BerriAI/litellm · warning · HTTPException

Violated guardrail policy

Error message

Violated guardrail policy

What it means

Parameter validation for DALL-E 3 image generation: same mechanism as other image models - each non-default param must be in get_supported_openai_params(model) or the call raises ValueError (listing supported params) unless drop_params=True. DALL-E 3 supports quality and style but rejects DALL-E 2-era or GPT-image-era params (e.g. response_format restrictions, background, moderation) that are not in its supported set.

Source

Thrown at enterprise/enterprise_hooks/aporia_ai.py:130

        """

        response = await self.async_handler.post(
            url=self.aporia_api_base + "/validate",
            data=_json_data,
            headers={
                "X-APORIA-API-KEY": self.aporia_api_key,
                "Content-Type": "application/json",
            },
        )
        verbose_proxy_logger.debug("Aporia AI response: %s", response.text)
        if response.status_code == 200:
            # check if the response was flagged
            _json_response = response.json()
            action: str = _json_response.get(
                "action"
            )  # possible values are modify, passthrough, block, rephrase
            if action == "block":
                raise HTTPException(
                    status_code=400,
                    detail={
                        "error": "Violated guardrail policy",
                        "aporia_ai_response": _json_response,
                    },
                )

    async def async_post_call_success_hook(
        self,
        data: dict,
        user_api_key_dict: UserAPIKeyAuth,
        response,
    ):
        from litellm.proxy.common_utils.callback_utils import (
            add_guardrail_to_applied_guardrails_header,
        )

        """

View on GitHub (pinned to 6c2dcb801b)

Solutions

1. Trim the call to only the params listed in the error message (dall-e-3: prompt, n, quality, size, style, user).
2. Set drop_params=True to auto-drop unsupported keys.
3. Switch to gpt-image-1 if you need the newer parameter set (background, output_format, moderation).
4. Check get_supported_openai_params('dall-e-3') at startup and validate your param dict against it.

Example fix

# before
litellm.image_generation(model="dall-e-3", prompt="cat", output_format="png", background="transparent")

# after
litellm.image_generation(model="dall-e-3", prompt="cat", quality="hd", style="natural")
# or: litellm.drop_params = True

DALLE3_SUPPORTED = {"prompt", "n", "quality", "size", "style", "user"}

def validate_dalle3_params(params: dict) -> list[str]:
    return [k for k in params if k not in DALLE3_SUPPORTED and k != "prompt"]  # non-empty => will raise

def param_set_is_supported(params: dict, supported: set[str]) -> bool:
    return set(params).issubset(supported)

try:
    img = litellm.image_generation(model="dall-e-3", prompt=p, **params)
except ValueError as e:
    if "not supported" in str(e):
        params = {k: v for k, v in params.items() if k in DALLE3_SUPPORTED}
        img = litellm.image_generation(model="dall-e-3", prompt=p, **params)
    else:
        raise

When it happens

Trigger: Calling litellm.image_generation(model='dall-e-3', ...) with unsupported params such as response_format='b64_json' on endpoints that disallow it, background, output_format, or any param outside dall-e-3's supported list, without drop_params=True.

Common situations: Generic image wrappers forwarding every kwarg; migrating dall-e-2 code that relied on response_format; new gpt-image-1 params copy-pasted into dall-e-3 calls; router-level default params applied model-agnostically.

Related errors

• Unknown hook: {hook_name}. Available hooks: {list(ENTERPRISE
• `banned_keywords_list` can either be a list or filepath. Non
• Unclassified keys in {PRICES_PATH.name}: {', '.join(unclassi
• No Braintrust API token provided. Pass via Authorization hea
• Braintrust API error: {e.response.text}

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/f79a4f0dcc424f22. Report an issue: GitHub.