* feat: enhance rule file loading with .roo/rules directory support
- Introduced functions to safely read files and check for directory existence.
- Added capability to read all text files from a specified directory in alphabetical order.
- Updated `loadRuleFiles` to prioritize loading rules from a `.roo/rules/` directory, falling back to existing rule files if necessary.
- Enhanced `addCustomInstructions` to support loading mode-specific rules from a `.roo/rules-{mode}/` directory, improving flexibility in rule management.
This change improves the organization and retrieval of rule files, allowing for better modularity and maintainability.
* Updated strings and translations
* Add tests
* Revert changes to system prompt translations
* Fix path resolution
* Make instruction structure clearer
---------
Co-authored-by: Matt Rubens <mrubens@users.noreply.github.com>
* Remove unused imports
* Restore protected status for mcpHub property
* More mcpHub tidy-up
* Remove need for access to private "view" property
Increase type flexibility so we can directly use postMessageToWebview()
* Make ensureCacheDirectoryExists private again
expose new public writeDataToCache() method to commonize code.
also decided to leave updateApiConfiguration as public, since widely used in diverse ways.
* Match naming of readModelsFromCache.
Symmetric naming of functions that complement each other.
* Decision: Leave readModelsFromCache as public
* Make updateGlobalState and setGlobalState private again
Stop using deprecated (private) method. Use contextProxy set/getValue methods instead.
Utility functions provided for brevity.
* Add clarifying comment to explain presence of utility functions
* Prefer this/provider.log over outputChannel.appendLine
log is already public, outputChannel should be private.
But also prefer log internally for the additional value of console logging, plus code brevity.
* Lock down access to properties as far as reasonably possible
* PR feedback - don't relax type checks
---------
Co-authored-by: Matt Rubens <mrubens@users.noreply.github.com>
* feat: Add support for .roorules configuration files
- Add .roorules to the list of supported rule files
- Improve mode-specific rule file handling to prioritize .roorules over .clinerules
- Update file loading logic to track which rule file was successfully loaded
- Add test cases for .roorules functionality
This change maintains backward compatibility with existing .clinerules while
introducing support for the new .roorules format as the preferred configuration
method.
* refactor: simplify rule file loading by removing unused rules
- Removed .cursorrules and .windsurfrules from the list of rule files in loadRuleFiles function.
* refactor: update loadRuleFiles to return single rule file content
- Modified loadRuleFiles function to return content from the first available rule file instead of combining multiple rule files.
- Updated tests to reflect the new behavior of loading only the .roorules file content when available, ensuring clarity in rule file handling.
* fix: update prompts for deprecated rule file references
- Updated prompts in the English locale to reflect the deprecation of .clinerules in favor of .roorules.
- Added notes in the prompts indicating that .clinerules will stop working soon, ensuring users are aware of the upcoming changes.
* Translations
* Update links
* Revert README changes
* Add missing spans around links
---------
Co-authored-by: Matt Rubens <mrubens@users.noreply.github.com>
Track the current workspace path in history items to enable filtering and organization of history by workspace. This change adds the workspace property to the history schema and ensures it's saved with each history entry.
Problem:
Closing auxiliary windows (e.g., from "open in editor") could prematurely dispose the shared singleton McpHub instance, shutting down MCP servers unexpectedly while other panels (like the sidebar) might still be active.
Solution:
Implemented reference counting directly within the McpHub class to manage its own lifecycle based on active clients (ClineProvider instances).
- Added `refCount`, `registerClient()`, and `unregisterClient()` methods to McpHub.
- McpHub now disposes itself only when the last registered client unregisters (`refCount` reaches 0).
- ClineProvider instances now call `mcpHub.registerClient()` upon initialization and `mcpHub.unregisterClient()` upon disposal.
- Removed the direct `mcpHub.dispose()` call from ClineProvider.dispose.
Benefit:
This ensures the shared McpHub instance remains active as long as at least one ClineProvider instance is using it. Cleanup now correctly occurs only when the last provider is closed or during full extension deactivation. This centralizes the resource's lifecycle logic within the resource class itself.
Files Changed:
- src/services/mcp/McpHub.ts
- src/core/webview/ClineProvider.ts
* refactor: Change read_file result format to XML
Modifies how the result of the `read_file` tool is presented
in the conversation history sent to the AI model.
Previously, the format was:
[read_file for 'path/to/file'] Result:
{content_or_error}
This commit changes the format to use XML tags for better
structure and potentially easier parsing by the model:
<file>
<path>path/to/file</path>
<content>
{content_or_error}
</content>
</file>
This change only affects the `read_file` tool result formatting
within the user context message constructed in `src/core/Cline.ts`.
Other tool result formats remain unchanged.
* fix: Update error message for better clarity in multi-search-replace strategy
Refines the error message returned when no sufficiently similar match is found during the multi-search-replace operation. The message now includes a clearer instruction to use the read_file tool for obtaining the latest file content before attempting to apply the diff again.
* refactor: Update readFileTool to return results in XML format
Modifies the `readFileTool` function to format the output as XML, enhancing the structure of the returned file content. This change aligns with previous updates to ensure consistent result formatting across tools.
* refactor: Simplify result handling for tool responses in Cline
This update refines the handling of tool responses in the `Cline` class by removing the XML formatting for `read_file` results and consolidating the logic for pushing results to the user message content. The changes ensure that all tool results are processed uniformly, improving code clarity and maintainability.
* test: Update read_file tests to validate XML formatted results
This commit modifies the assertions in the `read_file` tool tests to check for the expected XML format in the results. The changes ensure that the output structure aligns with recent updates to the tool's response formatting, enhancing test accuracy and reliability.
* test: Refactor assertions in read_file tests to use expected XML variable
This commit updates the `read_file` tool tests to utilize a predefined variable for the expected XML output, improving readability and maintainability of the test code. The changes ensure consistency in the expected results across multiple test cases.
* fix: Improve handling of escaped markers in apply_diff
unescapeMarkers() is supposed to fix escaped merge conflict markers so
that they match the code. However the function has a bug which requires
SEARCH and REPLACE strings in the markers to be replaced. This is not
part of merge conflict markers, so many valid cases were previously
missed.
* Revert test changes
* Add new test
---------
Co-authored-by: Matt Rubens <mrubens@users.noreply.github.com>
Currently all errors in apply_diff diff structure are assumed to be
related to merge conflicts. This is unfortunate because it is way more
likely that LLM simply made an error structuring the diff. The LLM is
then directed to fix escaping in the diff from which it doesn't recover.
It is safe to assume that missing or extra "<<<<<<< SEARCH" or ">>>>>>>
REPLACE" strings are likely bad diff structure, because while "<<<<<<<"
and ">>>>>>>" are common merge conflict markers, no tools add "SEARCH"
and "REPLACE" strings after them. This is likely output of the LLM
itself.
LLM not recovering has been observed in Claude 3.5 Sonnet and Claude 3.7
Sonnet models.
To address the issue, error message now mentions merge conflict markers
and their escaping only if it's clear that errors may come from this
source. In the rest of cases the error message repeats the expected diff
structure back to the LLM.
The multi-search-replace diff strategy previously did not correctly strip line numbers (`number | `) from the SEARCH block when the REPLACE block was empty. This occurred because the condition for stripping required both blocks to consistently have line numbers.
This prevented successful deletion operations when the SEARCH block content was copied from `read_file` output (which includes line numbers) and the REPLACE block was empty.
This commit updates the line number stripping condition in `applyDiff` to also trigger if the SEARCH block has line numbers and the REPLACE block is empty or contains only whitespace, resolving the bug.
Additionally, a new test case has been added to `multi-search-replace.test.ts` to specifically verify this deletion scenario. All tests now pass with this updated logic.
* fix: remove -p flag from test script to prevent git operation errors
The -p flag in npm-run-all was causing tests to run in parallel, which led to 'Cannot log after tests are done' errors with git operations. These errors don't appear when running test:extension alone.
The issue occurs because git-based tests create temporary directories and run async operations that can interfere with each other when executed in parallel. Running tests sequentially resolves this cleanly.
While it might increase total test time slightly, it ensures more reliable and consistent test results.
* refactor(terminal): improve mock streams and fix test issues
- Create shell-specific mock streams (bash, cmd, pwsh) with proper line ending handling
- Fix open handles in tests by properly managing timeouts
- Standardize stderr redirection across all shell implementations using stdio option
- Improve test reliability and output cleanliness
* fix(tests): add skipVerification option to PowerShell tests to debug Linux issues
* fix(tests): use explicit variable name in PowerShell test to fix Linux compatibility
* Refactor terminal tests to use purpose-based approach instead of command mapping
* Remove reference to non-existent test file
* fix: use printf instead of echo -e for more consistent behavior across platforms
* fix: use single quotes for PowerShell commands to preserve variables on Linux
* Update code-qa workflow to run tests on both Windows and Ubuntu
* fix: use platform-specific PowerShell command execution for Linux and Windows
* Fix toggleToolAlwaysAllow to handle path normalization for cross-platform compatibility
* Fix McpHub tests to handle normalized paths on Windows
* Suppress console.error messages in McpHub tests
* fix: make Bedrock ARN regex patterns Windows-compatible
Fixed an issue where AWS Bedrock tests were timing out on Windows but passing on Linux. The root cause was path separator handling in regex patterns used for model ID extraction from ARNs.
1. Updated model ID extraction regex to handle both forward slashes (Linux) and backslashes (Windows)
2. Modified ARN matching regex to be platform-independent
3. Ensured consistent region prefix handling for all supported regions
This change maintains functionality while ensuring cross-platform compatibility.
* fix: make WorkspaceTracker test cross-platform compatible
Fixed an issue where the WorkspaceTracker test 'should initialize with workspace files' was failing on Windows but passing on Linux. The problem was in the mock implementation of toRelativePath that only handled forward slashes.
- Updated the toRelativePath mock to use path.relative which properly handles platform-specific path separators
- Ensured all paths are converted to forward slashes for consistency in test assertions
- The fix maintains cross-platform compatibility while preserving the test's intent
* fix: make WorkspaceTracker tests cross-platform compatible
Fixed cross-platform compatibility issues in the WorkspaceTracker tests that were causing failures on Windows but passing on Linux:
1. Updated the toRelativePath mock implementation to:
- Use path.relative which properly handles platform-specific path separators
- Convert paths to forward slashes for consistency in test assertions
2. Enhanced the 'should not update file paths' test to be platform-agnostic by:
- Using more flexible assertions that don't depend on specific path formats
- Checking file path length and content rather than exact string matches
- Properly typed the test assertions to fix TypeScript errors
These changes preserve the test intent while ensuring they run successfully across different operating systems.
* fix: make McpHub tests cross-platform compatible
Fixed cross-platform compatibility issues in the McpHub tests that were causing failures on Windows but passing on Linux:
1. Made the toggleToolAlwaysAllow tests more platform-agnostic by:
- No longer relying on specific path formats which differ between Windows and Linux
- Using the last write call instead of searching for a specific path string
- Adding more robust assertions that verify structure instead of exact path matches
- Properly handling array existence checks
2. These tests would fail on Windows because paths are formatted with backslashes instead of
forward slashes, causing path equality checks to fail.
The changes maintain test intent while ensuring cross-platform compatibility.
* handle escaping of slash and quote
* fix: ensure consistent line endings in git fallback strategy
Fixed an issue where tests would fail on GitHub Windows runners but pass on local Windows machines due to line ending differences. The fix ensures consistent line ending handling by:
1. Normalizing CRLF to LF when reading files in the git fallback strategy
2. Disabling Git's automatic line ending conversion
3. Maintaining consistent line ending usage throughout text operations
* feat: run tests sequentially on Windows, parallel otherwise
Fix switching profiles to ensure only the selected profile is switched for all modes, and optimize certain i18n keys.
Signed-off-by: feifei <liuyansheng1121@gmail.com>
