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()
Bug Description
Steps to Reproduce
In
implementations/aml_investigation/03_evaluation.ipynb, cell under running the experiment I got this error.The eval fails at
AnalystOutputvalidation in AmlInvestigationTask: the schema says comma-separated str, but the agent output often uses a JSONlist. 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
AnalystOutputobject validation.Actual Behavior
Environment
Coder platform for the project
Screenshots/Logs
Additional Context
Possible Solution