Add Pydantic based evaluate and retry API call block for inappropriate responses

[Kwanky-Millet]

By inappropriate responses, you mean LLM outputs being none?

[venkateshtata]

Not just "none", pydantic allows langchain to compare the LLM response to the description of the variable that we are trying to extract value for. We need to implement that mechanism to NetHandle. For examples, checking things like if the retriever agent's response contain referenced records of the user or did it fail ? and if its a fail, we need to add a retry