Add model property to TokenUsage for cost calculation

[Copilot]

Q	A
Bug fix?	no
New feature?	yes
Docs?	no
Issues	Pays into #1565
License	MIT

TokenUsage lacked model information, making it impossible to calculate costs accurately when using multiple models with different pricing (e.g., GPT-4o vs embeddings).

Changes

• TokenUsageInterface/TokenUsage: Added model property and getModel() method
• TokenUsageAggregation: Returns model name if all usages share the same model, null otherwise
• All bridge extractors: Extract and populate model from API responses (OpenAI, Anthropic, Gemini, VertexAI, DeepSeek, Perplexity, Mistral, Ollama, Generic)
• Tests: Added coverage for model property and aggregation behavior
• Examples: Updated print_token_usage() function in examples/bootstrap.php to display the model property in all token usage examples

Usage

$result = $platform->request('gpt-4o', 'Hello');
$tokenUsage = $result->getMetadata()->get('token_usage');

// Model now available for cost calculation
$model = $tokenUsage->getModel(); // "gpt-4o-2024-08-06"

// Calculate costs based on actual model used
$pricing = [
    'gpt-4o' => ['input' => 2.50, 'output' => 10.00],
    'text-embedding-3-small' => ['input' => 0.02, 'output' => 0],
];

foreach ($pricing as $prefix => $prices) {
    if (str_starts_with($model, $prefix)) {
        $cost = ($tokenUsage->getPromptTokens() / 1_000_000 * $prices['input'])
              + ($tokenUsage->getCompletionTokens() / 1_000_000 * $prices['output']);
    }
}

The model parameter is optional (nullable) for backward compatibility.

Q	A
Bug fix?	no
New feature?	yes
Docs?	no
Issues	Fix #1583
License	MIT

Original prompt

---

This section details on the original issue you should resolve

<issue_title>[Platform] Token usage - information about used model</issue_title>
<issue_description>Hi, I’m trying to calculate the total cost of a request. For example, a single request may include two or three calls to the OpenAI GPT-4o model and to the embeddings model.

However, when I look at the token usage in the response:

$tokenUsage = $response->getMetadata()->get('token_usage');

I get an array of TokenUsage objects that does not contain any information about which model was used. Since each model has different pricing, how can I correctly calculate the total cost?

Is there another way to retrieve this information?</issue_description>

Comments on the Issue (you are @copilot in this section)

• Fixes [Platform] Token usage - information about used model #1565

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

[OskarStark]

@copilot can you update the examples

[Copilot]

@copilot can you update the examples

Updated the examples to display the model property in the token usage output. The print_token_usage function in examples/bootstrap.php now shows the model as the first row in the token usage table. (13ba1a7)

[OskarStark]

Add changelog entries to components @copilot

[OskarStark]

@copilot it's for 0.4 for all components

[OskarStark]

@copilot

$tokenUsage = $response->getMetadata()->get('token_usage');

returns a TokenUsageAggregation object, which contains a private array of TokenUsage objects.

Each TokenUsage object should have a model property.

There is also an issue with iterating over the TokenUsage objects, since the array is private.

[OskarStark]

@copilot rebase and make changes if needed

[RVXD]

When will this feature be available? Need token_usage stats.

[RVXD]

Thanks for all the work on this!