BerriAI/litellm · error · OCIError

raw_response.text

Error message

raw_response.text

What it means

After the HTTP request to the OCI embedText endpoint returns, the adapter checks the status code. Any non-200 response is surfaced as an OCIError whose message is the raw response body (`raw_response.text`). This is a pass-through of an upstream OCI API error, not a client-side validation failure.

Source

Thrown at litellm/llms/oci/embed/transformation.py:260

            inputType=input_type,
            truncate=optional_params.get("truncate", "END"),
            outputDimensions=optional_params.get("outputDimensions"),
        )
        return request.model_dump(exclude_none=True)

    def transform_embedding_response(
        self,
        model: str,
        raw_response: httpx.Response,
        model_response: EmbeddingResponse,
        logging_obj: LiteLLMLoggingObj,
        api_key: str | None,
        request_data: dict,
        optional_params: dict,
        litellm_params: dict,
    ) -> EmbeddingResponse:
        if raw_response.status_code != 200:
            raise OCIError(
                status_code=raw_response.status_code,
                message=raw_response.text,
            )

        try:
            json_response: Final = raw_response.json()
        except Exception as e:
            raise OCIError(
                status_code=raw_response.status_code,
                message=f"Failed to parse OCI embed response as JSON: {e}",
            )

        try:
            parsed: Final = OCIEmbedResponse(**json_response)
        except Exception as e:
            raise OCIError(
                status_code=500,
                message=f"OCI embed response does not match expected schema: {e}",

View on GitHub (pinned to 6c2dcb801b)

Solutions

1. Read the message body — it contains OCI's own error details (e.g. 'NotAuthorizedOrNotFound', 'TooManyRequests') which pinpoint the fix.
2. For 401/403: verify OCI_COMPARTMENT_ID, your user/API key, and that the tenancy has access to the model.
3. For 429: add backoff/retry or reduce batch frequency.
4. For 404: confirm the modelId and region in the endpoint URL match a model your compartment can serve.

from litellm.exceptions import litellm as _l  # generic
try:
    resp = litellm.embedding(model="oci/...", input=texts)
except Exception as e:
    code = getattr(e, "status_code", None)
    if code == 429:
        backoff_and_retry()  # rate limited
    elif code in (401, 403):
        audit_oci_credentials()  # config problem, do not retry
    else:
        raise

When it happens

Trigger: Invalid OCI credentials (401), wrong compartment OCID (404), model not entitled to your tenancy (403), rate limiting (429), or OCI service errors (5xx) — any non-200 from the embedText call. The status_code on the raised OCIError mirrors the upstream status.

Common situations: Expired or misconfigured OCI config file / API keys, using an ON_DEMAND model the tenancy has not accepted terms for, wrong region in the api_base, or throttling during bulk embedding runs.

Related errors

• {str(e)}
• {err.response.text}
• error_msg (upstream response error message)
• {e.response.text}
• {response.text}

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/ff87bbda70688c77. Report an issue: GitHub.