Try Gemini 1.5 models, the latest multimodal models in Vertex AI, and see what you can build with up to a 2M token context window. Try Gemini 1.5 models, the latest multimodal models in Vertex AI, and see what you can build with up to a 2M token context window.

Package cloud.google.com/go/vertexai/genai/tokenizer (v0.13.1)

Package tokenizer provides local token counting for Gemini models. This tokenizer downloads its model from the web, but otherwise doesn't require an API call for every [CountTokens] invocation.

CountTokensResponse

type CountTokensResponse struct {
	TotalTokens int32
}

CountTokensResponse is the response of [Tokenizer.CountTokens].

Tokenizer

type Tokenizer struct {
	// contains filtered or unexported fields
}

Tokenizer is a local tokenizer for text.

func New

func New(modelName string) (*Tokenizer, error)

New creates a new [Tokenizer] from a model name; the model name is the same as you would pass to a [genai.Client.GenerativeModel].

func (*Tokenizer) CountTokens

func (tok *Tokenizer) CountTokens(parts ...genai.Part) (*CountTokensResponse, error)

CountTokens counts the tokens in all the given parts and returns their sum. Only [genai.Text] parts are suppored; an error will be returned if non-text parts are provided.

Example

package main

import (
	"fmt"
	"log"

	"cloud.google.com/go/vertexai/genai"
	"cloud.google.com/go/vertexai/genai/tokenizer"
)

func main() {
	tok, err := tokenizer.New("gemini-1.5-flash")
	if err != nil {
		log.Fatal(err)
	}

	ntoks, err := tok.CountTokens(genai.Text("a prompt"), genai.Text("another prompt"))
	if err != nil {
		log.Fatal(err)
	}

	fmt.Println("total token count:", ntoks.TotalTokens)

}