BerriAI/litellm · error · BytezError

{error}

Error message

{error}

What it means

After a successful HTTP response, the Bytez transform checks the JSON body for a top-level 'error' key; if present it raises BytezError with that error stringified and the HTTP status code. Bytez reports application-level failures (model errors, inference crashes, account issues) inside an otherwise 200-ish response, so this is the primary Bytez failure surface.

Source

Thrown at litellm/llms/bytez/chat/transformation.py:197

        self,
        model: str,
        raw_response: httpx.Response,
        model_response: ModelResponse,
        logging_obj: LiteLLMLoggingObj,
        request_data: dict,
        messages: list[AllMessageValues],
        optional_params: dict,
        litellm_params: dict,
        encoding: Any,
        api_key: str | None = None,
        json_mode: bool | None = None,
    ) -> ModelResponse:
        json: Final = raw_response.json()

        error: Final = json.get("error")

        if error is not None:
            raise BytezError(
                message=str(json["error"]),
                status_code=raw_response.status_code,
            )

        # set meta data here
        model_response.created = int(time.time())
        model_response.model = model

        # Add the output
        output: Final = json.get("output")

        message: Final = model_response.choices[0].message

        message.content = output["content"][0]["text"]

        messages = adapt_messages_to_bytez_standard(messages=messages)

        # NOTE We are approximating tokens, to get the true values we will need to update our BE

View on GitHub (pinned to 6c2dcb801b)

Solutions

1. Read the message — it embeds Bytez's own error text, which names the real cause (model not found, billing, inference error).
2. For model issues, verify the model id matches a Bytez-supported repository path and correct it.
3. For quota/billing, top up or switch accounts/keys.
4. For transient inference failures, retry with backoff.

Example fix

# before
litellm.completion(model="bytez/some-wrong/model-id", messages=m)

# after
litellm.completion(model="bytez/meta-llama/Llama-3-8b", messages=m)  # valid Bytez model path

from litellm.llms.bytez.chat.transformation import BytezError

try:
    resp = litellm.completion(model="bytez/...", messages=m)
except BytezError as e:
    msg = str(e)
    if "model" in msg.lower() and "not" in msg.lower():
        fix_model_id_and_retry()
    elif e.status_code in (402, 429):
        handle_billing_or_rate_limit()
    else:
        retry_with_backoff()  # transient inference failures

When it happens

Trigger: Non-streaming litellm.completion against a bytez/ model where the response JSON contains {"error": ...} — e.g. model failed to load, out-of-credits, invalid model path, or inference exception on Bytez's side.

Common situations: Incorrect model repository ids; Bytez account quota exhaustion; transient model cold-start or GPU failures; content the hosted model rejects internally.

Related errors

• Failed to transform Braintrust response: {str(e)}
• APIError: {exception_provider} - {error_str}
• LiteLLM: provider returned a response with no 'choices'. Raw
• {completion_response["error"]}
• Azure AI Speech transcription failed with RecognitionStatus=

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/31fb585882335b86. Report an issue: GitHub.