Custom Guardrail

Use this if you want to write code to run a custom guardrail

Quick Start

1. Write a CustomGuardrail Class

The simplest way to create a custom guardrail is by implementing the apply_guardrail method. This method is called to check text content and can block requests by raising an exception.

Example CustomGuardrail Class

Create a new file called custom_guardrail.py and add this code to it:

import os
from typing import Optional, List
from litellm.integrations.custom_guardrail import CustomGuardrail
from litellm.types.guardrails import PiiEntityType
from litellm._logging import verbose_proxy_logger
from litellm.llms.custom_httpx.http_handler import (
    get_async_httpx_client,
    httpxSpecialProvider,
)

class myCustomGuardrail(CustomGuardrail):
    def __init__(self, api_key: Optional[str] = None, api_base: Optional[str] = None, **kwargs):
        self.api_key = api_key or os.getenv("MY_GUARDRAIL_API_KEY")
        self.api_base = api_base or os.getenv("MY_GUARDRAIL_API_BASE", "https://api.myguardrail.com")
        super().__init__(**kwargs)

    async def apply_guardrail(
        self,
        text: str, # IMPORTANT: This is the text to check against your guardrail rules. It's extracted from the request or response across all LLM call types.
        language: Optional[str] = None, # ignore 
        entities: Optional[List[PiiEntityType]] = None, # ignore
        request_data: Optional[dict] = None, # ignore
    ) -> str:
        """
        Check text content against your guardrail rules.
        Raise an exception to block the request.
        Return the text (optionally modified) to allow it through.
        """
        result = await self._check_with_api(text, request_data)
        
        if result.get("action") == "BLOCK":
            raise Exception(f"Content blocked: {result.get('reason', 'Policy violation')}")
        
        return text

    async def _check_with_api(self, text: str, request_data: Optional[dict]) -> dict:
        async_client = get_async_httpx_client(llm_provider=httpxSpecialProvider.LoggingCallback)
        
        headers = {
            "Content-Type": "application/json",
            "Authorization": f"Bearer {self.api_key}",
        }
        
        response = await async_client.post(
            f"{self.api_base}/check",
            headers=headers,
            json={"text": text},
            timeout=5,
        )
        
        response.raise_for_status()
        return response.json()

Advanced: Using Individual Event Hooks

If you need more fine-grained control, you can implement individual event hooks instead of (or in addition to) apply_guardrail:

• async_pre_call_hook - Modify input or reject request before making LLM API call
• async_moderation_hook - Reject request, runs in parallel with LLM API call (helps lower latency)
• async_post_call_success_hook - Apply guardrail on input/output, runs after making LLM API call
• async_post_call_streaming_iterator_hook - Pass the entire stream to the guardrail

See examples of individual event hooks here | See detailed spec of methods here

2. Pass your custom guardrail class in LiteLLM config.yaml

In the config below, we point the guardrail to our custom guardrail by setting guardrail: custom_guardrail.myCustomGuardrail

• Python Filename: custom_guardrail.py
• Guardrail class name : myCustomGuardrail. This is defined in Step 1

guardrail: custom_guardrail.myCustomGuardrail

model_list:
  - model_name: gpt-4
    litellm_params:
      model: openai/gpt-4o
      api_key: os.environ/OPENAI_API_KEY

guardrails:
  - guardrail_name: "my-custom-guardrail"
    litellm_params:
      guardrail: custom_guardrail.myCustomGuardrail  # 👈 Key change
      mode: "during_call"               # runs apply_guardrail method
      api_key: os.environ/MY_GUARDRAIL_API_KEY
      api_base: https://api.myguardrail.com

Mode Options

• during_call - Default mode, runs apply_guardrail method (or async_moderation_hook if using individual hooks)
• pre_call - Runs async_pre_call_hook for input modification
• post_call - Runs async_post_call_success_hook for output validation

Advanced: Multiple modes with individual event hooks

If you're using individual event hooks, you can configure multiple guardrails with different modes:

guardrails:
  - guardrail_name: "custom-pre-guard"
    litellm_params:
      guardrail: custom_guardrail.myCustomGuardrail
      mode: "pre_call"                  # runs async_pre_call_hook
  - guardrail_name: "custom-during-guard"
    litellm_params:
      guardrail: custom_guardrail.myCustomGuardrail  
      mode: "during_call"               # runs async_moderation_hook
  - guardrail_name: "custom-post-guard"
    litellm_params:
      guardrail: custom_guardrail.myCustomGuardrail
      mode: "post_call"                 # runs async_post_call_success_hook

3. Start LiteLLM Gateway

Docker Run

Mount your custom_guardrail.py on the LiteLLM Docker container

This mounts your custom_guardrail.py file from your local directory to the /app directory in the Docker container, making it accessible to the LiteLLM Gateway.

docker run -d \
  -p 4000:4000 \
  -e OPENAI_API_KEY=$OPENAI_API_KEY \
  --name my-app \
  -v $(pwd)/my_config.yaml:/app/config.yaml \
  -v $(pwd)/custom_guardrail.py:/app/custom_guardrail.py \
  my-app:latest \
  --config /app/config.yaml \
  --port 4000 \
  --detailed_debug \

litellm pip

litellm --config config.yaml --detailed_debug

4. Test it

Langchain, OpenAI SDK Usage Examples

Blocked Request

This request will be blocked if it violates your guardrail policy:

curl -i -X POST http://localhost:4000/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-1234" \
-d '{
    "model": "gpt-4",
    "messages": [
        {
            "role": "user",
            "content": "Content that violates policy"
        }
    ],
   "guardrails": ["my-custom-guardrail"]
}'

Expected response when blocked:

{
  "error": {
    "message": "Content blocked: Policy violation",
    "type": "None",
    "param": "None",
    "code": "500"
  }
}

Successful Call

This request passes the guardrail:

curl -i http://localhost:4000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-1234" \
  -d '{
    "model": "gpt-4",
    "messages": [
      {"role": "user", "content": "What is the weather like today?"}
    ],
    "guardrails": ["my-custom-guardrail"]
  }'

Advanced: Testing individual event hooks

If you're using individual event hooks, you can test each mode separately:

Test "custom-pre-guard"

Modify input

Expect this to mask the word litellm before sending the request to the LLM API. This runs the async_pre_call_hook

curl -i  -X POST http://localhost:4000/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-1234" \
-d '{
    "model": "gpt-4",
    "messages": [
        {
            "role": "user",
            "content": "say the word - `litellm`"
        }
    ],
   "guardrails": ["custom-pre-guard"]
}'

Successful Call

curl -i http://localhost:4000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-1234" \
  -d '{
    "model": "gpt-3.5-turbo",
    "messages": [
      {"role": "user", "content": "hi what is the weather"}
    ],
    "guardrails": ["custom-pre-guard"]
  }'

Test "custom-during-guard"

Unsuccessful call

Expect this to fail since litellm is in the message content. This runs the async_moderation_hook

curl -i  -X POST http://localhost:4000/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-1234" \
-d '{
    "model": "gpt-4",
    "messages": [
        {
            "role": "user",
            "content": "say the word - `litellm`"
        }
    ],
   "guardrails": ["custom-during-guard"]
}'

Expected response:

{
  "error": {
    "message": "Guardrail failed words - `litellm` detected",
    "type": "None",
    "param": "None",
    "code": "500"
  }
}

Successful Call

curl -i http://localhost:4000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-1234" \
  -d '{
    "model": "gpt-3.5-turbo",
    "messages": [
      {"role": "user", "content": "hi what is the weather"}
    ],
    "guardrails": ["custom-during-guard"]
  }'

Test "custom-post-guard"

Unsuccessful call

Expect this to fail since coffee will be in the response content. This runs the async_post_call_success_hook

curl -i  -X POST http://localhost:4000/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-1234" \
-d '{
    "model": "gpt-4",
    "messages": [
        {
            "role": "user",
            "content": "what is coffee"
        }
    ],
   "guardrails": ["custom-post-guard"]
}'

Expected response:

{
  "error": {
    "message": "Guardrail failed Coffee Detected",
    "type": "None",
    "param": "None",
    "code": "500"
  }
}

Successful Call

curl -i  -X POST http://localhost:4000/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-1234" \
-d '{
    "model": "gpt-4",
    "messages": [
        {
            "role": "user",
            "content": "what is tea"
        }
    ],
   "guardrails": ["custom-post-guard"]
}'

✨ Pass additional parameters to guardrail

info

✨ This is an Enterprise only feature Contact us to get a free trial

Use this to pass additional parameters to the guardrail API call. e.g. things like success threshold

1. Use get_guardrail_dynamic_request_body_params

get_guardrail_dynamic_request_body_params is a method of the litellm.integrations.custom_guardrail.CustomGuardrail class that fetches the dynamic guardrail params passed in the request body.

from typing import Any, Dict, List, Literal, Optional, Union
import litellm
from litellm._logging import verbose_proxy_logger
from litellm.caching.caching import DualCache
from litellm.integrations.custom_guardrail import CustomGuardrail
from litellm.proxy._types import UserAPIKeyAuth

class myCustomGuardrail(CustomGuardrail):
    def __init__(self, **kwargs):
        super().__init__(**kwargs)

    async def async_pre_call_hook(
        self,
        user_api_key_dict: UserAPIKeyAuth,
        cache: DualCache,
        data: dict,
        call_type: Literal[
            "completion",
            "text_completion",
            "embeddings",
            "image_generation",
            "moderation",
            "audio_transcription",
            "pass_through_endpoint",
            "rerank"
        ],
    ) -> Optional[Union[Exception, str, dict]]:
        # Get dynamic params from request body
        params = self.get_guardrail_dynamic_request_body_params(request_data=data)
        # params will contain: {"success_threshold": 0.9}
        verbose_proxy_logger.debug("Guardrail params: %s", params)
        return data

2. Pass parameters in your API requests:

LiteLLM Proxy allows you to pass guardrails in the request body, following the guardrails spec.

OpenAI Python

import openai
client = openai.OpenAI(
    api_key="anything",
    base_url="http://0.0.0.0:4000"
)

response = client.chat.completions.create(
    model="gpt-3.5-turbo",
    messages=[{"role": "user", "content": "Write a short poem"}],
    extra_body={
        "guardrails": [
            "custom-pre-guard": {
                "extra_body": {
                    "success_threshold": 0.9
                }
            }
        ]
    }
)

Curl

curl 'http://0.0.0.0:4000/chat/completions' \
    -H 'Content-Type: application/json' \
    -d '{
    "model": "gpt-3.5-turbo",
    "messages": [
        {
            "role": "user",
            "content": "Write a short poem"
        }
    ],
    "guardrails": [
        "custom-pre-guard": {
            "extra_body": {
                "success_threshold": 0.9
            }
        }
    ]
}'

The get_guardrail_dynamic_request_body_params method will return:

{
    "success_threshold": 0.9
}

Advanced: Individual Event Hooks

Pro: More flexibility Con: You need to implement this for each LLM call type (chat completions, text completions, embeddings, image generation, moderation, audio transcription, pass through endpoint, rerank, etc. )

For more fine-grained control over when and how your guardrail runs, you can implement individual event hooks. This gives you flexibility to:

• Modify inputs before the LLM call
• Run checks in parallel with the LLM call (lower latency)
• Validate or modify outputs after the LLM call
• Process streaming responses

Example with Individual Event Hooks

from typing import Any, AsyncGenerator, Literal, Optional, Union

import litellm
from litellm._logging import verbose_proxy_logger
from litellm.caching.caching import DualCache
from litellm.integrations.custom_guardrail import CustomGuardrail
from litellm.proxy._types import UserAPIKeyAuth
from litellm.types.utils import ModelResponseStream, CallTypes


class myCustomGuardrail(CustomGuardrail):
    def __init__(
        self,
        **kwargs,
    ):
        # store kwargs as optional_params
        self.optional_params = kwargs

        super().__init__(**kwargs)

    async def async_pre_call_hook(
        self,
        user_api_key_dict: UserAPIKeyAuth,
        cache: DualCache,
        data: dict,
        call_type: Optional[CallTypes],
    ) -> Optional[Union[Exception, str, dict]]:
        """
        Runs before the LLM API call
        Runs on only Input
        Use this if you want to MODIFY the input
        """

        # In this guardrail, if a user inputs `litellm` we will mask it and then send it to the LLM
        _messages = data.get("messages")
        if _messages:
            for message in _messages:
                _content = message.get("content")
                if isinstance(_content, str):
                    if "litellm" in _content.lower():
                        _content = _content.replace("litellm", "********")
                        message["content"] = _content

        verbose_proxy_logger.debug(
            "async_pre_call_hook: Message after masking %s", _messages
        )

        return data

    async def async_moderation_hook(
        self,
        data: dict,
        user_api_key_dict: UserAPIKeyAuth,
        call_type: Literal["completion", "embeddings", "image_generation", "moderation", "audio_transcription"],
    ):
        """
        Runs in parallel to LLM API call
        Runs on only Input

        This can NOT modify the input, only used to reject or accept a call before going to LLM API
        """

        # this works the same as async_pre_call_hook, but just runs in parallel as the LLM API Call
        # In this guardrail, if a user inputs `litellm` we will mask it.
        _messages = data.get("messages")
        if _messages:
            for message in _messages:
                _content = message.get("content")
                if isinstance(_content, str):
                    if "litellm" in _content.lower():
                        raise ValueError("Guardrail failed words - `litellm` detected")

    async def async_post_call_success_hook(
        self,
        data: dict,
        user_api_key_dict: UserAPIKeyAuth,
        response,
    ):
        """
        Runs on response from LLM API call

        It can be used to reject a response

        If a response contains the word "coffee" -> we will raise an exception
        """
        verbose_proxy_logger.debug("async_pre_call_hook response: %s", response)
        if isinstance(response, litellm.ModelResponse):
            for choice in response.choices:
                if isinstance(choice, litellm.Choices):
                    verbose_proxy_logger.debug("async_pre_call_hook choice: %s", choice)
                    if (
                        choice.message.content
                        and isinstance(choice.message.content, str)
                        and "coffee" in choice.message.content
                    ):
                        raise ValueError("Guardrail failed Coffee Detected")

    async def async_post_call_streaming_iterator_hook(
        self,
        user_api_key_dict: UserAPIKeyAuth,
        response: Any,
        request_data: dict,
    ) -> AsyncGenerator[ModelResponseStream, None]:
        """
        Passes the entire stream to the guardrail

        This is useful for guardrails that need to see the entire response, such as PII masking.

        See Aim guardrail implementation for an example - https://github.com/BerriAI/litellm/blob/d0e022cfacb8e9ebc5409bb652059b6fd97b45c0/litellm/proxy/guardrails/guardrail_hooks/aim.py#L168

        Triggered by mode: 'post_call'
        """
        async for item in response:
            yield item

CustomGuardrail methods

Component	Description	Optional	Checked Data	Can Modify Input	Can Modify Output	Can Fail Call
apply_guardrail	Simple method to check and optionally modify text	✅	INPUT or OUTPUT	✅	✅	✅
async_pre_call_hook	A hook that runs before the LLM API call	✅	INPUT	✅	❌	✅
async_moderation_hook	A hook that runs during the LLM API call	✅	INPUT	❌	❌	✅
async_post_call_success_hook	A hook that runs after a successful LLM API call	✅	INPUT, OUTPUT	❌	✅	✅
async_post_call_streaming_iterator_hook	A hook that processes streaming responses	✅	OUTPUT	❌	✅	✅

Frequently Asked Questions

Q. Is apply_guardrail relevant both in the request and in the response (pre_call, during_call and post_call hooks)?

A. Yes, one function works in both - See implementation here

Q. What do I get in the inputs of apply_guardrail? What does each field represent (what is text, language, entities, request_data)?

A. The main one you should care about is 'text' - this is what you'll want to send to your api for verification - See implementation here

**Q. Is this function agnostic to the LLM provider? Meaning does it pass the same values for OpenAI and Anthropic for example?

A. Yes

Q. How do I know if my guardrail is running?

A. If you implement apply_guardrail, you can query the guardrail directly via the /apply_guardrail API.