feature: Add extra Vllm dashboard production oriented metrics

Describe the feature

This on my mind since last fall, but here it is.

Current gap:

The default vLLM dashboard covers basic metrics but lacks production-grade observability for latency percentiles, scheduler state, and time-to-first-token (TTFT).

Proposed solution:

Integrate (merge or just add separately) with the OCI AI Blueprint dashboard (Apache 2/UPL licensed) which includes:

• End-to-end request latency (P99/P95)
• Scheduler state visualization
• TTFT latency panels
• Model name variable filtering for multi-model deployments

See print screens for previews

Dashboard Source:

View the dashboard from the OCI AI Blueprint project here >> vLLM dashboard

Note

I will be including this as extra dashboard in my upcoming vllm-prod-stack Terraform tutorial on a new cloud . It would be great to have it upstream instead of a custom add-on.

Why do you need this feature?

Use case:

Production deployments need these metrics for SLO monitoring and autoscaling decisions. Currently maintaining this as a separate ConfigMap in our Terraform stack.

Additional context

screenshots

Start by reviewing the referenced OCI AI Blueprint vLLM dashboard JSON and the repository’s existing dashboard ConfigMap and Terraform integration. Determine where an additional production-oriented dashboard belongs and how its latency, scheduler, TTFT, and model-filtering panels should be validated. Done means the dashboard is integrated upstream and supports the proposed production metrics without requiring the separate custom add-on.