README

Cost tracking for Laravel AI agents. Calculates API costs from usage metadata across providers.

Laravel AI Costs provides a zero-config way to calculate API costs for Laravel AI agent responses. Pricing for 2,600+ models is resolved automatically from the LiteLLM community pricing database (cached daily). Pass in a response, usage object, or raw token counts — get back a clean DTO with USD costs.

How Pricing Works

Pricing is resolved in order:

Local config overrides — your published ai-costs.php (if any)
LiteLLM database — 2,600+ models, cached once daily
Exception — if neither source knows the model

Requirements

PHP 8.2+
Laravel 11+ or 12+
laravel/ai ^0

Installation

composer require aaix/laravel-ai-costs

Optionally publish the config to customize model pricing:

php artisan vendor:publish --tag=ai-costs-config

Usage

With the Trait (recommended)

Replace Promptable with TracksAiCost on your agent — it wraps prompt() and tracks costs automatically:

use Aaix\LaravelAiCosts\Concerns\TracksAiCost;
use Laravel\Ai\Contracts\Agent;

class MyAgent implements Agent
{
    use TracksAiCost; // replaces Promptable

    // ...
}

Every prompt() call is tracked transparently:

$agent = MyAgent::make();
$agent->prompt('Analyze this data...');
$agent->prompt('Summarize the results...');

$agent->lastCost();       // AiCostResult for the last prompt
$agent->lastCost()->totalCostUsd;  // 0.000345
$agent->costs();          // array of all AiCostResult objects
$agent->totalCostUsd();   // sum of all prompts
$agent->resetCosts();     // clear tracked costs

Direct Calculator

use Aaix\LaravelAiCosts\Services\AiCostCalculator;

// From a laravel/ai response (model & provider auto-resolved)
$cost = AiCostCalculator::fromResponse($response);

// From a laravel/ai Usage object
$cost = AiCostCalculator::fromUsage($usage, 'claude-sonnet-4-6');

// From raw token counts
$cost = AiCostCalculator::fromTokens(10000, 500, 'deepseek-chat');

AiCostResult DTO

All methods return an AiCostResult readonly DTO:

Property	Type	Description
`inputCostUsd`	`float`	Input token cost in USD
`outputCostUsd`	`float`	Output token cost in USD
`totalCostUsd`	`float`	Combined cost in USD
`inputTokens`	`int`	Number of input tokens
`outputTokens`	`int`	Number of output tokens
`model`	`string`	Model identifier
`provider`	`string`	Auto-detected or explicit provider
`totalCostCents()`	`float`	Combined cost in cents

Configuration

Out of the box, no configuration is needed — pricing comes from LiteLLM automatically.

To override pricing for specific models, publish the config and add entries:

php artisan vendor:publish --tag=ai-costs-config

// config/ai-costs.php
'models' => [
    // Local overrides take precedence over LiteLLM
    'my-custom-model' => ['input' => 1.00, 'output' => 3.00],
    'gpt-4o'          => ['input' => 2.50, 'output' => 10.00], // pin a specific price
],

Model names have dots removed for config key compatibility (e.g. gpt-4.1 becomes gpt-41). The calculator handles this normalization automatically. Prefix matching is supported — my-custom* matches my-custom-model, my-custom-v2, etc.

Cache

LiteLLM pricing is cached for 24 hours by default. You can adjust this via environment variables:

AI_COSTS_CACHE_TTL=86400

To manually refresh:

use Aaix\LaravelAiCosts\Support\LitellmPricingProvider;

LitellmPricingProvider::clearCache();

Testing

php artisan test --filter=AiCostCalculator

Contributing

Contributions are welcome! Please open an issue or submit a pull request.

License

The MIT License (MIT). Please see License File for more information.

aaix / laravel-ai-costs

Maintainers

Package info

Statistics

Security