[hotfix][mcp] Fix MCPTool Jackson deserialization by ignoring unknown properties

[avichaym]

What is the purpose of the change

MCPTool deserialization fails with UnrecognizedPropertyException when a default ObjectMapper encounters the name and description fields serialized from the parent Tool class. These fields are not declared in MCPTool's @JsonCreator constructor.

This breaks checkpoint/restore and any JSON round-trip of MCPTool.

Brief change log

• Add @JsonIgnoreProperties(ignoreUnknown = true) to MCPTool
• Add regression test testJsonDeserializationWithDefaultMapper that uses a default ObjectMapper (the existing testJsonSerialization masked the bug by disabling FAIL_ON_UNKNOWN_PROPERTIES globally)

Does this pull request potentially affect one of the following parts

• Runtime: yes (checkpoint/restore of MCP tool state)
• Integration: yes (MCP module)

Documentation

N/A — bug fix only

Documentation

• doc-needed
• doc-not-needed
• doc-included

[wenjin272]

Hi, @avichaym, thanks for your contribution, and introducing Amazon integrations to Flink-Agents. I'll strive to start the review before the Chinese Spring Festival holiday.

I notice there are some test failures because of code style check, you can format the code by mvn spotless:apply.

[avichaym]

thank you @wenjin272 ... i fixed the code-styling issues. Now the python related tests fails in all 3 PR's due to what seems to be a worker issue , not related to the code ->

The test_long_term_memory_async_execution_in_action test fails because the nomic-embed-text model download didn't complete in time. The CI log shows the pull was still at 35% (95/274 MB at 7.3 KB/s) when
the test started:

pulling 970aa74c0a90: 35% ▕██████ ▏ 95 MB/274 MB 7.3 KB/s 6h48m

The pull_model() helper in test_utils.py has a 120s timeout and silently and seems to swallow the timeout exception, so the test proceeds without the model and fails with model "nomic-embed-text" not found. This is
probably a runner network speed issue

[yanand0909]

Thanks @avichaym, this looks good to me.