Come with a developer guide on adding new validators in Pythia

[cloudronin]

Currently Pythia ships with a handful of validators, we need to come up with a guide that explain how additional custom validators can be built in Pythia using a descriptive developer guide

[pratacosmin]

The next steps explain how a new validator can be added using the current format. This change require a new image build/deployment.

Add new validator

1. Add the actual implementation under pythia/validators/

Create a file with the validator implementation following the template

### validator implementation template
def validate(text):
    try:
        #some config
        err_msg, is_valid, risk_score = validate_text(text)

        return {
            "isValid": is_valid,
            "errorMessage": err_msg,
            "riskScore": risk_score

        }
    except Exception as e:
        print(e)
        return {
            "isValid": False,
            "errorMessage": str(e),
            "riskScore": 1
        }

2. Create a method with the validator name to be called

This part will allow us to call specific validators using only the name and also allow us to prepare the input and also how the validator is user (input, output or both base on the config) Add the method in the ValidatorCall clss in validator_call.py

from pythia.validators import validator_impl as validator_method

class ValidatorCall:
    def detect_pii(self, validator, **kwargs):
        question = get_question(kwargs=kwargs)
        input_data_reference = get_input_data(validator=validator, kwargs=kwargs)
        output_data_response = get_output_data(validator=validator, kwargs=kwargs)
        validator_results = []
        if input_data_reference:
            validator_result = validator_method.validate(text=input_data_reference)
            validator_result["validatedField"] = validator["input"]
            validator_result["validator"] = validator
            validator_results.append(validator_result)
        if output_data_response:
            validator_result = validator_method.validate(text=output_data_response)
            validator_result["validatedField"] = validator["output"]
            validator_result["validator"] = validator
            validator_results.append(validator_result)
        return validator_results

3. Add the validator to the config file

We control what validators are executed using the config.yaml file therefore we have to add the new created validator the file and enable him. Also we can specify what field to be use for the input or output ( null if no support)

    - name: validator_name
      description: "Custom Validaotr"
      source: custom
      enabled: true
      input: input_reference
      output: input_response