Dashboard

Spectra includes a built-in single-page application (SPA) dashboard for exploring AI request data, analyzing costs, managing the pricing catalog, and viewing per-user analytics. The dashboard is served directly from your Laravel application and requires no separate frontend build or deployment.

Accessing the Dashboard

By default, the dashboard is available at /spectra on your application's domain. The path, domain, and middleware are all configurable in config/spectra.php:

'dashboard' => [
    'domain' => null,                    // Optional subdomain (e.g., 'spectra')
    'enabled' => true,
    'path' => env('SPECTRA_PATH', 'spectra'),
    'middleware' => ['web'],             // Add 'auth' for authentication
    'layout' => 'full',                  // Display mode
    'date_format' => 'M j, Y g:i:s A',
],

Authorization

Dashboard access is controlled by the viewSpectra gate. By default, only the local environment is permitted. To allow access in production or staging environments, define the gate in a service provider:

Gate::define('viewSpectra', function ($user) {
    return $user->isAdmin();
});

If your application provides Spectra security via another method, such as IP restrictions or a VPN, you may need to change the closure signature to function ($user = null) to allow unauthenticated access to the gate check.

Screens

Requests

The main request explorer displays all tracked AI requests in a paginated, filterable list. Each row shows the provider with its display name and logo, the model used, token usage, calculated cost, latency, and HTTP status code. You can filter the list by provider, model, HTTP status, custom tags, trace ID, and date range. This screen is the primary interface for investigating individual AI operations and understanding usage patterns.

Request Detail

Clicking any request in the explorer opens the detail view, which provides a comprehensive breakdown of the request:

• Overview — Provider, model, model type, token counts, cost, latency, and status at a glance.
• Request payload — The full JSON body sent to the AI provider, including messages, parameters, and tool definitions.
• Response payload — The full JSON response with the AI's generated output.
• Code frames — The file and line number where the request originated in your application code, making it easy to trace an AI call back to its source.
• Tags and metadata — Any custom tags or context attached to the request.
• Media — Generated images or videos, if media persistence is enabled.

Cost Analysis

The cost analysis screen provides a breakdown of AI spending across providers, models, and time periods. You can view total spend by provider, cost per model, and spending trends over time. This screen helps you identify which models and features are driving the most cost and spot unexpected spending patterns.

Cost Comparison

The cost comparison tool lets you evaluate "what-if" scenarios by comparing your current model usage against alternative models. You can see what you are spending now, what you would spend with a different model at the same usage levels, and identify potential savings by switching models. This is particularly useful for optimizing costs across your AI operations.

Pricing Management

The pricing management screen lets you view and edit the pricing catalog directly in the dashboard. You can browse all providers and their models, view per-tier prices (standard, batch, flex, priority), edit prices inline, add new models to a provider, delete models you no longer need, and clear the pricing cache. Changes take effect immediately without requiring a deployment.

Providers

The providers screen shows all configured providers with their associated models, model snapshots, and operational status.

Trackables

The trackables screen provides per-user or per-entity analytics. You can see which users are making the most requests, view cost breakdowns per user, and drill into a specific user's request history. This is valuable for understanding usage distribution across your user base and identifying heavy consumers.

Layout Modes

The dashboard.layout configuration option controls which model types are displayed in the dashboard. This is useful if your application only uses a subset of model types and you want a focused view:

Layout	Shows
full	All model types with a type distribution chart
text	Text completions only
embedding	Embeddings only
image	Image generation only
video	Video generation only
audio	TTS and STT metrics

API Endpoints

All dashboard data is served through JSON API endpoints under /{spectra-path}/api/. These endpoints are protected by the same viewSpectra gate and can be used to build custom dashboards or integrate Spectra data into other tools.

Analytics

Method	Endpoint	Description
GET	/api/config	Dashboard configuration and feature flags
GET	/api/stats	Aggregated usage statistics across all dimensions
GET	/api/costs	Cost breakdown by provider, model, and time period
GET	/api/cost-comparison	Model cost comparison data for what-if analysis

Requests

Method	Endpoint	Description
GET	/api/requests	Paginated request list with full filter support
GET	/api/requests/{id}	Single request detail with payloads and metadata
GET	/api/requests/{id}/video	Download a generated video file
GET	/api/tags	All available tags for filter dropdowns

Trackables

Method	Endpoint	Description
GET	/api/trackables	List all tracked users and entities with usage summaries
GET	/api/trackables/view/{id}	Detailed analytics for a specific entity

Providers and Models

Method	Endpoint	Description
GET	/api/providers	Provider list with logos and metadata
GET	/api/models	Model list grouped by provider and type

Pricing Management

Method	Endpoint	Description
GET	/api/pricing	Full pricing catalog with all tiers
POST	/api/pricing/models/{id}	Update pricing for a specific model
PATCH	/api/pricing/models/{id}	Update model metadata
DELETE	/api/pricing/models/{id}	Delete a model from the catalog
POST	/api/pricing/providers/{name}/models	Add a new model to a provider
POST	/api/pricing/cache/clear	Clear the pricing cache

Disabling the Dashboard

To run Spectra in headless mode — tracking requests and calculating costs without serving the dashboard UI — disable it in the configuration:

'dashboard' => [
    'enabled' => false,
],

Or via environment variable:

SPECTRA_DASHBOARD_ENABLED=false

When the dashboard is disabled, all tracking, cost calculation, and API functionality continues to work. Only the web routes and UI assets are not registered.