Adds json_count_leaves utility function

[aditchawdhary]

Purpose

Add json_count_leaves utility function to count leaf elements in nested JSON structures, with integration into multimodal cache system for enhanced debugging capabilities.

Test Plan

# Run the new tests for json_count_leaves function
python -m pytest tests/utils_/test_utils.py::test_json_count_leaves -v

# Run all utils tests to ensure no regressions
python -m pytest tests/utils_/test_utils.py -v

Test Result

• All new tests pass for various data structures (nested dicts, lists, tuples, mixed types)
• Edge cases covered (empty structures, single values, deeply nested data)
• Integration with multimodal cache system works correctly
• No regressions in existing test suite

Changes made:

• Added json_count_leaves function in vllm/utils/jsontree.py
• Enhanced cache debug logging to include leaf count in vllm/multimodal/cache.py
• Added get_item_complexity method for structural complexity analysis
• Comprehensive test coverage in tests/utils_/test_utils.py

Example usage:

from vllm.utils.jsontree import json_count_leaves

data = {"images": [{"pixels": [[1, 2], [3, 4]]}]}
count = json_count_leaves(data)  # Returns 4

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

[gemini-code-assist]

Code Review

This pull request introduces a json_count_leaves utility function to count leaf elements in nested JSON-like structures and integrates it into the multimodal cache system for enhanced debugging. The implementation is clean and comes with a comprehensive set of unit tests. My main feedback is to improve the robustness of the new json_count_leaves function by switching from a recursive to an iterative implementation. This will prevent potential RecursionError exceptions when dealing with very deeply nested data structures.

[gemini-code-assist]

The current implementation uses json_iter_leaves, which is recursive. This can lead to a RecursionError for deeply nested JSON structures (the default limit in Python is often 1000). To make this utility function more robust, an iterative approach using an explicit stack is recommended. This avoids the recursion limit and is generally safer for functions processing arbitrary tree-like data structures.

Suggested change

	return sum(1 for _ in json_iter_leaves(value))
	count = 0
	stack = [value]
	while stack:
	node = stack.pop()
	if isinstance(node, dict):
	stack.extend(node.values())
	elif isinstance(node, (list, tuple)):
	stack.extend(node)
	else:
	count += 1
	return count

[DarkLight1337]

Thanks, can you fix pre-commit?