Struct AnthropicTokenCounter 

pub struct AnthropicTokenCounter { /* private fields */ }

Token counter for Anthropic Claude models.

Uses cl100k_base tokenizer with a 1.1x approximation factor, since Claude’s actual tokenizer isn’t publicly available. This provides conservative estimates (slightly over-counting).

Context Windows

Model	Context Window
claude-3-5-sonnet	200K
claude-3-opus	200K
claude-3-haiku	200K
claude-2.x	100K

Example

use multi_llm::{AnthropicTokenCounter, TokenCounter};

let counter = AnthropicTokenCounter::new("claude-3-5-sonnet-20241022")?;
let tokens = counter.count_tokens("Hello, world!")?;

Accuracy Note

Token counts are approximate. The 1.1x factor provides a safety margin to avoid accidentally exceeding context limits.

Implementations

impl AnthropicTokenCounter

pub fn new(model: &str) -> LlmResult<Self>

Create token counter for Anthropic Claude models

Trait Implementations

impl Debug for AnthropicTokenCounter

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl TokenCounter for AnthropicTokenCounter

fn count_tokens(&self, text: &str) -> LlmResult<u32>

Count tokens in a text string.

fn count_message_tokens(&self, messages: &[Value]) -> LlmResult<u32>

Count tokens in a list of messages (includes formatting overhead).

fn max_context_tokens(&self) -> u32

Get the maximum context window size for this tokenizer.

fn validate_token_limit(&self, text: &str) -> LlmResult<()>

Validate that text doesn’t exceed the token limit.

fn truncate_to_limit(&self, text: &str, max_tokens: u32) -> LlmResult<String>

Truncate text to fit within a token limit.