BerriAI/litellm · error · ValueError

{label_error.message}

Error message

{label_error.message}

What it means

Before a metric is registered, _valid_metric_labels runs _validate_single_metric_labels for the (metric_name, labels) pair; on mismatch it pretty-prints the invalid labels versus the valid set for that metric and raises ValueError with the generated label_error.message. This is the per-call guard that backs the aggregate config validation, so it typically fires from programmatic registration rather than env parsing.

Source

Thrown at litellm/integrations/prometheus.py:832

                    if self._validate_single_metric_name(metric_name) is None:
                        label_filters[metric_name] = config.include_labels

        return label_filters

    def _validate_configured_metric_labels(self, metric_name: str, labels: list[str]):
        """
        Ensure that all the configured labels are valid for the metric

        Raises ValueError if the metric labels are invalid and pretty prints the error
        """
        label_error: Final = self._validate_single_metric_labels(metric_name, labels)
        if label_error:
            self._pretty_print_invalid_labels_error(
                metric_name=label_error.metric_name,
                invalid_labels=label_error.invalid_labels,
                valid_labels=label_error.valid_labels,
            )
            raise ValueError(label_error.message)

        return True

    #########################################################
    # Pretty print functions
    #########################################################

    def _pretty_print_validation_errors(self, validation_results: ValidationResults) -> None:
        """Pretty print all validation errors using rich"""
        try:
            from rich.console import Console
            from rich.panel import Panel
            from rich.table import Table
            from rich.text import Text

            console: Final = Console()

            # Create error panel title

View on GitHub (pinned to 77b7c6c40c)

Solutions

1. Use the pretty-printed invalid-vs-valid label table printed above the raise to correct the pair
2. Restrict dynamic labels to the intersection with the metric's valid labels before registering
3. Register custom metadata labels globally so they count as valid everywhere
4. Pin the litellm version and re-verify label sets on upgrade

Example fix

# before
labels = ['end_user', 'random_tag']  # random_tag not registered -> ValueError
track_metric('litellm_proxy_total_requests_metric', labels)

# after
valid = callback._validate_single_metric_labels('litellm_proxy_total_requests_metric', labels)
if valid is None:
    labels = [l for l in labels if l in valid_labels_for_metric]  # keep only supported
track_metric('litellm_proxy_total_requests_metric', labels)

err = callback._validate_single_metric_labels(metric_name, candidate_labels)
if err is not None:
    raise ValueError(f'labels {err.invalid_labels} invalid for {err.metric_name}; valid: {err.valid_labels}')

When it happens

Trigger: Calling the label-validation path with e.g. ['end_user'] for a metric that does not support end_user, a tag not enabled via custom_prometheus_metadata_labels / tags, or any label outside _all_defined_labels() for that metric.

Common situations: Custom code building label sets dynamically from request metadata; teams enabling new tags in traffic but not in config; metric label sets shrinking between litellm versions.

Understand the failure class

Background: Schema validation failed / invalid input schema: payload rejected because its shape doesn't match the expected schema — this error's family across 28 libraries.

Related errors

• Configuration validation failed: {all_error_messages}
• Prometheus exclude configuration validation failed: {joined_
• {error.message}
• otel.attributes: include_list and exclude_list are mutually
• otel.attributes: {TOKEN_TYPE_ATTRIBUTE} is a structural toke

AI-assisted analysis of BerriAI/litellm@77b7c6c40c (2026-08-18). Data as JSON: /api/errors/c08808161c58dd9d. Report an issue: GitHub.