oraios/serena · error · ValueError

Unknown token count estimator: {self}

Error message

Unknown token count estimator: {self}

What it means

RegisteredTokenCountEstimator._create_estimator() maps each registered estimator enum value to a concrete TokenCountEstimator implementation; an unmatched value raises ValueError. This should be unreachable unless a new enum member was added without a corresponding case in _create_estimator — i.e. an internal consistency bug, not user error.

Source

Thrown at src/serena/analytics.py:109

    CHAR_COUNT = "CHAR_COUNT"

    @classmethod
    def get_valid_names(cls) -> list[str]:
        """
        Get a list of all registered token count estimator names.
        """
        return [estimator.name for estimator in cls]

    def _create_estimator(self) -> TokenCountEstimator:
        match self:
            case RegisteredTokenCountEstimator.TIKTOKEN_GPT4O:
                return TiktokenCountEstimator(model_name="gpt-4o")
            case RegisteredTokenCountEstimator.ANTHROPIC_CLAUDE_SONNET_4:
                return AnthropicTokenCount(model_name="claude-sonnet-4-20250514")
            case RegisteredTokenCountEstimator.CHAR_COUNT:
                return CharCountEstimator(avg_chars_per_token=4)
            case _:
                raise ValueError(f"Unknown token count estimator: {self}")

    def load_estimator(self) -> TokenCountEstimator:
        estimator_instance = _registered_token_estimator_instances_cache.get(self)
        if estimator_instance is None:
            estimator_instance = self._create_estimator()
            _registered_token_estimator_instances_cache[self] = estimator_instance
        return estimator_instance


class ToolUsageStats:
    """
    A class to record and manage tool usage statistics.
    """

    def __init__(self, token_count_estimator: RegisteredTokenCountEstimator = RegisteredTokenCountEstimator.TIKTOKEN_GPT4O):
        self._token_count_estimator = token_count_estimator.load_estimator()
        self._token_estimator_name = token_count_estimator.value
        self._tool_stats: dict[str, ToolUsageStats.Entry] = defaultdict(ToolUsageStats.Entry)

View on GitHub (pinned to 7fcbca7e62)

Solutions

1. Upgrade Serena to a version that implements the requested estimator in _create_estimator().
2. Change the token_count_estimator setting (in serena_config.yml) to a supported value, e.g. tiktoken-gpt-4o, anthropic-claude-sonnet-4, or char-count.
3. If you added the enum member yourself, add a matching 'case' returning a TokenCountEstimator subclass in _create_estimator().

Example fix

// before
case _: raise ValueError(f"Unknown token count estimator: {self}")
// after
case RegisteredTokenCountEstimator.MY_NEW_ESTIMATOR:
    return MyNewEstimator()
case _: raise ValueError(f"Unknown token count estimator: {self}")

def estimator_supported(value, supported: set) -> bool:
    return value in supported

try:
    estimator = registered_estimator.load_estimator()
except ValueError as e:
    log.warning('Falling back to CharCountEstimator: %s', e)
    estimator = CharCountEstimator(avg_chars_per_token=4)

When it happens

Trigger: Calling load_estimator() (directly or via analytics token counting) with a RegisteredTokenCountEstimator value that has no case in _create_estimator — typically after upgrading Serena or adding a custom/registered estimator enum member.

Common situations: A version mismatch where config/analytics references an estimator newly added upstream while running older code; contributor added a RegisteredTokenCountEstimator member but forgot the match arm; corrupted/patched enum value.

Related errors

• Unknown language backend '{backend_str}': valid values are {
• Invalid line_ending: {value!r}. Valid values are: {valid}
• Invalid language server: '{orig_language_str}'.\nValid value
• Unhandled comment normalisation: {comment_normalisation}

AI-assisted analysis of oraios/serena@7fcbca7e62 (2026-08-29). Data as JSON: /api/errors/b712189041c87fae. Report an issue: GitHub.