import json from haystack.core.serialization import import_class_by_name import pydantic from pydantic import ValidationError from typing import Optional, List from colorama import Fore from haystack import component """ output_validator.py Purpose: Ensures that a JSON output generated by an LLM model is formatted correctly. Copied from https://haystack.deepset.ai/tutorials/28_structured_output_with_loop ! ; """ # Define the component input parameters @component class OutputValidator: def __init__(self, pydantic_model: type[pydantic.BaseModel]): self.pydantic_model = pydantic_model self.iteration_counter = 0 def to_dict(self): return { "pydantic_model": f"{self.pydantic_model.__module__}.{self.pydantic_model.__name__}", "type": "components.output_validator.OutputValidator", } @classmethod def from_dict(cls, data): print(Fore.YELLOW + "Inside from_dict...") model_class = import_class_by_name(data["pydantic_model"]) return cls( pydantic_model=model_class, ) # Define the component output @component.output_types( valid_replies=List[str], invalid_replies=Optional[List[str]], error_message=Optional[str], ) def run(self, body: str): self.iteration_counter += 1 ## Try to parse the LLM's reply ## # If the LLM's reply is a valid object, return `"valid_replies"` try: output_dict = json.loads(body) self.pydantic_model.model_validate(output_dict) print( Fore.GREEN + f"OutputValidator at Iteration {self.iteration_counter}: Valid JSON from LLM - No need for looping: {body}" ) return {"valid_replies": body} # If the LLM's reply is corrupted or not valid, return "invalid_replies" and the "error_message" for LLM to try again except (ValueError, ValidationError) as e: print( Fore.RED + f"OutputValidator at Iteration {self.iteration_counter}: Invalid JSON from LLM - Let's try again.\n" f"Output from LLM:\n {body} \n" f"Error from OutputValidator: {e}" ) return {"invalid_replies": body, "error_message": str(e)}