Stay in control of your AI routing

Sidebar legend

• Overview – house with upward arrow icon → quick health snapshot.
• Providers – stacked server icon → manage vendor credentials.
• Structured Outputs – layout grid icon → build reusable JSON schemas.
• Workflows – connected nodes icon → orchestrate model routing.
• Playground – play button in a square → run ad-hoc workflow tests.
• Request Logs – clipboard sheet icon → inspect historical calls.
• API Keys – key icon → issue and revoke credentials.

Projects

• Use projects to separate environments or product lines. Each project has its own API keys, provider credentials, workflows, and logs.
• Switch projects from the top navigation. The active project scopes every dashboard view. Projects live under the house icon in the top bar.

API Keys

• Keys follow the mr_live_XXXXXXXX format and are scoped to a project.
• Copy the key the moment you create it—ModelRiver stores only a hashed value.
• Track the Last Used timestamp to identify dormant integrations. Revoke compromised or unused keys instantly.

Providers

• Add provider credentials under Settings → Providers. ModelRiver encrypts them at rest and masks the value in the UI.
• Custom providers let you specify a base URL for self-hosted or private gateways.
• Missing credentials trigger a dashboard warning and cause API requests to fail with a helpful error. Find the stacked server icon in the sidebar to jump straight to providers.

Structured Outputs

• Design reusable JSON schemas to receive consistent responses from LLMs.
• Include descriptions and examples to guide models, and attach structures to workflows with a single click.
• ModelRiver validates responses before returning them to your application. If a provider fails, ModelRiver can synthesise a schema-compliant fallback so downstream systems stay stable. Look for the layout grid icon labelled “Structured Outputs” in the sidebar.

Workflows

• Workflows bundle model choices, fallbacks, structured outputs, and cache fields.
• Cache fields pull values from the request (for example, metadata.segment or messages.0.content) and surface them in logs and responses.
• Workflows are referenced from the API via {"workflow": "name"}, letting you update routing logic without code changes.
• Use the Playground to test changes without touching production logs. Open the connected nodes icon to manage workflows; the play button icon leads to the Playground.

Request Logs

• Every API call is captured with provider/model used, latency, token usage, cached data, and structured output metadata.
• Filter and export logs to debug production issues, power analytics, or reconcile billing.
• Transient dashboard tests run in the Playground do not create persisted logs. Visit the clipboard sheet icon to audit request history.

Stats & Analytics

• The overview tab charts request counts and token consumption for the last 30 days.
• Monitor fallback usage to spot provider instability and adjust workflows accordingly.
• Combine stats with cached fields to understand which customer segments or workflows generate the most volume. The Overview card (house icon) shows charts and quick stats at a glance.

Collaboration tips

• Grant dashboard access to operations and product teams so they can iterate on prompts, workflows, and structures themselves.
• Use consistent naming conventions (for example, team-purpose-environment) to keep workflows manageable at scale.
• Review logs regularly to ensure cached data doesn’t include sensitive information that should be stripped upstream.