What causes "Cannot import cohere package, please `pip install cohere`." in run-llama/llama_index?

Install the SDK: pip install cohere (and pin it in your dependency file).

run-llama/llama_index · error · ImportError

Cannot import cohere package, please `pip install cohere`.

Error message

Cannot import cohere package, please `pip install cohere`.

What it means

Raised by CohereRerankRelevancyMetric.__init__ when it tries `from cohere import Client` and the optional cohere package is not installed. llama-index-core declares cohere as an optional dependency (note the `# pants: no-infer-dep` marker), so the rerank relevancy metric fails at construction time on any environment without it.

Source

Thrown at llama-index-core/llama_index/core/evaluation/retrieval/metrics.py:453

    _client: Any = PrivateAttr()

    def __init__(
        self,
        model: str = "rerank-english-v2.0",
        api_key: Optional[str] = None,
    ):
        try:
            api_key = api_key or os.environ["COHERE_API_KEY"]
        except IndexError:
            raise ValueError(
                "Must pass in cohere api key or "
                "specify via COHERE_API_KEY environment variable "
            )
        try:
            from cohere import Client  # pants: no-infer-dep
        except ImportError:
            raise ImportError(
                "Cannot import cohere package, please `pip install cohere`."
            )

        super().__init__(model=model)
        self._client = Client(api_key=api_key)

    def _get_agg_func(self, agg: Literal["max", "median", "mean"]) -> Callable:
        """Get agg func."""
        return _AGG_FUNC[agg]

    def compute(
        self,
        query: Optional[str] = None,
        expected_ids: Optional[List[str]] = None,
        retrieved_ids: Optional[List[str]] = None,
        expected_texts: Optional[List[str]] = None,
        retrieved_texts: Optional[List[str]] = None,
        agg: Literal["max", "median", "mean"] = "max",

View on GitHub (pinned to afd0fef371)

Solutions

Install the SDK: pip install cohere (and pin it in your dependency file).
If you do not need Cohere reranking, drop 'cohere_rerank_relevancy' from the metrics list and use the built-in hits/mrr/precision/recall metrics instead.
Gate the metric behind an availability check (importlib.util.find_spec('cohere')) so evaluation configs degrade gracefully on environments without the package.

Example fix

# before
metric = CohereRerankRelevancyMetric()  # ModuleNotFoundError -> wrapped ImportError

# after
pip install cohere
metric = CohereRerankRelevancyMetric()

Defensive patterns

Strategy: validation

Validate before calling

import importlib.util

if importlib.util.find_spec("cohere") is None:
    raise RuntimeError("cohere package required for rerank relevancy: pip install cohere")

Try / catch

try:
    metric = CohereRerankRelevancyMetric(api_key=key)
except ImportError as e:
    if "pip install cohere" in str(e):
        log.warning("cohere not installed; skipping rerank relevancy metric")
        metric = None
    else:
        raise

Prevention

Add cohere to your lockfile whenever the cohere_rerank_relevancy metric is registered.
Gate optional-dependency features behind find_spec checks in eval harnesses.
Run dependency smoke tests in CI that construct every configured metric.

When it happens

Trigger: Constructing CohereRerankRelevancyMetric (directly or through RetrieverEvaluator with the 'cohere_rerank_relevancy' metric) on an environment where `pip install cohere` was never run — typically a slim Docker image or a fresh venv holding only llama-index-core.

Common situations: Adding cohere-rerank evaluation to an existing project without updating requirements.txt; CI images that cache old dependency locks; upgrading llama-index and losing a transitive cohere install that used to come from another integration package.

Related errors

AI-assisted analysis of run-llama/llama_index@afd0fef371 (2026-08-15). Data as JSON: /api/errors/b700b04a3c072417. Report an issue: GitHub.