Guardrails Top Level Settings

[CalebCourier]

Adds a settings singleton to the guardrails package. Currently only supports one setting use_server that allows users to specify when to run guards on the guardrails-api vs on the client.

Tested with the following:

config.py

from guardrails import Guard
from guardrails.hub import RegexMatch

name_case = Guard(
    name='name-case',
    description='Checks that a string is in Name Case format.'
).use(
    RegexMatch(regex="^[A-Z][a-z\\s]*$")
)

start command

guardrails start --config=config.py

client side script

import os
import logging
from rich import print
from guardrails import configure_logging
from guardrails import Guard, settings

os.environ["GUARDRAILS_BASE_URL"] = "http://localhost:8000"

print(os.environ.get("GUARDRAILS_API_KEY", "No API key found"))
print(os.environ.get("GUARDRAILS_BASE_URL"))

settings.use_server = True
# settings.use_server = False
configure_logging(None, log_level=logging.DEBUG)


name_case = Guard(name='name-case')


result = name_case.parse(
    llm_output="z-dawg"
)

print("result: ", result)
print("number of validators run: ", len(name_case.history.last.validator_logs))

[JosephCatrambone]

Thank you!