts to check args.settings instead of args.experiments
- Added newTaskRequireTodos to SystemPromptSettings interface
- Pass newTaskRequireTodos setting through Task.ts and generateSystemPrompt.ts
- Updated all related tests to use settings instead of experiments
- Fixed TypeScript errors in test files by adding newTaskRequireTodos property
This ensures the tool description correctly shows todos parameter as required/optional
based on the VSCode setting value, fixing the issue where Roo would try to use
new_task without the todos parameter when it was required.
---------
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Merge Resolver <merge-resolver@roo-code.com>
Co-authored-by: Roo Code <roomote@roocode.com>
* fix: add explicit max_output_tokens for GPT-5 Responses API
- Added max_output_tokens parameter to GPT-5 request body using model.maxTokens
- This prevents GPT-5 from defaulting to very large token limits (e.g., 120k)
- Updated tests to expect max_output_tokens in GPT-5 request bodies
- Fixed test for handling unhandled stream events by properly mocking SDK fallback
* fix: add missing translations for reasoningEffort.minimal in Indonesian and Dutch locales
* fix: correct GPT-5 response ID persistence and usage
- Renamed metadata field from 'previous_response_id' to 'response_id' for clarity
- Fixed logic to correctly use the response_id from the previous message as previous_response_id for the next request
- This resolves the 'Previous response with id not found' errors that occurred after multiple turns in the same session
* feat: add robust error handling for GPT-5 previous_response_id failures
- Automatically retry without previous_response_id when it's not found (400 error)
- Clear stored lastResponseId to prevent reusing stale IDs
- Handle errors in both SDK and SSE fallback paths
- Log warnings when retrying to help with debugging
* fix: handle GPT-5 response ID race condition with nano model
- Add promise-based synchronization for response ID persistence
- Wait for pending response ID from previous request before using it
- Resolve promise when response ID is received or cleared
- Add 100ms timeout to avoid blocking too long on ID resolution
- Properly clean up resolver on errors to prevent memory leaks
This fixes the race condition where fast nano model responses could cause
the next request to be initiated before the response ID was fully persisted.
* fix: address PR review comments for GPT-5 implementation
- Extract usage normalization helper to reduce duplication
- Suppress conversation continuity for first message (but respect explicit metadata)
- Deduplicate response ID resolver logic
- Remove dead enableGpt5ReasoningSummary option references
- DRY up GPT-5 event/usage handling with normalizeGpt5Usage helper
- Centralize default GPT-5 reasoning effort using model info
- Fix Indonesian locale minimal string misplacement
- Add clarifying comments for Developer prefix usage
- Add TODO for future verbosity UI capability gating
- Fix failing test in reasoning.spec.ts
* fix(openai-native): address Roomote inline feedback\n\n- Delegate standard GPT-5 SSE event types to shared processor to reduce duplication\n- Add JSDoc for response ID accessors\n- Standardize key error messages for GPT-5 Responses API fallback\n- Extract persistGpt5Metadata() in Task to simplify metadata writes\n- Add malformed JSON SSE parsing test\n
* fix(openai-native,gpt5): correct usage cost calc (use calculateApiCostOpenAI incl. cache); enforce 'skip once' continuity via suppressPreviousResponseId; dedupe responseId resolver on SSE 400; feat: gate reasoning.summary by enableGpt5ReasoningSummary; centralize default reasoning effort; types/ui: add ModelInfo.supportsVerbosity and gate Verbosity UI by capability; refactor: avoid duplicate usage emission in SSE done/completed
* fix(gpt5): default enableGpt5ReasoningSummary=true to preserve tests and expected behavior
* fix(gpt5): canonicalize GPT-5 metadata key to previous_response_id and align enableGpt5ReasoningSummary default docs
* fix(openai-native): remove review artifact comments and guard GPT-5 in completePrompt
* feat: Before requesting, ensure checkpoint is initialized
* Generate a checkpoint before modifying the code
* refactor: streamline checkpoint handling and enhance getCheckpoints method
* Blocked waiting for checkpoint initialization timing to change
* cancel checkpoint restore limit
* fix: ensure checkpoint service is undefined on initialization error and improve checkpoint diff handling
* refactor: simplify checkpoint service initialization and cleanup unused variables in CheckpointMenu
* fix: prevent race condition in checkpoint service initialization
- Only assign service to cline.checkpointService after successful initialization
- Add proper cleanup on initialization failure
- Prevents service from being in inconsistent state if Git check fails
* fix: remove checkpoint save from presentAssistantMessage for update_todo_list case
---------
Co-authored-by: Daniel Riccio <ricciodaniel98@gmail.com>
* fix: detect Claude models by name for API protocol selection
- Modified getApiProtocol to accept modelId parameter
- Added check for 'claude' in model name (case-insensitive)
- Updated Task.ts to pass model ID to getApiProtocol
- Added comprehensive tests for the new logic
Fixes#5830
* fix: limit Claude model detection to vertex and bedrock providers only
- Modified getApiProtocol to only detect Claude models by name when provider is vertex or bedrock
- Added comprehensive unit tests for getApiProtocol function as requested in PR review
- This ensures Claude models are only auto-detected for providers that need it
* feat: add BatchDiffApproval component for multi-file diff application
- Introduced a new component `BatchDiffApproval` to handle the approval of batch changes across multiple files.
- Integrated the `BatchDiffApproval` component into `ChatRow` to display batch diff requests.
- Updated experimental settings to include a toggle for multi-file apply diff functionality.
- Enhanced localization files to support new strings related to batch changes in multiple languages.
- Updated tests to cover the new multi-file apply diff feature.
* revert this
* fix: update applyDiff parameter type to accept string or DiffItem
* refactor: keep original file name for apply diff tool
* revert this
* Update src/core/webview/__tests__/ClineProvider.test.ts
* revert this
* fix: keep the original path if the experiment is disabled
* test: add dynamic strategy selection tests for MultiSearchReplaceDiffStrategy and MultiFileSearchReplaceDiffStrategy
* fix: mock applyDiffTool module and ensure legacy tool resolves successfully in tests
* remove this
* ellipsis suggestion
Co-authored-by: ellipsis-dev[bot] <65095814+ellipsis-dev[bot]@users.noreply.github.com>
* refactor: mirror concurrent file reads
---------
Co-authored-by: Daniel Riccio <ricciodaniel98@gmail.com>
Co-authored-by: Daniel <57051444+daniel-lxs@users.noreply.github.com>
Co-authored-by: ellipsis-dev[bot] <65095814+ellipsis-dev[bot]@users.noreply.github.com>
