BerriAI/litellm · error · DashScopeError

response_json.get("message", str(response_json))

Error message

response_json.get("message", str(response_json))

What it means

DashScope's rerank API signals failures with a top-level envelope {"code": ..., "message": ..., "request_id": ...} and no "results". LiteLLM detects that shape and re-raises DashScopeError with the upstream message (str(response_json) as fallback) and the HTTP status — the text you see is DashScope's own error description.

Solutions

1. Act on the upstream message text: InvalidApiKey -> fix/rotate the key; Throttling -> slow down or raise quota; NotFound/model errors -> use a rerank-capable model (gte-rerank / qwen3-rerank families)
2. Keep the request_id from your logs when escalating to DashScope support
3. For throttling, add client-side rate limiting and retry with backoff honoring Retry-After when present

try:
    out = litellm.rerank(model=m, query=q, documents=docs)
except Exception as e:
    msg = str(e)
    status = getattr(e, 'status_code', None)
    if status == 401 or 'InvalidApiKey' in msg:
        rotate_credentials()
    elif status == 429 or 'Throttl' in msg:
        backoff_and_retry()
    else:
        raise

When it happens

Trigger: Invalid API key (401 envelope); throttling/QPS limits ('Throttling' code); requesting a model that does not support the rerank endpoint (code like 'Algorithm.NotFound'); unsupported parameters in the rerank body.

Common situations: Using an embedding/chat model name on the rerank route; cn-region credentials against the intl endpoint; bursty background jobs hitting per-account QPS; newly rotated key not yet deployed.

Related errors

• DashScope API key is required. Set 'DASHSCOPE_API_KEY' env…
• {error.message}
• DASHSCOPE_API_KEY is not set
• documents is required for DashScope rerank
• err.response.text

AI-assisted analysis of BerriAI/litellm@77b7c6c40c (2026-08-18). Data as JSON: /api/errors/d3d5070ebbea55d0. Report an issue: GitHub.

Appendix: source

Thrown at litellm/llms/dashscope/rerank/transformation.py:187

        litellm_params = litellm_params or {}
        try:
            response_json: Final = raw_response.json()
        except Exception:
            raise DashScopeError(
                status_code=raw_response.status_code,
                message=raw_response.text,
            )

        logging_obj.post_call(
            input=request_data.get("query"),
            api_key=api_key,
            additional_args={"complete_input_dict": request_data},
            original_response=response_json,
        )

        # DashScope error envelope: {"code": "...", "message": "...", "request_id": "..."}
        if "code" in response_json and "results" not in response_json:
            raise DashScopeError(
                status_code=raw_response.status_code,
                message=response_json.get("message", str(response_json)),
            )

        results: Final = response_json.get("results")
        if results is None:
            raise DashScopeError(
                status_code=raw_response.status_code,
                message=f"No results in DashScope rerank response: {response_json}",
            )

        # qwen3-rerank returns:
        #   {"index": int, "relevance_score": float}
        # plus, when return_documents=true was sent:
        #   "document": {"text": "..."}
        # which already matches LiteLLM's RerankResponseDocument shape.
        transformed_results: Final[list[dict]] = []
        for r in results:

View on GitHub (pinned to 77b7c6c40c)