Integrating Cost Data with Existing Tools

The AI spend page integrates data from browser extensions that track spending outside the router. When a user installs the extension, it observes calls and reports them alongside the router's internal quotes. This creates a comprehensive view of organizational AI expenditure across all entry points.

For enterprise environments using Azure OpenAI or Azure AI Foundry with Entra authentication, the platform supports on-behalf-of credential management. Instead of storing shared keys, the system exchanges identity provider tokens for scoped access tokens, ensuring that only authorized permissions are used for downstream model calls.

Configuration Example: On-Behalf-Of Access

// VS Code configuration showing OBO token scope
{
  "providers": [
    {
      "name": "Azure OpenAI",
      "type": "azure",
      "obo": "https://cognitiveservices.azure.com/.default"
    }
  ]
}

This configuration allows the platform to exchange tokens using the jwt-bearer grant with requested_token_use=on_behalf_of, ensuring secure access without exposing raw API keys.

Comparing Open-Source and Hosted Options

OpenSmartRoute distinguishes itself as an open-source LLM router available as a Python SDK, osr serve image, and hosted platform. Unlike proprietary alternatives that require vendor-specific integration, this solution offers flexibility in deployment while maintaining compatibility with major cloud providers.

The following table compares key features relevant to cost optimization:

Frequently Asked Questions

How does the router know which models to recommend?

The router analyzes task complexity and historical performance data to select models that balance cost and reliability. It uses reverse auction strategies where providers bid on success probability and cost, allowing the system to choose the optimal option for each request.

Can I compare my current spending against optimized costs?

Yes, the /platform/dashboard/spend page displays observed calls and spend alongside the router's quote. This comparison shows the gap between actual expenditure and potential savings through better model selection.

Is the platform suitable for enterprise environments?

The platform supports enterprise features including on-behalf-of authentication, detailed audit logging, and CSV exports for finance teams. It integrates with Azure OpenAI and other major providers while maintaining open-source flexibility.

Next Steps

To begin optimizing your AI spending, visit the /docs/QUICKSTART_PLATFORM page to set up the dashboard and connect your existing billing data.

Nvidia and Samsung back Nous Research with $90 million funding

Nous Research raised $90 million in Series B funding led by Robot Ventures. Nvidia Corp and Samsung Electronics joined the investment round.