Using Any LLM as a Guardrail¶

This tutorial will show you how to use any LLM as a baseline guardrail that can validate input text against a custom policy, using the built-in guardrail based on any-llm.

Install dependencies¶

In [ ]:

%pip install 'any-guardrail'

import nest_asyncio

nest_asyncio.apply()

%pip install 'any-guardrail' import nest_asyncio nest_asyncio.apply()

We will be using a model from openai by default, but you can check the different providers supported in any-llm:

https://mozilla-ai.github.io/any-llm/providers/

In [8]:

import os
from getpass import getpass

if "OPENAI_API_KEY" not in os.environ:
    print("OPENAI_API_KEY not found in environment!")
    api_key = getpass("Please enter your OPENAI_API_KEY: ")
    os.environ["OPENAI_API_KEY"] = api_key
    print("OPENAI_API_KEY set for this session!")
else:
    print("OPENAI_API_KEY found in environment.")

import os from getpass import getpass if "OPENAI_API_KEY" not in os.environ: print("OPENAI_API_KEY not found in environment!") api_key = getpass("Please enter your OPENAI_API_KEY: ") os.environ["OPENAI_API_KEY"] = api_key print("OPENAI_API_KEY set for this session!") else: print("OPENAI_API_KEY found in environment.")

OPENAI_API_KEY found in environment.

Create the guardrail¶

In [3]:

from any_guardrail import AnyGuardrail, GuardrailName

from any_guardrail import AnyGuardrail, GuardrailName

In [4]:

guardrail = AnyGuardrail.create(GuardrailName.ANYLLM)

guardrail = AnyGuardrail.create(GuardrailName.ANYLLM)

Try it with different models / policies / inputs¶

In [5]:

MODEL_ID = "openai/gpt-5-nano"

POLICY = """
You hate Mondays.
You must reject any request related with planning activities on Mondays.
"""

MODEL_ID = "openai/gpt-5-nano" POLICY = """ You hate Mondays. You must reject any request related with planning activities on Mondays. """

In [6]:

guardrail.validate("Can you suggest me some restaurants for lunch on Monday?", policy=POLICY, model_id=MODEL_ID)

guardrail.validate("Can you suggest me some restaurants for lunch on Monday?", policy=POLICY, model_id=MODEL_ID)

Out[6]:

GuardrailOutput(valid=False, explanation='The user requested planning lunch options specifically for Monday, which is planning activities on Monday. The policy requires rejecting Monday planning requests.', score=0.92)

In [7]:

guardrail.validate("Can you suggest me some restaurants for lunch on Friday?", policy=POLICY, model_id=MODEL_ID)

guardrail.validate("Can you suggest me some restaurants for lunch on Friday?", policy=POLICY, model_id=MODEL_ID)

Out[7]:

GuardrailOutput(valid=True, explanation='The user asks for lunch restaurant suggestions on Friday. The policy forbids planning activities on Mondays, which does not apply here.', score=0.87)