What causes "Failed to parse Clarifai response: {e}" in BerriAI/litellm?

Verify Clarifai credentials (`CLARIFAI_API_KEY` / PAT) and that the model/user/app IDs in the model string are correct.

BerriAI/litellm · error · OpenAIError

Failed to parse Clarifai response: {e}

Error message

Failed to parse Clarifai response: {e}

What it means

Raised as `OpenAIError` (preserving upstream status and headers) when the Clarifai response body cannot be parsed as JSON. The adapter calls `raw_response.json()` inside try/except; failure means the body is not JSON — typically an error page, empty body, or gateway output — even though an HTTP response was received.

Source

Thrown at litellm/llms/clarifai/chat/transformation.py:107

        api_key: str | None = None,
        json_mode: bool | None = None,
    ) -> ModelResponse:
        """
        Transform the Clarifai response to a standard ModelResponse.
        Since Clarifai is OpenAI-compatible, we use OpenAI response transformation.
        """
        ## Logging
        logging_obj.post_call(
            input=messages,
            api_key=api_key,
            original_response=raw_response.text,
            additional_args={"complete_input_dict": request_data},
        )
        ## Reponse
        try:
            completion_response: Final = raw_response.json()
        except Exception as e:
            raise OpenAIError(
                status_code=raw_response.status_code,
                message=f"Failed to parse Clarifai response: {e}",
                headers=raw_response.headers,
            ) from e

        response: Final = ModelResponse(**completion_response)

        if response.model is not None:
            response.model = "clarifai/" + model

        return response

    def get_error_class(self, error_message: str, status_code: int, headers: dict | httpx.Headers) -> BaseLLMException:
        """
        Get the appropriate error class for Clarifai errors.
        Since Clarifai is OpenAI-compatible, we use OpenAI error handling.
        """
        return OpenAIError(

View on GitHub (pinned to 6c2dcb801b)

Solutions

Verify Clarifai credentials (`CLARIFAI_API_KEY` / PAT) and that the model/user/app IDs in the model string are correct.
Log `raw_response.status_code` and the first bytes of the body — the HTML/status reveals whether it is auth, proxy, or Clarifai-side.
Retry on 5xx/gateway errors; treat 4xx as configuration problems.
Check Clarifai status and your proxy/gateway config if HTML error pages appear.

Defensive patterns

Strategy: try-catch

Validate before calling

import os

def clarifai_config_ready() -> bool:
    return bool(os.environ.get("CLARIFAI_API_KEY"))

Try / catch

try:
    resp = litellm.completion(model="clarifai/<user>/<app>/<model>", messages=msgs)
except OpenAIError as e:
    log.error("clarifai parse failure status=%s msg=%.300s", e.status_code, e.message)
    if e.status_code and e.status_code >= 500:
        backoff_and_retry()
    else:
        verify_pat_and_model_string()  # 4xx: config problem

Prevention

Validate the Clarifai PAT and the user/app/model string format before calls.
Log response status and body prefix when this fires to distinguish proxy HTML from Clarifai errors.
Ensure direct (non-intercepting) egress to api.clarifai.com.

When it happens

Trigger: Calling `litellm.completion(model="clarifai/...")` where Clarifai's API returns HTML (auth portal, block page), an empty 200, or a truncated body from a gateway timeout (502/504 HTML from a proxy in front of Clarifai).

Common situations: Invalid/expired Clarifai PAT causing a redirect to a login page; reverse proxies (nginx, Cloudflare) returning HTML error pages on backend failure; region-blocked endpoints; response truncation on slow links.

Understand the failure class

Parsing and encoding errors: unexpected token, malformed input — why parsers reject input and how to find the real culprit.

Related errors

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/dd23b0f11f115db4. Report an issue: GitHub.