fix(mcp): isolate strict schema conversion from non-strict fallback

[adityasingh2400]

Summary

ensure_strict_json_schema mutates the input schema in place and may raise partway through (e.g. when a nested object has additionalProperties: true). When that happens, MCPUtil.to_function_tool catches the exception and proceeds with is_strict=False — but the schema it serves to the model has already been partially mutated, leaking strict-mode artifacts like additionalProperties: false and required: [...] onto a tool we explicitly mark as non-strict.

Repro on main:

schema = {
    "type": "object",
    "properties": {"x": {"type": "object", "additionalProperties": True}},
}
tool = MCPTool(name="test_tool", inputSchema=schema)
ft = MCPUtil.to_function_tool(tool, FakeMCPServer(), convert_schemas_to_strict=True)
# ft.strict_json_schema == False  (correct: conversion failed)
# ft.params_json_schema now contains the mutations from the failed conversion:
#   {"type": "object", "properties": {...}, "additionalProperties": False, "required": ["x"]}

The fix passes a fresh copy.deepcopy(schema) to ensure_strict_json_schema, so a partial in-place mutation cannot leak into the non-strict fallback.

Test

tests/mcp/test_mcp_util.py::test_to_function_tool_failed_strict_conversion_keeps_original_schema — fails on main with the partially-mutated schema, passes after this change.

[seratch]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. What shall we delve into next?

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".