fix: add path-based repetition detection for read_file tool

[roomote]

Related GitHub Issue

Closes: #11071

Description

This PR attempts to address Issue #11071 where GLM4.5 (and potentially other models) gets stuck in a loop reading the same file repeatedly with different parameters (e.g., different line ranges).

The Problem:
The existing ToolRepetitionDetector only catches byte-for-byte identical tool calls. When a model reads the same file with slightly different parameters (like different line ranges), each call is considered unique and bypasses detection.

The Solution:
This PR adds path-based repetition detection specifically for read_file tool calls:

1. Track consecutive read_file calls that target the same file path(s)
2. Block execution when the same file is read more than N times consecutively, regardless of other parameters (like line ranges)
3. Keep existing identical-call detection as the primary mechanism
4. Reset tracking when a different tool is used, allowing recovery

Key Implementation Details:

• Added checkReadFilePathRepetition() method that extracts file paths and tracks repetitions
• Supports both params-based and nativeArgs-based formats
• Uses same limit as identical call detection by default, but can be configured separately
• Added translation key for the new error message

This approach is provider-agnostic and will work regardless of which API provider is used (LM Studio, OpenAI-compatible, or others).

Feedback and guidance are welcome!

Test Procedure

1. Run the new test cases:

   cd src && npx vitest run core/tools/__tests__/ToolRepetitionDetector.spec.ts

2. All 39 tests pass, including 8 new tests for path-based detection:
   • Detection of same file with different line ranges
   • Detection with nativeArgs format
   • Allowing different files consecutively
   • Resetting counter when different tool is used
   • Custom readFilePathLimit configuration
   • Recovery after limit is reached
   • Non-interference with other tools

Pre-Submission Checklist

• Issue Linked: This PR is linked to an approved GitHub Issue (see "Related GitHub Issue" above).
• Scope: My changes are focused on the linked issue (one major feature/fix per PR).
• Self-Review: I have performed a thorough self-review of my code.
• Testing: New and/or updated tests have been added to cover my changes (if applicable).
• Documentation Impact: I have considered if my changes require documentation updates (see "Documentation Updates" section below).
• Contribution Guidelines: I have read and agree to the Contributor Guidelines.

Documentation Updates

• No documentation updates are required.

Additional Notes

This fix provides a more robust solution that works at the tool detection level rather than targeting specific providers. The user reported the issue affects both LM Studio and OpenAI-compatible endpoints, so a provider-agnostic approach is more appropriate.

---

Important

Adds path-based repetition detection for read_file tool in ToolRepetitionDetector to prevent loops when reading the same file with different parameters.

• Behavior:
  • Adds path-based repetition detection for read_file in ToolRepetitionDetector.ts.
  • Blocks execution if the same file is read more than N times consecutively, regardless of parameters.
  • Resets tracking when a different tool is used.
• Implementation:
  • Adds checkReadFilePathRepetition() to track file path repetitions.
  • Supports both params and nativeArgs formats.
  • Configurable readFilePathLimit added to constructor.
• Tests:
  • Adds 8 new tests in ToolRepetitionDetector.spec.ts for path-based detection.
  • Tests include different line ranges, nativeArgs format, and custom limits.
• Localization:
  • Adds readFilePathRepetitionLimitReached message to tools.json.

^{This description was created by for 538ecae. You can customize this summary. It will automatically update as commits are pushed.}

[roomote]

Rooviewer   See task on Roo Cloud

Review complete. No issues found.

The implementation is well-designed and follows existing patterns:

• Path-based repetition detection correctly catches models reading the same file with different parameters
• Path extraction handles both native protocol and legacy formats
• Tests are comprehensive (8 new tests, all passing)
• Recovery mechanism works correctly

_{Mention @roomote in a comment to request specific changes to this pull request or fix all unresolved issues.}