[BUG] validation error for AnalystOutput

[rjavadi]

Bug Description

 Item 9 failed: 1 validation error for AnalystOutput
flagged_transaction_ids
  Input should be a valid string [type=string_type, input_value=['5d6de366f9d9b374', '7c2...0e', '3d3bf35982b11b91'], input_type=list]
    For further information visit https://errors.pydantic.dev/2.12/v/string_type

Steps to Reproduce

In implementations/aml_investigation/03_evaluation.ipynb , cell under running the experiment I got this error.

The eval fails at AnalystOutput validation in AmlInvestigationTask: the schema says comma-separated str, but the agent output often uses a JSON list. Fixing it means either widening the schema (e.g. str | list[str] + a validator that normalizes to one representation) or tightening the agent/output instructions so it always emits a comma-separated string.

Code snippet (if applicable):

Expected Behavior

We shouldn't get error message for AnalystOutput object validation.

Actual Behavior

Environment

Coder platform for the project

Screenshots/Logs

Additional Context

Possible Solution

# In cases.py:

@field_validator("flagged_transaction_ids", mode="before")
    @classmethod
    def _normalize_flagged_transaction_ids(cls, value: Any) -> str:
        """Accept comma-separated str or JSON list from model output; store as str."""
        if value is None:
            return ""
        if isinstance(value, str):
            return value
        if isinstance(value, list | tuple | set):
            return ",".join(str(item).strip() for item in value if str(item).strip())
        return str(value).strip()