voyageai_tokenizer

Suggest edits

Bases: BaseTokenizer

Source Code in griptape/tokenizers/voyageai_tokenizer.py
@define()
class VoyageAiTokenizer(BaseTokenizer):
    MODEL_PREFIXES_TO_MAX_INPUT_TOKENS = {
        "voyage-3": 32000,
        "voyage-large-2": 16000,
        "voyage-code-2": 16000,
        "voyage-2": 4000,
        "voyage-lite-02-instruct": 4000,
    }
    MODEL_PREFIXES_TO_MAX_OUTPUT_TOKENS = {"voyage": 0}

    api_key: Optional[str] = field(default=None, kw_only=True, metadata={"serializable": False})
    client: Client = field(
        default=Factory(
            lambda self: import_optional_dependency("voyageai").Client(api_key=self.api_key),
            takes_self=True,
        ),
        kw_only=True,
    )

    def count_tokens(self, text: str) -> int:
        return self.client.count_tokens([text], model=self.model)

  • MODEL_PREFIXES_TO_MAX_INPUT_TOKENS = {'voyage-3': 32000, 'voyage-large-2': 16000, 'voyage-code-2': 16000, 'voyage-2': 4000, 'voyage-lite-02-instruct': 4000} class-attribute instance-attribute

  • MODEL_PREFIXES_TO_MAX_OUTPUT_TOKENS = {'voyage': 0} class-attribute instance-attribute

  • api_key = field(default=None, kw_only=True, metadata={'serializable': False}) class-attribute instance-attribute

  • client = field(default=Factory(lambda self: import_optional_dependency('voyageai').Client(api_key=self.api_key), takes_self=True), kw_only=True) class-attribute instance-attribute

count_tokens(text)

Source Code in griptape/tokenizers/voyageai_tokenizer.py
def count_tokens(self, text: str) -> int:
    return self.client.count_tokens([text], model=self.model)
← Prev

simple_tokenizer

↑ Up

tokenizers

Next →

tools

Could this page be better? Report a problem or suggest an addition!