ai-agents-pydantic-conversion--when-using-crewai-with-a-small-llm-and-complex-out-7fa9c2bb

THEN Inspect the underlying Pydantic validation error by catching ConverterError separately, or switch to a more capable LLM (e.g., GPT-4 instead of Llama 3.2 3B). Alternatively, patch the converter in crewai/utilities/converter.py to pass the agent object (not None) to fallback functions, so the recovery logic works and a meaningful error is raised.

# In your own code, wrap task execution to catch ConverterError and log the original validation error:
from crewai.utilities.converter import ConverterError
try:
    result = crew.kickoff()
except ConverterError as e:
    print(f"Pydantic conversion failed: {e}")
    # Inspect e.__cause__ for original ValidationError