bug: state_to_json() doesn't correctly serialize dataclasses

[tgasser-nv]

Did you check docs and existing issues?

• I have read all the NeMo-Guardrails docs
• I have updated the package to the latest version before submitting this issue
• (optional) I have used the develop branch
• I have searched the existing issues of NeMo-Guardrails

Python version (python --version)

Python 3.13.2

Operating system/version

MacOS 15.6.1

NeMo-Guardrails version (if you must use a specific version and not the latest

0.16.0

Describe the bug

The state_to_json() function in colang/v2_x/runtime/serialization.py does not correctly serialize dataclasses. This function has no coverage in unit-tests, and this line shadowed the bug by always passing None into the state_to_json() function. Parsing a None is effectively a no-op for the state_to_json() function.

Once the text = result.text line was removed from nemoguardrails/actions/v2_x/generation.py, this now gives an error with the stacktrace below. We need to fix this, but it isn't necessary to merge the Type fixes. This Issue is tracked in the unittest.skip(reason=...) field

tests/v2_x/test_passthroug_mode.py:83 (TestPassthroughLLMActionLogging.test_passthrough_llm_action_invoked_via_logs)
self = <tests.v2_x.test_passthroug_mode.TestPassthroughLLMActionLogging testMethod=test_passthrough_llm_action_invoked_via_logs>

    def test_passthrough_llm_action_invoked_via_logs(self):
        chat = TestChat(
            config,
            llm_completions=["user asked about capabilites", "a random text from llm"],
        )
        rails = chat.app
    
        logger = logging.getLogger("nemoguardrails.colang.v2_x.runtime.statemachine")
    
        with self.assertLogs(logger, level="INFO") as log:
            messages = [{"role": "user", "content": "What can you do?"}]
>           response = rails.generate(messages=messages)

v2_x/test_passthroug_mode.py:95: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
../nemoguardrails/rails/llm/llmrails.py:1324: in generate
    return loop.run_until_complete(
/opt/homebrew/Cellar/python@3.13/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/base_events.py:725: in run_until_complete
    return future.result()
../nemoguardrails/rails/llm/llmrails.py:964: in generate_async
    output_state = {"state": state_to_json(output_state), "version": "2.x"}
../nemoguardrails/colang/v2_x/runtime/serialization.py:217: in state_to_json
    result = json.dumps(d, indent=indent)
/opt/homebrew/Cellar/python@3.13/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/json/__init__.py:238: in dumps
    **kw).encode(obj)
/opt/homebrew/Cellar/python@3.13/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/json/encoder.py:200: in encode
    chunks = self.iterencode(o, _one_shot=True)
/opt/homebrew/Cellar/python@3.13/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/json/encoder.py:261: in iterencode
    return _iterencode(o, 0)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

self = <json.encoder.JSONEncoder object at 0x169872950>
o = ParsedTaskOutput(text='a random text from llm', reasoning_trace=None)

    def default(self, o):
        """Implement this method in a subclass such that it returns
        a serializable object for ``o``, or calls the base implementation
        (to raise a ``TypeError``).
    
        For example, to support arbitrary iterators, you could
        implement default like this::
    
            def default(self, o):
                try:
                    iterable = iter(o)
                except TypeError:
                    pass
                else:
                    return list(iterable)
                # Let the base class default method raise the TypeError
                return super().default(o)
    
        """
>       raise TypeError(f'Object of type {o.__class__.__name__} '
                        f'is not JSON serializable')
E       TypeError: Object of type ParsedTaskOutput is not JSON serializable

/opt/homebrew/Cellar/python@3.13/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/json/encoder.py:180: TypeError

v2_x/test_story_mechanics.py::test_when_else_deep_hierarchy_case_match 
v2_x/test_story_mechanics.py::test_when_conflict_issue 
v2_x/test_story_mechanics.py::test_flow_event_competition 
[gw3] [ 94%] PASSED test_streaming.py::test_main_llm_supports_streaming_flag_config_combinations[chat-False-False-False] 
v2_x/test_story_mechanics.py::test_flow_bot_question_repetition 

Steps To Reproduce

1. Remove @unittest.skip () decorator from test_passthrough_llm_action_invoked_via_logs() in tests/v2_x/test_passthroug_mode.py
2. Run poetry run pytest tests/v2_x/test_passthroug_mode.py

Expected Behavior

Test Passes

Actual Behavior

Test fails with stacktrace

tests/v2_x/test_passthroug_mode.py:83 (TestPassthroughLLMActionLogging.test_passthrough_llm_action_invoked_via_logs)
self = <tests.v2_x.test_passthroug_mode.TestPassthroughLLMActionLogging testMethod=test_passthrough_llm_action_invoked_via_logs>

    def test_passthrough_llm_action_invoked_via_logs(self):
        chat = TestChat(
            config,
            llm_completions=["user asked about capabilites", "a random text from llm"],
        )
        rails = chat.app
    
        logger = logging.getLogger("nemoguardrails.colang.v2_x.runtime.statemachine")
    
        with self.assertLogs(logger, level="INFO") as log:
            messages = [{"role": "user", "content": "What can you do?"}]
>           response = rails.generate(messages=messages)

v2_x/test_passthroug_mode.py:95: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
../nemoguardrails/rails/llm/llmrails.py:1324: in generate
    return loop.run_until_complete(
/opt/homebrew/Cellar/python@3.13/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/base_events.py:725: in run_until_complete
    return future.result()
../nemoguardrails/rails/llm/llmrails.py:964: in generate_async
    output_state = {"state": state_to_json(output_state), "version": "2.x"}
../nemoguardrails/colang/v2_x/runtime/serialization.py:217: in state_to_json
    result = json.dumps(d, indent=indent)
/opt/homebrew/Cellar/python@3.13/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/json/__init__.py:238: in dumps
    **kw).encode(obj)
/opt/homebrew/Cellar/python@3.13/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/json/encoder.py:200: in encode
    chunks = self.iterencode(o, _one_shot=True)
/opt/homebrew/Cellar/python@3.13/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/json/encoder.py:261: in iterencode
    return _iterencode(o, 0)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

self = <json.encoder.JSONEncoder object at 0x169872950>
o = ParsedTaskOutput(text='a random text from llm', reasoning_trace=None)

    def default(self, o):
        """Implement this method in a subclass such that it returns
        a serializable object for ``o``, or calls the base implementation
        (to raise a ``TypeError``).
    
        For example, to support arbitrary iterators, you could
        implement default like this::
    
            def default(self, o):
                try:
                    iterable = iter(o)
                except TypeError:
                    pass
                else:
                    return list(iterable)
                # Let the base class default method raise the TypeError
                return super().default(o)
    
        """
>       raise TypeError(f'Object of type {o.__class__.__name__} '
                        f'is not JSON serializable')
E       TypeError: Object of type ParsedTaskOutput is not JSON serializable

/opt/homebrew/Cellar/python@3.13/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/json/encoder.py:180: TypeError

v2_x/test_story_mechanics.py::test_when_else_deep_hierarchy_case_match 
v2_x/test_story_mechanics.py::test_when_conflict_issue 
v2_x/test_story_mechanics.py::test_flow_event_competition 
[gw3] [ 94%] PASSED test_streaming.py::test_main_llm_supports_streaming_flag_config_combinations[chat-False-False-False] 
v2_x/test_story_mechanics.py::test_flow_bot_question_repetition