aicost

A Rust library to compute LLM usage costs based on the genai crate types.

doc-for-llm.md

Features

• Supports multiple providers (OpenAI, Anthropic, Gemini, DeepSeek, Groq, xAI, Fireworks, Together, Zai).
• Handles normal tokens, cached tokens (read/write), and reasoning tokens.
• Provides granular cost breakdown via the AiCost struct.
• Integrates with genai via ModelIden and Usage.

Usage

use aicost::compute;
use genai::chat::Usage;

// From a genai::chat::Usage object
let usage = Usage {
    prompt_tokens: Some(1000),
    completion_tokens: Some(500),
    ..Default::default()
};

let cost = compute("openai", "gpt-4o", &usage)?;

println!("Total: ${:.6}", cost.total);
println!("Input Total: ${:.6}", cost.input_total);
println!("Output Total: ${:.6}", cost.output_total);

Using ModelIden

If you are using the genai crate, you can compute costs directly using ModelIden.

let cost = aicost::compute_iden(&model_iden, &usage)?;

Structure

The AiCost struct provides a detailed breakdown:

pub struct AiCost {
    pub total: f64,

    pub input_total: f64,
    pub input_normal: f64,
    pub input_cache_read: f64,
    pub input_cache_write: f64,

    pub output_total: f64,
    pub output_normal: f64,
    pub output_reasoning: f64,

    pub input_cache_saving: f64,
}

This Repo