langchain - 💡(How to fix) Fix Add comprehensive unit tests for ModelCallLimitMiddleware [1 comments, 1 participants]

Checked other resources

• This is a feature request, not a bug report or usage question.
• I added a clear and descriptive title that summarizes the feature request.
• I used the GitHub search to find a similar feature request and did not find it.
• I checked the LangChain documentation and API reference to see if this feature already exists.
• This is not related to the langchain-community package.

Package (Required)

• langchain

Feature Description

I would like to add comprehensive unit tests for ModelCallLimitMiddleware in the langchain_v1 package.

The current test-to-source ratio for this middleware is 0.85 (227 lines of tests / 267 lines of source), which is significantly lower than the average ratio of ~1.6 for other middleware implementations in the same directory.

After analyzing the existing tests in tests/unit_tests/agents/middleware/implementations/test_model_call_limit.py, I identified the following coverage gaps:

High Priority Gaps

1. Async methods are not tested independently

   • abefore_model() has no dedicated test
   • aafter_model() has no dedicated test

2. after_model counting logic is not fully tested

   • No test verifying count increments from 0 to 1, 1 to 2, etc.
   • No test verifying the return dict structure

3. Helper function _build_limit_exceeded_message() is not directly tested

   • No test for thread-only limit exceeded
   • No test for run-only limit exceeded
   • No test for both limits exceeded
   • No test for neither limit exceeded (edge case)

Medium Priority Gaps

1. Boundary condition tests are missing

   • count exactly at limit (should exceed)
   • count one below limit (should not exceed)
   • zero count with low limit

2. Combined scenario tests are incomplete

   • thread exceeded but run not exceeded
   • run exceeded but thread not exceeded
   • both exceeded simultaneously

Use Case

Improved test coverage for ModelCallLimitMiddleware would:

1. Prevent regressions - Ensure async methods continue to work correctly as the codebase evolves
2. Validate counting logic - Confirm that after_model/aafter_model correctly increment both thread_model_call_count and run_model_call_count
3. Verify error messages - Ensure _build_limit_exceeded_message produces accurate messages for all limit combinations
4. Cover edge cases - Boundary conditions and combined scenarios that could easily break with future changes
5. Bring test coverage in line with other middleware - The current 0.85 ratio is well below the ~1.6 average

Proposed Solution

I plan to add a new test file tests/unit_tests/agents/middleware/implementations/test_model_call_limit_additional.py with the following test classes:

• TestBuildLimitExceededMessage - Direct tests for the helper function (5 tests)
• TestAsyncMethods - Tests for abefore_model and aafter_model (5 tests)
• TestAfterModelCounting - Tests for counting logic in after_model/aafter_model (5 tests)
• TestBoundaryConditions - Boundary value analysis tests (4 tests)
• TestCombinedScenarios - Tests for combined thread + run limit scenarios (5 tests)
• TestEdgeCases - Tests for single-limit and missing-count scenarios (3 tests)

Total: ~27 new test cases, ~400 LOC

This would bring the test-to-source ratio from 0.85 to ~2.4, in line with other middleware implementations.

All tests will:

• Use pytest with @pytest.mark.asyncio for async methods
• Use FakeToolCallingModel for end-to-end tests (no network calls)
• Follow the existing test patterns in the repository
• Include proper type annotations and docstrings

Alternatives Considered

1. Extending the existing test file - Considered adding tests directly to test_model_call_limit.py, but a separate file keeps the PR scope clear and avoids cluttering the existing well-organized test file.

2. Only adding async tests - Considered a smaller scope, but the other gaps (counting logic, helper function, boundary conditions) are equally important and would likely need separate issues otherwise.

Additional Context

• Source file: libs/langchain_v1/langchain/agents/middleware/model_call_limit.py
• Existing tests: libs/langchain_v1/tests/unit_tests/agents/middleware/implementations/test_model_call_limit.py
• Similar test coverage improvements have been merged recently (e.g., PR #36644 which added init tests with middleware)

I am happy to implement this if assigned. Thank you for considering this request!